Winding Drums The analysis and design of fabricated steel cylindrical drums for mine winding engines By L. T. J. Atkinson & G. L. Taylor* ‘VEEC (rome Exim Ley Fane ta Gino BG ‘CEG. (esas Enpoerng) Li, Finer ane Ghalmoa Biaiow Reprinted by kind permission of the Publishers of "Colliery Engineering" fron e series of articles published between December 1966 and August 1967. GEC Mechanical Handling Limited BIRCH WALK ERITH KENT ENGLAND Telephone Errh 36823 + Teleprems, Cables Propac, Dentors Telex 263237_ Winding Drums The analysis and design of fab- ricated steel cylindrical drums for mine winding engines. By L.T.J. ATKINSON & G.L. TAYLOR ' 1. Introduction Is sms series of articles it is proposed 10 deal with two main aspects of winding engine drum design in which the accent is upon cylindrical drum winders as opposed \w friction winders, the latier needing a completely different theoretical analysis from the former. The bulk of the early work deals with the mathemati al analysis of shells in which. out of necessity, the computer plays # big purt: this forms the basis of the first_muin aspect of design. In the final stages. the ‘more practical aspect is dealt with in as much as certain physical structures are analysed in which the weak points of ‘design are outlined and remedies are suggested. During the course of these articles the reader may find that theoretical analysis cannot be discussed without certain points of practical design being introduced and vice-versa; this is, however, unavoidable and no apologies can he given for some degree of repetition as this is necessary: jn order to correlate certain features of design. 1.1. The basic problem To the layman, and indeed to the uninitiated designer in this field of ‘engineering, a winding engine drum ‘appears to be a simple piece of engineering construction epparently requiring no special knowledge for both its conception and its design. This could not be further from the truth as this series will ultimately show. It has been heard suid that winding engines are so much “old iron” but one must suppose. however, that, should the uninitiated be allowed a glimpse at the insides of some of the older drums. resplendent in their display of com= Pression rings made from fiat bottom tail’ securely riveted 10 the shell plates. then there may be some excuse for such remarks, There is nothing wrong in using rails for compression rings provided it is used in the cortect fashion and that it is present in the right Quantity as the reader will luter appreciate when com- Pression rings are being discussed. Modern winding engine drums (see Fig. 1) should give no excuse for such derogatory remarks (at least it is hoped that this is so) because they should have been Fig. 1. & modern winding engine under constuction. This it » ouble-drum, single-clutch machine for which the ums ore ‘manufactured entirely from ralé stee! plete scientifically designed and carefully manufactured with 8 much care as is justifiably possible. In this respect, ukhough great care is normally taken to ensure sound designs, too little knowledge has been available to wind= ing engine designers in the past to enable them to predict with greater certainty whether or not the finished product is going to withstand the duty for which it is required. Jt is relatively easy 10 overdesign, to be on the “safe side".'although this does not always solve the problem as one is often confronted with side effects and, in these modern times, one of the worst of these side effects is the economy of manufacture. This is basically where the problem starts, as indeed is the case in most com: petitive industries, but, here we are faced with the human element in that lives may be dependent upon whether of not the safe functioning of equipment can be assured The continual battle for economy must be weighed against sefety and the latter must. of course, win every time but it will be shown that, the manufacturing economy can account for itself quite effectively if greoier fatiention is paid towards obtaining lighter and less bulky structures providing ulways. of course. that such re- sulting structures are proved adequate for their appointed duty. Faced with the circumstances outlined above. many Gesigners will be presented with the problem of how such economies can be achieved and one answer is. of course. t0 fabricaie their drums rather than adhere to the ever fuithful castings which can be more expensive and sometimes very troublesome to manufacture. The Beneral trend these days is towards fabrication because4 Fig. 2. Some of the forces acting on 2 paraile! winding drumForces (¢). (e). (g). (h) and (j) are not shown iis often very much quicker to obtain # welded structure than it is 10 obtain patterns and subsequent castings. But whut if our designer is working for a company who ere currenily in the habit of manufacturing all drums from welded steel plate? He is presumably faced with an even bigger problem but this need not be so if attention 's paid lo-design detail along the lines suggested by these atticles, 1.2 Finding the loads - Before any unalysis of the stresses in a structure can he undertuhen. it is fist necessary to know what loading configuration is being applied to that structure. This may appent very obvious bul if one considers the situation, av )ied to 8 winder drum, it will be appreciated that although it 3s necessary to know the loading configura- tion how is it 0 be obtained? There are main types of Joad which can be con- sidered, namely:— (4) Tangential pull of over and/or underlay ropes. (b) Torsional load applied via main driving motor. tc) Centrifugal forces due to rotation, (a) Crushing of shell due to winding on ropes under tension. (2) Forces due to thermal expansion of brake path. (1) Forces produced as brakes are applied, In addition to these there are also: (g) Forces produced due to dead weight. th) Forces produced due to drumshaft deflection by virtue of (p). (i) Auxiliary dynamic loads from various sources. Fig. 2 gives some idea of the directions of the loads shich can be encountered. ‘DF all these only (A), (e) and (1) have any prime signi- ‘cance but only (2) is being considered for the time veing although mention will be made of the remainder ts these articles progress. Why choose only (d)? Let a nodestly small drum of 100in diameter by 100in Jong be considered having wound on to it a lin diameter rope inder a constant tension of 10,000 Ib: this will produce vm the shell @ gross impressed load of 6.280.000 1b of ‘ome 3,000 tons, Compared with this the other forces ex.) for (e)) are of no particular significance when rnalysing the shell Having now established that the crushing forces are ‘portant it is necessary to define the magnitude and, vithout doubt, this has been the greatest single cause of design controversy for very many years. Many readers will undoubtedly be familiar with that excellent text book “Electric Winders” by H. H. Brough. ton which was first published in 1928 and subsequently revised in 1948 (see references). This book is 38 years cld and it is surprising to see how litle change there has been in the approach towards designing winders. On page 273 of the 1948 revision Broughton discusses the “Strength of Drums” and his opening paragraph states that, (quote) “Except in special cases, if a drum de designed to have strength sufficient to resist the crushing action of the rope coiled upon its surface, it will also be strong enough to resist bending”. (unquote). Unfortu- nately, no mention is made as to which plane of bending is referred to but it is taken to mean bending in a plane parallel to, and passing through, the axis of the drum, or is it known what is implied by “special cases” for instances will be given later of “normal” drums which, although adequate from 2 compressive point of view, can suffer very high bending stresses in the axial plane, During the past few years, many drum failures have ‘been investigated and it has been suggested that the designers of some of the older drums have utilised the results obtained many years ago by E. O. Waters. In ‘one experiment carried out by Waters, and outlined by Broughton, it is significant to note that a rope of only in diameter was used, it being wound on to a 10in diameter drum under 2 ‘constant tension of 1,000 Ib. AS each successive layet was wound on the tensions in each previous layer were, in some way, measured until after 4 layers had been completed the ‘compressive stress in the Shell was only about 54% of what it would have been if the tensions in the ropes of each layer had remained at 1,000 1b, The figures given inthe text book do, of course, imply a load reduction in the under layers as new layers are wound on but one must be very wary of using such results when designing full sized machine as some drum failures have probably indicated already.* In connection with the above results Waters may have failed 10 appreciate that, although his experiments may have been technically correct, the rope which was used bore no resemblance whatsoever to the constructional features of a true winding rope, No other data exists in Broughton’s text to indicate the thickness of the shell its length, the material from which it was made or the DeraFig 3. Although the South Atvican winder nas drums manufactured fom cast steel. same ideo can be erned of the size of the installation which is typical of that county manner in which its ends were supported. All these can alfect the final load upon the shell and probably the ‘most important was the rope which was used. or the method used 10 measure the tensions. Waters’ experiments were a starting point and were ‘aimed at giving designers something more tangible with which to work but it is unfortunate that drums. designed by using his methods, would have yielded shells which were 100 weak to withstand the superimposed loads. 1.3 The South African scene Cylindrical Drum winders in South Africa are ex: emely common (see Fig, 3) and are also very important in relation to the output of gold and other minerals from the mines of that country and it is because of this that the machines are used to their fullest capacity, It is not surprising then that, should any failures occur, they are considered with the utmost importance and urgency and ) every attempt is made to rectify such costly stoppages. Ki was, presumably, with this in mind that Mr. John Dolan, the Consulting Electrical & Mechanical Engineer to the Rand Mines Limited, wrote a paper, dated June 11. 1957, the object being to arrive at suitable equations with which to enable a more accurate assessment 10 be ‘made of the loads, and hence the stresses, on a winder drum shell. Commenting very briefy on this paper. Do‘an firstly evolves @ “factor” equation which enables the direct Joad upon the shell to bs computed. d=pending upon the multifarious parameters of design. Waters’ experiment is also mentioned and Do'an has applied his own factor equation to a similar set of circumstances and presum- ably assumes w shell thickness for his final results. The figures given by Dolan definitely show that Waters’ results Wouid give too thin a shell and this is shown to be so when laler in the paper several analysed with a > obtaining service str machines. some of which have failed. and Dolan’s figures show that a minimum compressive siress of same 35.000 Ib. sq. in. was obtained on these machines using his analytical methods. Compurisons between the valves of these stresses certainly imply that shells desizned to Waters’ factors would definitely have been ton weak. This paper was followed up by another fur more comprehensive work which was read by Mr. Dolan und published in “The South Africun Mechanical Engine in December of 1963. It represents a very great deal of work which had presumably been carried out in the intervening years between 1957 and 1963 and deals’ at great length with many practical tests on various winder drums, some of which were still operating satisfactorily and others which had failed in service. The work also contains an extremely comprehensive list of literally hundreds of machines giving full details of the duty of each and the stresses.to which the shells are subjected. calculated by means of Dolan’s methods, At this stage, the reader may feel that the subject matter is becoming mainly historical. This is quite true, up to # point. be- cause it is necessary 10 briefly outline the researches which have been cartied out to date so as to emphasise the very important points which have either been ov looked ‘or considered virtually insoluble due 10 the lack of knowledge at the time. Dolan’s researches have outlined certain weaknesses in some existing designs and yet. if one cares 10 look through his tabulated list of winders, armed with the knowledge of which of the drums ofthe listed machines have failed and which have not failed, some interesting facts emerge. Let several cases be mentioned in respect of the stresses obtained on various mild stzel shells as follows:— Uae, (1) Machine installed in 1946, shell stress 35,700 Ibyin?. failed, Ue hte (2) Machine installed in 1954, shell stress 32,609 Ibn failed. These are two cases selecisd from the lisi und the stresses obiained are approximatsly equal to the design stress limit menticned in Dolan’s earlier paper. Now examine the following:— Fomen, (3) Machine installed in 1959, shell sizess 4,300 Ibn’,failed qn ‘4) Muhine installed in 1931, shell stress 20.200 Ibn, failed, Both (3) und (4) are well below the design stress limi set_by Dolan, Finally:— v «51 Machine installed in 1925, shell stress 38.000 tbyin’ Apparently still serviceable. 43 (6) Machine installed in 19s, shell stress 48.600 Ibiin‘, Apparently still serviceable, The failure of (1) and (2) may be more easily under- siood when it is remembered that both stresses are near to Dolan’s design limit but this does not explain the reason behind the failuges of (3) and (4) and the non failures of (5) and (6) the stresses in which ate very much removed from Dolan’s limit of 35.000 Ibjint By far the greater proportion of machines in Dolan's tubulated list are apparently still performing satisfactorily the majority of shell stresses being under 30.000 Ib/in with quite a few under 20.000 Ibjin:. How then, ‘ch unrelated shell failures be explained? It is quite sear. from the foregoing, that circumferential compres- sive stresses are not the only criterion for design as s0 much depends upon the designer's individual approach to detail and the way in which a particular drum is con- stfucted and it is the authors’ opinion that much more could be done to overcome previously unrecognised siresses, the most important being the stresses produced by the bending moments imposed at the shell end connections. Before proceeding further it would be advantageous outline the fundamental approach made by Dolan towards obiaining the magnitude of the superimposed loads upon a shell and hence the evaluation of the circumferential stress in the shell material Let_us consider for the moment an infinitely long cylindrical shell. If such a shell was to have only external cylindrical surface subjected to a constant and uniformly distributed radially applied load. acting in- wards lowards the centre, then it is reasonable to suppose that the whole shell would suffer a reduction in diameter ich would be everywhere similar. This is better des- wibsd as a constant radial displacement of the shell surface. Further let it be assumed that this constant uniform load is produced by one layer of rope coils wound around the shell. Suppose now an additional layer of coils be wound on top of the first such as to produce an increuse in the constant uniform load and therefore an increase in the adial displacement of the shell surface, what has hap- 5 we Sapam at rvng shat h = ae cometh TRE section xox pened to the first layer coils? Quite clearly they too would have been given a radial displacement and be- cause of this they would have suffered a change of strain resulting in @ reduction of the load in each coil. This means. of course. that the total load from the two layers of rope. as impressed upon the shell. is not the sum of the individual layers: in fact it is less because of the eduction of load in the first layer. The same situation arises as more and more layers lied where the additional load imposed by a new layer is only partially felt as an additional load on the shell. In this way we can arrive at a proportion. or a factor not greater than unity. by which the loads in new layers are multiplied, the resultant load being that which is felt by the shell. It is upon this simple principle that Dolan has based his rope load “factors” but the deriva- tion of the factors is rather more involved and depends upon such things as size and type of rope. shell material and thickness and the relevant elastic moduli of both rope and shell. One of the main things that has been outlined in Dolan’s report is that the rope load “factors” are not constant between one drum and another (assuming of course that the machines are not identical) ic. the rope load “factors” for a 2in shell having wound upon it @ din dia. stranded rope would be different from that of the same shell having, say, a I4in locked coil rope. Dolan’s factors are applied in the following equation to give the total load upon the shel: Ks ks ks STAT Roar tT Ks aK TT Kedar Ks TMK sR inliKr where ne Ty Ty, To. Tesaverage tensions in Ist. 2nd, 3rd layers up to “n” layers. ‘oung's modulus of shell material multiplied by shell area under one coil pitch. ‘ope stretch modulus multiplied by rope metallic area. This equation, based upon the principles outlined above, has been verified by the authors’ Alll the foregoing research work, mentioned herein. has presumed the loads acting over the shell to be constant from one end to the other and from the point of view Brahe patn Diving moe hayestonan Fig. 4. The essential components of poralle! winding drumool obtaining compressive siresses on tong shells this is quite adequate bur what of the other sinewes which are inherent in-a deformed shell! It is not sulliciemt 1 de purely upon the basis af compressive stress as it will shown liner thut there are other stress2s of equal, if not of greater. importance which should not he ignored. It must be emphasised that the methods outlined ybove can niy produce an approximate value of circumferential compressive sirens at the mid point of fairly long shells and should only be used as a guide for designing winder drums. The reader may well ask haw, over all these years. have designers been able to justify their designs fay being able to fulfil the duty requirements laid down for them? As already mentioned many drums have fuiled. particularly in S, Africa where duties are severe. Possibly due to lack of knowledse ar lack of attention to fundamental detuils. it is difficult 19 know. but from the authors” point of view justification comes from long experience of successful designs coupled with an ap- Preciatinn of the nature of stresses in a shell und aliens tion to detail. In siew of this doubt, is there no heitér und more accurate approach towards designing drums such that designers ean be more sure of the stresses which eun be expected in service? The authors believe that this series will illustrate that a more accurate approsch is possible and that it will shed far more light upon this problem thun has hitherto been possible in the past. 1.4 Mathematical analysis of shells Mathematics is not everyone's idea of bliss, even less so when it comes to the point of using differential equa tons. The mathematical theory of shells must not be confused with simple beam theory..as can be the ten- dency when examining a sectional drawing of a drum Gee Fig. 4), it is a three dimensional problem which necessitates the solution of differentia! equations and ‘can hecome very complex unless one assumes certain conditions of symmetry. Generally speaking it is possible 10 assume certain symmetrical attributes in that winder drumshells are of equal thickness throughout. that they ure. for all practie cal purposes. initially round and that the shell material is homogeneous. that is to say. providing 2 shell is at all times subjected to circumferential compressive stress at every point then the effect of longitudinal shell joins ean be ignored. It is also assumed that there are no longitudinal tensile stresses present. produced by the stretching of the shell in this direction during deflection. his latter assumption is a fairly safe one as the practical end supports of a drum shell (cheeks) are never strong to allow such stresses 10 develop i.e. us the shell deflects the outer edges of the cheeks will move inwards towards one another." Mf faced with the problem of designing a shell using the known mathematical theories whut sort of approuch can he made” Clearly. if one looks at a drum design Urawing. seeing the ends of the shell fixed to rather com Plicaied cheeks of virtually incalculahle stifness. the problem looks insurmountable. unless, once again. other Fundamental assumptions are made, ‘The problem facing Dr, W. R. Crawford when he Published his article in "COLLIERY ENGINEERING’ ig. § This illustration shows the compression rings around the inner surface ofthe shell plate. An older ye bolt-on shells shown in July 1948 was how to present ing manageable form, 4 practical and usable theory based upon the more sophisticated theories of cylindrical shells. especially when confronted with the problem of compression rings 3). Unfortunately. Dr, Crawford has made some very rash ussumptions, us indeed he had to, in order 10 allow the simplified approach to he made possible. The assumptions made in his article encompass these already outlined above hut in addition there ate: (a) That the winding ropes impose 3 uniform external radial pressure over the whole surface of the drum, fie. no aecount is taken of the variation in rope pull as the rope is eoiled on to the drum, He has erred on the safe side here, as the shell load ix assumed as being produced” by wound on at constant tension, the tension being that which exists when the conveyance ix at pit bottom ie, the maximum." (1 That each end support behaves in the same fashion asa compression fing. Le. the supparis Uelleet radially” inwards when’ the shell hecomes TInaded. This is clearly incorreet as very litte radial slefleetion occurs at these points, If a fabricated rum is considered having dise plate end supports radial defleetion is virwally absent (Fig. 4. fer That the supported ends af the shell dev notvey ising ‘supper Detected form oF abel plate (a) rolute, This means that a tangent, drawn longi- tudinally. to the shell at a point where the shell is attached to the end supports remains horizontal before. during und after defection has taken place, Rotation definitely takes place, if it did not then ‘many more drumshells would probably have failed than those that have done so already (Fig. 6). |i should be stated at this point that Dr. Crawford way fully aware of the assumptions, already itemised ‘ubove, und he has stated this in his paper but unfortu- hhately, a practical shell does not behave in this ideal way. as the reader will be made aware of later. but no cary way could be found to enable existing theory to be utilised ‘in such a manner except by assuming certain ideal conditions. Although Crawford's work is being criticised on havis. i is. however. the first major attempt at applying true cylindrical shell theory to practical design and the “pswers can be obtained merely by the use of a slide Je. It also emphasises the importance of the longi- tudinal bending moments which are produced at the ends of the shell and which are correlated to the maxi mum compressive stresses at the mid point instead of, 28 in previous empirical work, basing the whole design ‘upon only the compressive stress. Dr. Crawford's assumptions relating to the end supports are the subject of the greatest criticism, Whether of not it is desirable to aim at rigid end connections during the design is debatable as this depends to a large extent lupon the position of the brake path. Quite clearly it is not desirable to allow too much distortion of this ‘member as braking troubles may possibly oceur but. be this as it may. some movement clearly takes place ‘an practical structures and in this respect it is quite impossible 10 impose a rigid anchor for the shell ends. (see Fig. 6) ‘Those readers who are familiar with Crawford's paper should Took at the cross sectional drawings Figs. 4 and § on page 263 of the July 1949 issue of "COLLIERY ENGINEERING" (See also Fig. 7 in this text). From these drawings it can be seen that the shell is shown Testing on a ledge forming part of the cheek to which it is usually bolted. using countersunk headed, nibbed bolts. Let it be assumed that the “end ring”, as Craw- ford has called it, is absolutely rigid. The’ shell end ending moments are trying to twist the end ring but the only media through which the moments can be ysmitted are the bolts themselves. jenerally speaking. the bolts are not strong enough to allow the full possible end moment to be transmitted in this way and they will extend elastically or yield*. As sonn as this happens the end of the shell is allowed to rotate, in tbe longitudinal sense, and ultimately “settles () Fig. 6. Par sections st shell support point “Crawlord” bases tvs theories on the fact tat the shel! suppor jomis do not rotate (ig, 64) whereas in practice some rotation ofthe joint is ineviable (Fig. 6b grossly ied) as oll practical end supports hhave some degree of elasticity down” at some value of end moment less than the pos- sible maximum, thus reducing the severity of bending stress in the material of the shell itself. It is quite pos- sible that, had this not happened. some drum shells may have failed in bending as the degree of stiffness in some of the older end rings (cheeks), being of massive cast section, was probably quite high. It can be proved that, as the end fixing moment is reduced, so is the shear stress at that point until a point is reached where the shell end moment is zero where ‘upon the shear stress is at its lowest value. Shear forces are, generally speaking, not all that severe on normal drums but can assume some importance in connection Wwith certain types of shell end fixation methods.* * Another criticism of Crawford's work is that, if one desires 10 design a shell with compression rings which do not have similar cross sectional areas and which are not all uniformly spaced along the shell, it is not pos- sible, This is quite a justifiable criticism as in fact very ‘often such a case does arise and one could be left wondering how this theory can be used. So far in this present work some historical aspects have been introduced mainly. as previously mentioned, to outline the state of researches existing until recently. All have been criticised in one way or another and justifiably s0, as falling short of the true state of affairs ‘existing on a practical drumshell, and to conclude this Bot fares m this type of ammecon dave been Known so cer ‘Fie Seporant nes of oon Be "wile Beake path Fig. 7. Part section of one of the older type east end supports ‘elered to as an end tnglirst article these are now summarised below for con- venient reference 1.5 Summarising 1) Wars His experiments produced results which. if used. would have given designs which were toa weak Dylan has more or less proved this. Only compressive stresses were analysed. bending land shear were not apparently considered. Dotan ‘Although 4 monumental work he is still only relating his results 10 the compressive stresses but as such his results are more accurate than anything prev- viously used, Only two people associated with him, Messrs. T. C. Kuun and R. S, Loubser of the CSIR of S. Africa appreciated the significance of axial stresses. Crawford The first article of any real value to the designer in this country (United Kingdom) introducing true shell theory which for the first time considers shell bend ing moments and shear siress as well as compressive Stresses. The main criticism is his treatment of the shell end fixation and the fact that asymmetric com: pression ring layouts cannot be allowed for. In Part Il of this series we shall deal more fully with the mathematical theory of shell design and investigate the problem of shell deflection, producing results from which defection, bending moment, shear force and compressive stress diagrams can be plotted. Bending ‘moments will be examined with a view to assessing the severity of the bending stresses and the magnitude of these stresses will be compared with those of the more 2 familiar circumferential compressive’ stresses. Shear stresses will also be commented upon. Ravenrners Waters, £, 0, Paper No, LIM read at Annual Meeting of the ‘ASME. in December 1920, outlining the results of tes Sime oul"te: determine loading factors Drguanign "Aur of leap Winden-— manus on the "Design. “Construction, Application “ang Operation Winging Engines and Mine Hants’ Firs published in 193% Stcond esiton 1998 Dolan, J. Bresident of the South African Insite of Mechanical Engineers (19s4/48)" Electes’ Honorary” Member 98k Con sulting Electrical and Mechanical Engineer to'the Rand Mines Tia From 1948-190. Firs paper ented “Winging Drum, Shell Loading due 10 Successive Layers of Suressed Roper June’ 11. 198 Second paper entitled “Winder Drum Treag, Design Invest ign” published in “The South African Methania? Engines? of ebember 196. Crawford. W. R. Article publthed in “COLLIERY ENGINEER: ING™ of july 1849 emttles “Design of Colliery Machinery and Equipment™ Part One. This frat of 3 teres of articles eals im particular with the design of Winding Drums Authors’ note There are many other references which could be given but it has been thought advisable to restrict these to the ‘most significant of recent years. the works of which may. or _may not be familiar to every reader. The above are significant due to the fact that some | designers in the United Kingdom may be using Craw- | ford’s results and most certainly designers in South “Africa will be using Dolan’s results as the latter has been the most authoritative work in that country to date, Much work has been done in respect of strain gauge testing by such bodies as the British Welding Research Association and the Design Advisory Service but no firm theories have resulted from any of this work2.4. The initial approach Before any serious line of research can be undertaken there has to be clear understanding in one’s mind as to the method of approach. Quite clearly this is a com- plex problem and it is only by building up a picture from very small beginnings that one can attempt a solution, The question of finding the stresses in the drum struc- ture is of course the main aim but finding the load on the shell is the biggest problem of all (see 1. By now nearly every reader will be familiar with the fact that, especially with multi-layer winding, not all ¢* the load from subsequent layers of rope is irhpressed upon the drum. This is due 10 drum deformation, by virtue of adding new layers, affecting a change of strain and hence a change (reduction) of load in the coils of rope in the under layers. (see 1.3). To allow for this a designer introduces “factors”, by which the average rope pull of any layer is multiplied, the resulting load being applied 10 the drum shell. The sum of all such loads from individual layers being the total load on the shell for the purpose of estimating compressive stresses, It was usual for the “factor” applied to the first layer to be unityt and those for subsequent layers to be less than unity, the values reducing gradually in some rather tie 7 / / / ~~~ ‘Fig. 8 Axiol plone of reterance of shell vague fashion which did not appear to be related to anything tangible. It was because of this and also to the fact that Dolan had already published his, work, that the present line of research was initiated. It was thought advisable to teach .some logical conclusion based upon accepted theoretical reasoning so that designers in the future would be better equipped to enable them to know, within reason, what stresses could be expected as a result of their designs. 2.2, Theoretical considerations of shell deflection Although it has been mentioned previously (1.4) that it is impossible to achieve rigidly fixed ends to a pract cal shell, it was decided to use this condition with wi to start the present line of research and to eventually build a computer programme around it, modifications being made to it as research progressed, ‘The starting point then was the simplest possible shell. 1 was of necessity assumed that, (@) the shell material was homogeneous and of con- stant thickness (b) the shell was initially purely cylindrical (©) the shell ends were rigidly attached to some solid unyielding support (@) the shell load was uniformly distributed and con- stant all over, and was acting radially inwards. It can be seen from this that the shell ends could suffer no radial displacement and were not free to rotate in the axial plane. (Fig. 6 part 1). Fig. 8 shows such a shell and clearly indicates the axial plane of reference. Fig. 9 shows the same shell in section and elevation, the axial plane in this case being the pane of the paper on ‘Which this text is witten in relation to the sectional view. How then does the shell behave under these condi tions? To establish this e must make use of the mathematical theory of elasticity as applied to shells. Without going fully into this theory, which is ade- quately covered elsewhere, it is sufficient to state that certain conditions of symmetry can be accepted as being appropriate to this particular problem. One of these conditions is that the load be symmetrical with respect to the axis of the shell: that is to say the valve of the load is identical at every point measured around a cir cumferential line. This does not imply that the value of the load around a circumferential line at one end of the fioroved snd i fact he “actor” sping at he mi pat Fae iy nai tnt aes oth Fig. 9. Section and elevation of shelshell has to be the same as that at the other end, the load can vary in any way whatsoever along the length of the shell but it must be symmetrically disposed around the cireumference at any panticular point This is very nearly obtained in practice as the variation ‘of load. or pull. between one end of a single coil of rope and the other may be only a matter of a few hundred pounds whereas the to1al rope pull may be several thousand pounds. A small percentage error indeed. Another condition is that longitudinal stresses are zero* as previously mentioned (1.4 para. 2), This leaves the 3 remaining forces with which we are concerned, (e) axial bending moments (f) radial shear forces (2) circumfezential forces (compression). I is further assumed that the shell is thin in relation to ils mean radius such that the compressive stress across a section is everywhere sensibly the same. This is true enough of most practical shells. From the fundamental equations of equilibrium we obtain the basic equation:— a “fo al ® aw et Ww ae shell remains constant along its length then equation (1) reduces t0:— ae @ and is the Hexural rigidity of the shell, comparable to EX/ in beam theory. E=Elastic modulus of the shell material el} thickness in inches oisson's ratio an radius of shell in inches suing deflection in inches oad function. In the case of a uniformly distri- buted load this becomes the value in Ib/in’. Z can however be any function of x and is indeed so when the load varies slong the shel x=any point measured from the end of the shell. It is not proposed to deal with the integration of equation (2) as this will detract somewhat from the general theme and any equations given will have to be accepted as being correct for the particular case in question, 2.3. Long shells Very nearly all shells applicable 10 winding engine parallel drum practice can be considered as “long” in terms of the shell theory being discussed but this is not always the case and care should be taken, when using such theories. 10 make sure that the length is sufficient {Pp sean aye SEE ee a aS | Sa a “ Fig. 10. An ‘intintely' long shell subjected to 2 single band of load uniformly distibuted around the circumference sumone ieee I ig, 11, nine’ shell subjected to.» uniformly applied constant {ond ai over outer shell and acting relay wares before using the simplified versions to follow. What should be the minimum length? Generally speak- ing for the case of fixed ended plain shells the overall unsupported length should not be less than 4/71 inchest. There. will be small errors even on shells of much greater length but these will be very small in comparison to the general stress. magnitudes. Before dealing with the primary equations let us imagine a shell of reasonable length having completely unsupported ends and subjected over its cylindrieal sur- face to a constant uniform external load similar to that shown in Fig. 9. Common sense tells us that the shell i® going to contract radially and in fact boiler theory a3 be used to obtain the radial deflection, This theory states that the circumferential stress boiler shell will be:— a Pape aan where p is the pressure is the mean dia 1 is the shell thickness rsd? the mean radius. Knowing E, the radial deflection can be found from the relationsbip| Stress EX strain ten Pee xe and the radial deflection A= 2 =r 2) 1t will be seen that this simple equation will continue to appear in this work and is of fundamental importance. In the case of long shells with supported ends it wil be:seen that the radial deletion at the mid point along the length will be very neatly equal to priJEr except for the very small differences created by the form ef equation governing the deflection values. From this it should be seen that the maximom compressive stress in ASSESS RAINE 8 minimum enh owas ehnge wih seo' very long plain shell cannot be any greater than pritt 41 the centre and it was on this basis only that all Jprevious workers (except Crawford) evaluated shell sresses after using the appropriate load factors previ- ‘ously mentioned. In view of this then, what is the effect on the com- pressive siress when some form of end support is pro- vided and what happens at the ends? 2.4, End supports Before discussing this aspect it should be pointed out, to those readers not familiar with the notations used, that the énds of a shell. in mathematical terminology, are called “boundaries” and in future throughout this series the term “boundary conditions” will frequently appear. This relates purely to what is done to the shell ends i.e. they may be rigidly clamped, completely free, of elastic, in fact anything may be done to the boundaries provid: ing it is subsequently possible 1 translate such conditions into mathematical notations for the purpose of analysis. 1k will be as well at this stage to try and understand how a long shell behaves under load. Let an infinitely long shell Fig. 10 be imagined and let it be subjected to a single band of load uniformly distributed around the circumference. The deflection pattern of the shell is clearly indicated but greatly ‘exaggerated and is seen to be symmetrically disposed on either side of the load point. At a certain distance from the load point the deflection changes sign and becomes negative, then positive, etc., put the magnitudes of these supplementary deflections re very small and become even smaller the further away from the load they are measured. 1f the direction of the load was reversed it would pro duce a similar deflection pattern to the above but every- where reversed in sign. The magnitude of the principal defection in both cases is directly proportional to the applied load, Let it now be supposed that the infinite shell is initially subjected (0 a uniformly applied constant load all over its outer surface and acting radially inwards; in this case the shell would be given a constant radial defiection as shown in Fig. 11 in which only one side of the shell section is shown. If we now select two points (x- and jy) equal distance apart to the end supports (assumed fixed) of a typical long shell and apply to each of these poims a single band of load as in Fig, 10 but acting radially out- wards and of sufficient magnitude to produce, each by itself, a maximum defection A then there exists a situa- tion similar to that arising out of a shell of length “I” having rigidly clamped ends and subjected to a similar uniformly applied load as in Fig. 9. The two identical eflection diagrams about x-x and y-y are similar to those in Fig. 10 but it will be seen that each diagram interacts with the other in the central portion of the shell. Fig. 11 clearly shows this and it is why the simple equation pr/E1 may not give exact values for mid point deflections. We are here dealing with a simplified form of the “Method of Superposition” in which individual defiec- tion diagrams are all “algebraically” added to give one whole pattern. This is what should be done with the diagram of Fig. 11 but it will, in general, be found that the discrepancy from 3 at the midpoint of the shell is of relatively small magnitude and this fact will be ‘established later on. It would be quite interesting. and entertaining. to dis- cuss the “Method of Superposition” at great length be- cause here is what appears to be a vety simple manual ‘method of solving shell problems. This is not so, especi- ally when compression rings are introduced and when the shell ends themselves are allowed to rotate in the axial planes. Graphical analysis is really an art in itself and as yet the authors have not fully explored the possibilities of ‘obtaining solutions to more complex loading problems. Not only is the method laborious but even when the diagrams have been prepared there is still the question of obtaining bending moments and shear forces, as a deflection diagram alone, although interesting, is not sufficient, For the purposes of calculating the slope, bending moment and shear forces the only sensible method of obtaining the Ist, 2nd and 3rd derivatives from a deflection diagram is by means of finite difference calculations but any reader conversant with this will appreciate the tedium and inaccuracies involved, unless only spot checks are required. Such spot checks will, however, be necessary at the boundaries but it is here where the utmost accuracy’ is required because of the large variations in the rate of change of the deflection pattern thus making finite difference calculations very difficult. There are much better ways of obtaining the same results by means of computer programmes and they are much faster and infinitely more accurate. Tt is because ‘of these computer techniques that the methods outlined above have. been disregarded except for one of two simple cases used later on. Before proceeding with the mathematical analysis of the simpler shells it would be as well 10 define the conventions used in order to arrive at the correct inter- pretation of results. CS) (——) Te teatng Ve beng Fig. 12. Convention for the shapes due to axial bending t tL ate got Fig. 13. Convention for the shapes due to shearsoos NY can gfton. co anion ol The following conventions will apply: Loads or forces acting on the shell Loads or forces acting radially inwards towards the shell axis will be classed as +ve. loads. Loads or forces acting radially outwards will be classed as —ve. loads. Bending Fig. 12 illustrates the convention for the shapes due to axial bending. Shear Fig, 13 illustrates shear convention. The above convention enables the general shape of deflection, bending moment and shear force diagrams to be visualised from only written results and will be Used throughout this work. Deftections Radial deflections measured inwards towards the axis ‘of the shell will be considered as +ve. (i. shell con- traction). Radial deflections measured outwards will be considered —ve. (ie. shell expansion) NB. Positive loads do not necessarily produce positive deflections at every point. This will be evident when, in part 3 of the series, partial loadings are considered, Let us now examine, mathematically, what happens to the shell in Fig. 9 in which the boundaries are rigidly - ou y Fig. 14. Graphs of functions WD. YG», 6(82), and (8x) clamped. Ignoring longitudinal tensile stresses the governing equations are:— Debection we be 1-2669+¥ 6x) pr Fz lw Ba) inches o Bending moment ‘Mx’ a R1Gx)-v 6a) Ib in/in of circumference (5) Shear force Qe'= fn B3)+¥ G2] Itfin of circumference O The expressions within the brackets are:— w (Bx)me* = [cos Bx+sin B x) oe ™M Y (B x)me* = [cos f x—sin fx) oe 8) O(Bx)met = con px =) n(Bx)me**sin fx (10) “ 1285 and po V/2S=P appro Say fot mil eel In the above @ x is simply the product of @ and the distance ‘x from either boundary. It should be noted that pr'/Et has appeared as a coefficient in equation (4) and, for a given shell geometry and uniform loading, is a constant. The term within thebrackets produces a variation in the actual defection “value in accordance with the distance ‘x’ measured from Kher boundary. for example. from conditions given, Felating to the shell of Fig. 9. the known deflection at the boundaries is zeto by design i when x=0, 8.50 and y (2.1) 1 and equation (4) pe becomes S [1—1ezero, When x is very large Ax Fy Un reef: becomes large and y (8x) becomes numerically very small and can be considered zero*. Equation (4) becomes, on Lo f-Oep r/Er Fy Unda To enable an easy assesment of the 4 functions of equations (7) to (10) to be made, a graph has been plotted with @x as abscissa and the values of the functions as ordinates for values of 8x from 0 up to 5 (Fig. 14) 2.5. A typical shell AL this stage it would be interesting to investigate typical shell 10 discover the magnitude of the stresses produced. It must be appreciated, however, that it is not possible to determine a shell thickness direct as a result of the equations but rather a shell thickness must first be given and then investigated, alterations being made up or down, as necessary, followed by a further investigation until a shell thickness has been arrived at within which the stresses are at a safe level. Fig. 15 shows the typical shell which is to be investi- ied together with the data (parameters) necessary for 2 siress analysis, the rope pull is assumed to be constant ang wound the full length of the shell and for the moment the resulting load on the shell is assumed con- stant from end to endt. Shell material Yield stress Young's modulus E Poisson's ratio v Rope dia. d posoet being 19 A over mild steel 35,000. Ibyin? UH 30x10 gS 03 . 1Sin 386 (rope will be close coiled) ” iar Bain Rope pull P 30,000 1 Eoastant. The first step is to convert the rope pull into an equivalent unit pressure on the shell P _ 30,000 Tread pressure per inch of rope length mm 300 1b. p= Equivalent pressure in Ib/in?=. = 200 Ibvin?. Tope dia, 1-285 1-285 pe, vii ViGxI For the moment only local stresses will be considered. Defection at boundaries=zero (by design) Detection halfway alone shell (using equation (4) 1285 % 36= 4625 (Gay 4-60) “y (89=-0011 ‘We cannot just insert these values in equation (4) and expect the answer to give the value of deflection at 01285 - x=36in because of the interaction of the opposite boundary pattern as already shown in Fig. 11. What we must do is to calculate the deflection as found from equation (4) and subtract from it the value of the de- fection as found from pr/Et. Double the resulting value and add it algebraically 10 p r/Et. This will give the true deflection at the mid point only.* This can be done purely by operating on the terms within the brackets of equation (4) as follows: — or weet Fra ¥ Gs) substituting gives:— we2l +0011) Er this differs from 1-0 p/E1 by O-O111 and the effect of the opposite diagram is to add as much as this again 10 the deflection: therefore the true deflection will be We 10222 prjEt which in this case will be 1.0222 « 200 x 1002 30% 10%%T =0-0682in whereas p r/E 1=0-0667in. Jean How can we cope with any other point not in the centre of the shell? Basically what we are doing is to firstly draw one half of the defection diagram app able to say the left hand boundary (see Fig. 16) which provides a set of ordinates for the deflection (full line), and superimposing upon it the ordinates of the diagram applicable to the right hand boundary (dotted line). ‘The deflection at a point A~A (not in the centre) would be the deflection as found from equation (4) for the full Tine diagram plus the small amount by which the deflec- tion of the dotted diagram exceeds pr'/E 1 at this point. Similarly the true deflection at a point B-B will be the full line diagram deflection minus the small amount by Which the deflection of the dotted diagram differs from prt, Having drawn the full line diagram from the results of equation (4) it is @ simple matter to measure off and modify accordingly to produce half of the nett diagram which will be symmetrical about the mid point of the shell at I, *To,ghan,anuwes by dre compuiaon te teary of hor sete shoal PETROS Sete pa Pad see Boos v © pee 8 , tote Uw 100% 2 S40 te Fig. 18. Typical shell investigated together with the data (pars. (meres) necessary for 2 sess analysis4 fi Hi 1 + 8 Fie. 16. Dellectons for fixed ended shell cena arene have Having found the deflections at every point how do We determine the sizesses? The circumferential compres. sive siress is easily found by multiplying the deflections by Elr which will produce a compressive stress diagram cf the same shape as the deflection diagram but to ifferent scale . The compressive stress at the point JJ, will be 0-0682 x 30x 10* 109 -= 20,460 Ibjin? 19) MP Note:—p r'/E 1 alone would produce a compressive stress of r/t=20,000 Ibyin®, Bending moments and stresses ‘At the boundaries 8 x=zero and 9 (8 x)=0, ¥ (2 x)=1 and substituting in equation (5) gives:— Bending moment fo-1) 200 “Toes 050 Ib injin of circumference. ‘The section modulus of a lin square bar is 0-1666in* and =6.050 = 36,300 Ibjin? ver i665 ~ 26500 Win? very nearly (The influence of the opposite boundary will very slightly modify this value). The bending moments at points other than at the boundaries are difficult to calculate accurately unless the Shell is extremely long. in which case equation (3) is used direct, but the reader is asked to wait until the theory of short shells is developed by means of the compuer programme in Part 3 of this series, the errors the bending stress will be -23 Dethcton ————— 4} k-—--- (In order to illstate the diagram interactions the ‘supplementary’ dellections around the been exaggerated Manual calculations are not very large but they are evident, as will be seen, Shear forces and stresses At boundaries 8 x=0 and (8 )=0, W (2x)= substituting in equation (6) gives:— and a P P__200 Shearing force=@ [0+ 1}=2———— = 1,555 Ibjin of cire cn oe (OH nm age 1S Ibn of cumference. As the shell is lin thick this will be the shear stress. The remarks concerning the calculation of bending moments at points other than at the boundaries apply equally as well to shear forces as these are also difficult to ‘calculate accurately. All that can be said at the ‘moment of bending moments and shear forces at the ‘midpoint of the shell is that they will both be sensibly zero for the shell in question. Fig, 16 also shows the true deflection diagram for this shell on to which has been added the line repre: senting the theoretical deflection pr/E7 and it is signi- ficant to note that this value of deflection occurs at dis- tances of 1'835/r# from either boundary. (18-35 inches.) This shell can be considered “long” but it does not imply that every 6-f-long shell is “long” because "7’ and ‘r' may change radically thus producing a different resultant value for the length given by 1-835 Jr. The reader should now be able to see the similarity between the diagrams of Figures 11 and 16 and should also be able to appreciate the slight deflection discrepan- ies which occur over the central portion of the shell due to diagram interactions. Here we have a shell of a given type having rigidly fixed boundaries subjected to a constant rope. pull of 30.000 Tb producing 2 very safe compressive stress of20.460 Ibsin? max.* but at the boundaries the shell is Subjected 10 a yield magnitude stress of 36,300 Ibyin*. this state of affairs was to continue the shell material Ould almost certainly yield locally and even if early Structural overload failure did not occur then the pos sibility exists of a very short fatigue life. So much depends upon the nature of the material and upon whether the shell is welded at the boundaries and whether the weld design and welding technique was correct. Knowing these points it would be possible to predict a fatigue life with reasonable certainty. tis quite clear from the above results that the boundary sifesses are 100 severe and an alternative shell thickness should be chosen and a Iin shell under similar circumstances would produce a stress of 24.200 Ib/in’§ but fatigue stress limits should be aimed at to provide a suitable life, Such high boundary stresses will enabie the reader to appreciate more fully the cause of bolt breakages in the type of shells which are bolted on to circumferential ledges forming part of the cheeks. ‘Although it is not strictly possible to obtain such boundary conditions, as previously stated (1.4, 22) it does give some idea as to the “possible” magnitude of the stresses involved. On the other hand there are rela- live degrees of rigidity of boundary fixation and it is the object of this series to develop ways and means of Duilding in” a suitable value of rigidity such that stresses are kept within safe calovlable limits, but firstly the theories must be explored and developed. conditions ‘One method of attaining this is to attach the shell to exible end supports in order to induce a rotation of the shell at this point and the other method is to allow the shell ends to rest freely upon the end supports.** 6. Reducing the boundary stresses—free end ‘The most obvious way is the latter but this is by no means always possible and the former method can be adopted as an alternative, but before discussing flexible fend supports we shall deal with the case of a shell, identical in every respect with Fig. 9 except that its ends are completely free but constrained only in the radial sense such that boundary deflection is zer0. ‘The appropriate equations will be:— Deflection *Wa" o () os 12) ww 13) ai on tpl and ave ame 64/3238 ‘The functions @(2 3), 9 (2 x) and ¥ (8 2) are as given before in equations (9), (10) and (8) respectively or as in Fig. 14. It should be clear that the “semi-critical length” is no longer 1-835 yt as given before for the fixed ended shell as the equation governing deflection has changed slightly, thus the deflection of pr/E tis only now exactly realised when 6(B x)—zer0 ie. when @ x=1'5728. Cor- responding to this the semi-critical length will be 1225/7 t This means that full defection to approximately priEt can be realised on shorter shells than would have been possible for fixed ended shells. How do the stresses and deflections compare with the previous example? The following calculations will show, Using the same shell geometry as before. Defiection at boundaries by design=zero Deflection ha'fway along shell using equation (11) fi x=01285 x36—462 from which @ (2 x)=—0-00086 ‘eat _x=36=[1-00086-+0-00086) p H/E 1=1-00172 x 0-0667=0-0667in neatly. Bending moment at boundaries is zero by design and is also very nearly zero at the centre. Shear force at boundaries using equation (13). x0 -, Br=0 and ¥(Bx)=1 Pp Shearing force 2 2777 tojin of circum Zp” Taoiaas 777 Wn of ceoum reference ’ ‘These results prove three things:— (a) The shea fore fo fe ends is bal (b) The central defections remain unchanged trom those of the Gxed ended shell within the limits of diagram interactions, ete. (© For # shell of this particular geometry the central Geflections are not influenced to any degree by variations in boundary conditions. Fig. 17 shows the deflection diagram for the free ended shell, the semicritcal length of 1-225 V7i=1225in. ‘What has been gained by allowing the ends to rotate without hindrance? Firstly the boundary moments have been reduced to zero and at the same time the shear siress has been halved. Circumferential compressive stress has remained virally unchanged at around 20,000 Ibn and whereas previously a lin thick shell would have been too thin, due to the large boundary moments, in the present case it would be perfecdy safe.* There are of course bending moments present along the shell but these are not very severe and the criterion for design will be the compressive stress limit, but only when the boundaries are released. Values of bending moments and shear forces for free ended shells will be given in part 3 of this series when investigating the computer results, Such results will be far more accurate than any manual calculation which could be obtained from equations (11) 10 (13) due to boundary interactions. ‘Although these two cases bear no direct resemblance to an actual winder drum, as the reader may know it, purely by virtue of the boundary conditions, it quite larly illustrates how important it is not’ to have If that for fixed * saber gene cay ah, gern, geting racial aleten af shalt ovtnima——— é — Fig. 17. Delleetion diagram for freely supported shell boundary fixauons which are too rigid and in fact it is preferable to adopt fixations which approach a simply supported condition. The reader may well appreciate that this is possible, in fact the older type bolt on shells provided a rather erude form of fixation approaching this condition but the question is, how can a flexible connection be introduced such as to enable an assess ment of the boundary, and other stresses, 1o be made? This problem is, for the moment, too complicated to introduce at this stage and will be treated later in the series. We should, however, confine ourselves to the simpler forms of construction which are, mathematically, more easily predictable and within the scope of slide rule caleulations. 2.7. Short shells In preparing this series the Authors did feel that the ‘mathematical equations relating to short shells should be introduced at this stage in order to obtain compari sons between the three major stresses for the cases of fixed ended shells of length greater than 4/rf and those of shells considerably shorter than this value. It has already been emphasised that there is a severe limitation placed upon the value of manual computation and at best the introduction of short shell theory can only tend to confuse the reader who is trying to grasp this subject. The simple approaches so far outlined are really sufficient to enable an understanding to be gained of the way in which loaded shells behave. In part 3 of this series computer results will be given for various cases and the programme which has been prepared can accommodate short shells equally as well as long shells as the mathematics contained within the programme is adequate to cover any case in question. Such computer programmes are vastly superior to manual calculations as they are infinitely more accurate and are capable of providing results which cannot be attempted in any other way. It is because of the usefulness of such programmes and the severe limitation of manual methods that we ask the reader to wait until part 3 of this series before examining comparable results for long and short shells.) 34 The first computer results In the second part of this series we Jcalt with only be simpler approximate, techniques of manual eampy: stwwo relative wy “long” shells having both fixed and e houndaries. It is now time to intenducs the results chained from the earlige computer programmes s0 that Semparisens cane made belween the boundary and finer siteses tay well ay defistions! relating 10 both long and shoet shells The authors were fortunate in having easy aevess ta the digital compuier situsted in the Erith offices of GEC. (Process: Engineering) Lid, This isa Ferranti machine using Mercury Aurocode language (See Fig. 18) sn operates wholly By means of punched tape. Much vseltl work way already: being carried out by its use in respect of winding engine calculations, such as drum- shafts and brakes, etc. and one of the authors was Simversant with the Autocode lunguuge used. For the curler work however the authors are ine sicied_ 10 the Senior Mathematician, Mr. B, M, Scott ‘shoss help in preparing and modifying the programmes used) in the fundamental analyses was invaluable. The results in this part of the series are a direct result of Mir. Seoit’s labours and the programme existing to date ts an such a form that Ht ean be used as a direct check fon some of the later programme material produced. Such cross checking is of course invaluable in determi the sceuracies of complex problems At the time of completing. the firsi computer pro- mmnie_thers was being designed, in the ofces of GEC. Frith, » double drum winder for South Africa and if was decided 10 utilise the data for this machine {is trial for the new programme. 9943 ur ach drum of this winder was 10 fA diameter by'3 ft Yin hetween anges having a shell Zin thick. Although the drum was being supplied with compression rings it was decided 10 omit them until such time as plain shells hl been analysed. Only the fist laver coils were ape lied. the loading being transformed into a constant uniform loud over the whole shell based upon the sverage rope pull, This loading was found to be 374 Ibjint. US Ow Ax mentioned in part 2 of this series the programme was based upon fixed boundaries, Ze. the slope and de- Acetion at these points is zero, und it wus upon this basis that the programme was initially run. The change In the programme to accept free ended vonditions wus then purely a routine matter. First of all Jet us deal with the results from the com- pier as upplied 19 the boundaries and mid point of a lived ended long shell. The results from the compuler are listed below: - ae Br Fig. 18 The “Mercury” computer at Entn At boundaries Bending moment =-13.579 Ibin/in of circumference = stress = = 20,400 Ibfin Shear force = 3186 Ib/in of circumference = stress = 1.593 Ibjin® Compressive stress = 0 by design At mil point Bending moment =~ 75 Ib infin of circumference = sttess = = 12S Ibjin? Shear force = 9.13 Ibjin of circumference sess 4.56 Ib/in* Radial deflection = 23.535 thou Circumferential Compressive stress Now let these results be compured with some results ‘obtained manually by the use of equations (4). (5) und (6) of part 2, but first let the value of 8 be calculated, 1285 vn Vx? —— Ar boundaries (deflection: 1285 yBe0. Yiasi=l, WBS, MBsIEL and mane Bending moment oP tonite oe iF TF 374 SSSI 713.580 Ib inlin of circumference POTS » 2 Shear foree = 5 (04 1] teT aie! compres 00 Brg se) [+s ssa shar = RAINS Ib‘in of circumference onin3 1 mid point 0-0244, Mr (8 x)=0-0029, # (8 x) 00137 Trom which: = 08 0 (0197 and 9 (8 Bending moment 00274 +0-02441— = 213.580 x ~0-003 = ~40:74 Ib infin of circumference Shear force ” 001374 il .0-0028) 001082 a 2 = 0-0108 x 3.185= 34-4 Ibfin of circumference Detleetion Pe (140-0246) = zy! } using the technique described in part 2 the total deflec- tion will be: Hoag 2 @ HO x 374 x 60? x 1,000 30x 10°x2 =23-53 thou. Ay mentioned in part 2 the errors in the values of ending moment and sheur force at the centre which involve the 2nd and 3rd derivatives of the deflections. are obsious although in this case not at all significant ‘iue 10 the low magnitude, The values obtained manually at the boundary are exactly identical 10 those obtained Irom the computer as also is the deflection obtained at the mid point. which proves that the computer pro- gramme is giving correct results The reader may like to calculate for himself the re- - 7 Fig. 19, Detlection, benging moment and shear force oiagrams ‘Sheil—60in radive 6919 long Zin thick No comp. rings Unitorm tose-—37 Ib Boundary conditions —tigidy clampes 230. 10" sults for a point ITin from the L-H. boundary for which the-computer results are:~ Deflection “w"=20.866 thou, Bending moment “M™ 2.400 Ib infin ference. Shear force "Q"=~ 182.6 Tofin of circumference —remembering. when calculating deflection. 10 allow for the influence of the opposite boundary diagram Slight discrepancies will be found in M and as already outlined. Fig, 19 shows the deflection diagram for this fixed ended ong” shell plus the bending moment and shear force iagrams related to it. and it should be noticed that despite this being the frst of three layers of rope the tensile bending siress at the boundaries is 20,400 bf whereas the compressive stress is 11.750 Ib/in? at the mig. point. ‘We will now compare the above results with what would be obisined when the boundaries are released $0 as 10 be constrained radially but free to rotate in the axial planes. The reader may wish to analyse the results given by using equations (11), (12) and (13) of pan 2. Fig. 20 shows the deflection, bending moment and shear force diigrams for the free ended “long” shell identical in every respect, but for the boundary candi- tions. 10 the shell above. of circum. 3.2 Short shells, ‘The reader has already been warned about using the simple theories already outlined, on fixed ended shells shorter thun about 4/71 in and free ended shells shorter than ubout 2.Sv71 in. Manual results can give moder- ately accurate deflection values on shells shorier than the above appropriate lengths by suitably applying the simple superposition methods previously outlined bu! care musi be taken at each boundary because the shorter the shell becomes the more euch diagram affects theFig. 20. Dellection, bending moment and shear force diagrams Shell—60in racive ‘in thick 68in long No comp. rings ‘opposite boundary in such a way as to modify the nett deflection at the centre. Such simple methods are no longer accurate and it will be found that the greatest inaccuracies occur when considering the bending and shear stresses It has already been mentioned that the theory of short shells should be used but it is not proposed to expand the mathematical theory in this series particularly as this is not intended as treatise on fundamental shell theory and especially as the computer programme can give similar and fat more accurate results. Let it be emphasised once more that almost all cylindrical drum winder shells are “long” in ierms of the theories being discussed, and that the simple theories already outlined are sufficient for an adequate understanding of the way in which shells generally behave under load. ‘The main reason for not dealing mathematically with the short shell theory lies in the fact that when dealing later on, with shell loadings, it will be found that the load can only be found by utilising an entirely different computer technique evolved by considering the progres- sive nature of the coiling of the rope. By virtue of this the earlier computer programmes become redundant and 4 full mathematical study of this croblem would be purely academic. Our aim is to arrive at the ‘rue state ‘of affairs existing on a drumshell and the purely theo- retical approach can no longer be used when consider- ing rope load factors combined with elastic ends, etc. Let us consider a shell identical in all respects 40 the two cases above except that its length has been reduced from T2in to 30in. If such a shell had fixed ends the semi-critical length would be 20in, therefore this shell is in the “short” cutegory. If the ends were free the semi-critical length would be 13.4in and this would place it in the “long” category Figs. 21 und 22 show the deflection. bending moment and shear force diagrams for the fixed ended and free ended “short” shells respectively. What can be learned from the relationship between aay one st the diagrams of Figs. 19, 20, 21 and 227 In order to consolidate the more imporiant points Table I has been prepared so as to enable a quick comparison 10 be made of the various stresses existing at the boundaries and also over the central portion of the shells but not necessatily at the mid point. Relative maximum deflec- tions are also giv tein: Debesiontm chet | Bedea trend [ese | ea sae (SHEED One Centar Companion ne) | 761 | TOR | Deteion an [OHH | OG) | CORT ORT The figures above prove that the value of shear stress fat the boundaries of a free ended long shell are half that for a fixed ended long shell. This almost applies to the short shell in that the shear reduction for free ends is just over halt. Except for the fixed ended short shell there is very litle difference in the value of central compressive stress, as was to be expected, but apart from this the most sig- nificant point arising from Table | is that the general level of bending and shear stresses in both the free ended shells is much lower than that of the fixed ended shells: this is particularly evident at the boundaries, of course, where the bending stresses are 2er0. The fixed ended short shell is rather a special case. It may be thought that itis better from all aspects as the compressive stress is fairly low. Apart from the fact that30in Jong shells ate a rarity on drum winders we are faced with higher bending and shear stresses at the boundaries (It should be remembered that this is still aly the first layer of rope of a multi-layer machine), Although the above results are not characteristic of 4 practical shell they do prove that from every aspect a fixed ended shell exhibits far higher boundary stresses. than the free ended version and it should be the aim when designing 10 avoid high degrees of fixation stiffness. 3.3 Partial loads 1 should now be accepted that the computer pro- gramme, so far produced, is accurate, as the manual check calculations have indicated, and that we are in a position to proceed with rather more complex loading configurations, in particular the effect of a variable load ss the shell. Allied with this we are also interested in discovering what happens to the shell deflection pattern due to partial loads, ie. part only of the first layer of coils. ‘The computer programme was revised to accept load variations slong the shell and it was decided to run it 50 as to obtain four sets of consecutive results based upon 25%, 502%. 75% and 100% of applied first layer coils Tespectively. Load variation was from 388 Ib/in* at the first coil of the layer to 360 Ib/in* for the last coil of the full layer, varying Tineally across the shell, the average being 374 Ib/in? as used in the previous examples. Each of the four deflection patterns is shown in Fig. 23° and, upon examination, produce some interesting facts. The first thing which will be noticed is that the shell is suffering a negative deflection, or expansion, when subjected 10 the 25% and 50% load patterns but is of very small magnitude compared to the general deflections re sn ha nh under the loads. Nearly all practical shells are made in halves with a joint running the length of the shell and, a5 the above theories are based upon complete homo. geneity of the shell material, some readers may feel that all theoretical work of this nature is rather pointless unless it can cover the case of non-homogeneous con- structions. It should be pointed out however that negative deflections can only be produced by partial loading and a such the general stress levels. particularly on a drum designed for multilayering. are quite low and as soon as 75% of the layer is complete the shell is subjected 10 positive defections all over. The second point, which may not be immediately obvious, is that the defection at @ point some 20 inches along the shell remains fairly constant subsequent to the completion of the half layer. “This point is of course situated at the semictitical length from the boundary (1-835 V7 of 20-1 inches) where the major boundary influence ceases, In fact the pattern of the boundary diagram changes very lite as the remainder of the layer is completed. By far the most interesting point arising from Fig. 23 is the variation in deflection of a particular coil as the length of the layer progresses. Let us examine a point 17 inches from the left hand boundary which is where the last coil of the 252% com- plete layer is situated. When $0% of the layer had been completed the coil at point 17 inches was subjected to an additional deflection of 10-64 thou. beyond the posi had adopted when it had just been wound on (Ge, when: 25% of the layer was complete). When this coil had just been wound on it would have had in it the full normal load dictated by the position of the winding conveyance in the mine shaft which, in this instance, responded 10 the equivalent of 381 Ibjin’. It is now given an additional deflection due to winding on more coils which, by virtue of Hooke’s Law, must alter its inherent tensile load; but what value does it finally end up at? 2a naan se 143 oe : Jnso0eg Fig. 21. Detection, bending moment and shear soe} force diagrams see ‘Sheli—60in radiut30in long } Zin thick. No comp. ngs -a 000 Unitorm load —374 Io/in® aN E Boundary conditions—vigly clamoed tou” emg0r4 19Fig, 22, Detlection. vending moment and sheer ‘orce diagrams ‘Sheli—60in radius 30in long in tick” No comp. ngs Unitorm toag-—37e lovin Boundary condhians—simoly suppomed £30" 10" s z Before answering this question it would be advisable to investigate the general behaviour of a rope during a winding cycle. )4 How ropes behave during coiling The pull in a rope, as it passes on to a drum, is made up of many complex factors. Rarely is there a smooth linear transition of load as the wind progresses. Such complex factors can be attributed to the follow- ing dynamic conditions:— (a) accelerations and (b) rope oscillations (1 shock loads when changing layers (2) shock loads when jumping grooves (second layer upwards) Usually (a) only occurs over a few coils at each end of the wind and does not materially affect the shell deflections to any large extent even if the first live coil is near the mid point of the shell, Normally the first live coil is near 10 2 boundary and any additional load due to accelerating is reflected in a slight increase in the shear value at that boundary. Retardations relieve the load in the rope and can be considered as safe from the shell ‘dations en barn sana 7 atts FE esa csg al Ae cot satcion 200 tementt 12 000 bf! wemoerin ste loading aspect. (We are of course considering coiling ‘onto a drum, therefore the conveyance in question is ascending and retardations would reduce the rope load.) The same remarks really apply to (c) as this definitely occurs at a boundary," therefore (c) can really be con- sidered as a boundary load and not a load which can materially affect the central areas of the shell. Rope oscillations (b) are almost always present except at very low winding speeds, and may take any form. ‘Some ‘machines, particularly those with very long roy Planes and those with shallow angle planes (horizontal Tope planes are an example) are most prone to the slightest disturbance and can suffer fairly severe oscilla: tions. If we can consider such oscillations as being purely cyclic then the average or mean rope pull will be equal to the static rope pull at that point, he. the rope pull at any particular point in the wind is equal to the static pull plus or minus some particular value dependent upon the amplitude of the oscillations. This means that the effect on the shell of fluctuating load is averaged out. The effect due to (d) is rather similar to (c) except that cm spel crmtnees ol toy compenment ange rum Fig. 23. Shell deflection due to pari! loeding ‘Sheit—60in radius "69in long. 2in thick No comp. rings.» Loading as shown, Boundary condhions-—rgidly clamped en30 x40"it occurs once oF perhaps twice every revolution. At each cross over pwvint the shell is subjected to an impulse load Which is transient in nature i. it is not sustained. How will such complex loadings affect the stress levels in a shell? 1c seems logical to discount (a) altogether due to it mainly affecting only the boundary shear stresses. Gener- ally speaking the cyclic acceleration of the suspended load is only about 3 ft:ses* thus the accelerating forces in the rope are approximately 10% of the static pull. Allowances can be made for this slight additional load, ay it has a caleulable value, but for the moment it is being ignored ‘Also for the moment (b) is being considered as though its effect is averaged out as previously mentioned. What of (¢) and (4)? They are both transient in nature and both have a local effect and could conceivably con+ tribute to a reduction in the fatigue life of a given ma- chine. In view of such local effects, however, they cannot considered in the overall picture with respect to final shell deformations which demand circumferentially uni- form loads. So far we have not mentioned dead coils land this is a question, not of complexity, but rather of determining the magnitude of the inherent load within each coil. It depends wholly upon the technique used in an applying suitable loads to these coils. Once the technique is known for any particular installation the loads become calculable and can then be allowed for in any ealeula- tion. Such techniques will not be dealt with here as it wi detract the reader from the general context but it remai to say that any dead coil loading can be allowed for in Subsequent. calculations if itis decided that the number of dead coils warrants such a measure. For readers not familiar with winding engine termin- ology a “dead” coil is one which plays no normal part in the winding cycle, ie. it never leaves the surface of the drum. Such coils ate introduced for the purposes of providing spare rope, or as a frietional supplement to the rope anchorage point or to purposely modify rope “fleet angling” (:¢. the angle at which the first “live” coil leaves the drum from the first layer) or a combi tion of all three. In the above brief description it should be quite clear that one type of complex load can give rise {0 another in some measure, but for the purposes of this series of antcles it has been considered sufficient to ignore the eflecis produced by them and to assume that the rope load is shock free and exhibits a smooth transition throughout the winding cycle.3.5 Rope load factors It has been thought for some time that the factors applying to individual layers have been in error, par- siculatly the unity factor normally associated with the first layer, and that the valves applicable to any layer were not constant over the full length of the drum she This is really quite obvious, if one thinks about because all the factors calculated by previous workers, Dolan included, have of necessity been based upon variations of coil loads in the under layers produced by increased shell deflections as more layers are applied, ie. to obiain a rope load factor there must be a defect ‘At the ends of a drum there is little or no deflection, therefore how can there be @ factor of less than unity? Wf it is agreed that there és a unity factor at the shell boundaries and that at the mid point, where deflection ‘curs, there is some factor less than unity then the value of such factors must vary in some way actoss the shell, and approximately in accordance with the deflection pattern. (This is very nearly so up to a point) Let us for the moment confine ourselves to the Srst layer only. At the end of section 33 and from Fig. 23, it was shown that the coil at poist 17in suffered an ‘additional deflection of 10°64 thou. which means that the circumferential length of this coil bas been reduced. As the rope forming this coil had previously been stretched due 10 the suspended load any reduction in length of this coil constitutes a reduction of stress in the rope and the change in stress is directly proportional to the change ) in the rope coil diameter, and hence rope coil radius. 4 Be reso’ cu Prone Men Ra Once again using the relationship. Steess=E x strain we can write, using appropriate notation:— Change of Stress=Er x A/r r Ar iue., Rope load change A= A Arb Change in uit loading == Where: Er=Rope stretch modulus Ar=Metallic area of rope per inch of shell length. Asradial deflection change ‘r=mean shell radius* Putting this into figures gives:— 15% 10x 0-66 x 0-01064 0 '=2925 Ibjint ‘This means that the load in this coil has been reduced by 29:25 Ib/in* (equivalent unit load), which leaves only 381-29-25=351-75 Ibjin’. Expressed as a “factor” this is 351-75 = no: 3a Fig. 24 Relationship between set-induced ‘actors and ratio Ki/ks based upon colling ‘over an “infinte™ shellapplied load in the coil at point 17in is wansferred to ihe shell A similar set of circumstances arise for the coils at points 34in and $2in in that these are subjected to addi tional deflections of 11 and 8:94 thou, respectively. re- sulting in “factors” of 0-918 and 0-933. This is surely proof enough of the presence of a “factor” in the first layer although the values given will be approximate due to the fact that each complete guarter layer was applied as a whole assuming no factors within that quarter layer. (It will be shown later that these values, although approximate for this case. are fairly representative of the values obtained by using more sophisticated methods of analysis.) For the above to be true there must of course be no rope slippage around the drum. If we examine a very severe case where the difference in rope pull between one end of complete coil and the other is at its maximum and then calculate whether slippage will occur it can be shown to be highly unlikely. Take the case of a winder having a very heavy 2in locked coil rope weighing, say, 10 Tb/ft and a drum diameter of 18 ft. The minimum rope pull is assumed 10 be 40,000 Ib. The total weight of rope in one complete coil around the drum is 565 Ib, producing @ rope pull at the beginning and end of the complete coil of 40,000 and 39.435 Ib respectively. Th Using the relationship, 2 Inserting our parameters gives 40,000 39,435 Taking logs (to base 10) of both sides ee = 1-01433 nearly log 1:01433=2nu log 271828 fue, 0006025 jx 044343 . 0.0060 # 04343 x2 '=0-0022 which is an extremely low coefficient of friction, Where: T,=inial pull T,=final pull e=2-71828 = coefficient of friction angle of wrap of rope around drum. If the coefficient of friction was to fall below this then slippage could occur. A conservative figure for coiled metal surfaces is 0-05 which is well above the limiting figure obtained. The reader should consider the case of a friction winder where the rope only has ap- Proximately 180* angle of wrap yet which can sustain very high out of balance loads compared to those with which we are concerned. From this example it can be appreciated that rope slippage around a drum shell is highly unlikely. It should be quite clear from the foregoing results that, in view of having eliminated rope slippage, there is a rope load “factor” relating to the first layer which varies from one end of the shell to the other and is sensibly unity at the boundaries. This does, of course, bring us 10 the big problem of how to find these “factors” and it is no simple matter 10 calculate them. Fig. 25:,This diagram ilustrates the way in which shell deflections ‘conuibute to rope oad Tactors 3.6 The manual search Both authors, having convinced themselves that such first layer factors existed, spent considerable time in tying to establish suitable equations from which the factors at the mid point of long shells could be derived. It had already been shown by Dolan, and subsequently Proved by the authors, that the better known factors from the second layer upwards depended upon the ratio of rope metallic area to shell thickness and it was thought reasonable that the factors for the Gist layer would depend similarly upon this. Some form of equation was needed based upon Known theory. Alternatively the existing computer programme could have been used in a revised form and’programmed such as to apply each ‘ew coil progressively with a view to obtaining the Snal coll loads by means of progressive feed back Joops. It was found, however, that with the type of programme as existing the cost would have been extremely prohibi- tive in terms of computer running hours. Of economic necessity it was decided to try a manual approach by firstly assuming an infinitely Jong shell and applying it to one coil at a time. As each new coil was added its effect on all previous coils was calculated and in turn the interaction of all coils upon each other was assessed. In view of the cumbersome nature of the expressions obtained it was only possible to apply the interaction feedback once only in determining the equi- brium conditions following the application of every new coil. As was to be expected a form of mathematical series was evolved but it was desired to extend this series such 88 to cover sulficient indivi as to allow the factor to “ constant value. Tt was decided to apply say 20 coils and derive an equation from the series which defined the deflection ‘of the 20th coil below the unstrained shell position. Tt was further necessary to arrive at a second equation which would define the additional defection of the 20th coll after tay 20 more coils had beea applied. Twenty coils have been mentioned as it was necessary to select a point far enough away from the first coil so as to preclude any interaction due to the shell deflection diagram adjacent to the first coil. Having obtained the two equations they were then combined to form a single equation which would, it was hoped, provide an expression for the fist layer factor. This was of the following quadratic form:— Fs=1-Z3=Z? (YX) a (14) where 5, Y and X are rather complicated summations 23of the heights of the defiection diagram ordinates for sch individual coil and Z is a function of such variables as rope type and shell geometry etc.” A small computer programme was prepared to evalu ate the above expressions over a large range of Z values and for 8 widely differing shells based on contract records. The results are shown on the graph of Fig. 24 and were extremely disappointing as may be appreciated by noting the behaviour of the group of curves forming the narrow band. ‘The factor values Fs derived from equation (14) have heen plotted as ordinate against the ratio KriKs as abscissa, where: Kr=metallie area of rope per inch of shell» stretch modulus Ks=shell thickness * Youngs modulus It is clear to see that the band of curves has a turning point ata factor value of approximately 0-83 which would mean that by using equation (14) the factors could only range between 1-0 and 0'83 which is quite ridicu- Tous. For example, if Kr was very small then the shell could deflect quite a lot more under the action of additional coils andior layers without affecting to any great degree the load in the coils of the under layers. If Kr was very large the converse would happen so that for very low values of Kriks we obtain high factors (when KriKs= 0 the factor=1) and when KriKS is very large we obtain sw factors. 'AS a matter of interest and convenience a computer curve fiting programme was utilised, based upon the mean values found for all 8 shells, to produce a quad- ratic factor equation involving the ratio KriKs as the independent variable and this produced:— rooosrs [2] -osne Mer BR ay The mininom value beg O15 at Kekonoses a aS See ats Kine see ea he ma an fr neigh rope SS IE AE 8 a a Stoney Soa us he Satay beet et ‘eek ats itt ty bon ate Foe 4 modest 36in radius by lin thick to 96in radius by 328 I as psy eed, wea mpg te der tion of the mathematical series, that it was only possible Sra Set oleh regent e values of approximately 0°3. 3.7 The importance of Krik's This ratio is extremely important and was used in the ov, lend fo pin in RE te deraon inverse sense in Dolan’s equation. (1.3). It is really defining a ratio of the load changes between rope and shell. If, once again, we examine the simple relationship. Stress=E x strain which produces, Load= (Ex area) x strait then for a given strain (4/r) to which both rope and shell are subjected the load “gained” by" the shell is related to the load “lost” by the rope in the form: Load lost by rope Er A? Load gained by shell Er 4i ‘=a constant ‘The above is a constant for one particular rope/shell combination. It should be clearly seen then that as Kr increases so oes the ratio Xr/Ks which means that for a given strain the load lost by the rope is a higher proportion of that sained by the shell; this will affect. the rope load Factor”. 3.8 An approximate proof of the 1st layer factor Fs The following proof is based upon loadings over an infinite sbell. - Let a uniformly distributed load of intensity “p" be applied over an infinite shell (Fig. 25): this will produce 2 radial deflection everywhere of If all the load to the right of xx is removed we are left with a deflection diagram (a) Fig. 25 which crosses sox at 4/2. This latter fact should be apparent from conditions of symmetry in that if instead, all the load to the left of ax bad been removed we would get a mirror image pattern and, as the total deflection A is equal to the sum of all deflections, then it is clear that at x-x the sum should equal 4 therefore the deflection curve crosses 7-r at 4/2. On examining the graph of Fig. 23 it can be seen that the initial deflection of coil 17 was 10-49 thou. and its final deflection 21°33 thou. which corroborates the above facts. AAs the presence of a “factor” has been accepted it is clear that the ultimate deflection is less than A and its value, is A x the factor since the factor being sought reduces the applied load from “p” to “Fs.” therefore the deflection becomes Fs.A. ‘At this stage we cannot assess the shape of the true load pattern adjacent to the coil at x-x (i. the cross hatched area is indeterminate in nature) and the only way to arrive at a suitable factor equation is to firstly assume that the maximum unfactored load “p" is opera tive to the left of =r producing a deflection A/2 for the ‘oil being examined at 2-x, and then secondly to assume that the minimum factored load Fs.p is in operation producing a deflection Fs.4/2. Clearly the deflection of this coil will lie somewhere between these two valves, the average being 4/4 (Fs+1). Let us for the moment consider the maximum possible factor. When the remaining load to the right of ar is applied and with the factor in operation the coil at zr = abwill have deflected from A/2 to Fs.4 ie. the additional defection is FsA-A/2=4 (F5-4), The load change in the coil will be-— dr Kr pe SE sept Bey 7 2 and since E«1=Ks the load change becomes:— ey ra Under the initial conditions depicted in Fig. 25 coil at x-r was assumed to have init the fll load "p 25 it was the Jas coil wound on up fo this point. In this case, after subsequent deflection to Fs.A, the new load in this coll will bes ree» pee and the factor Fe wil bes— pp ED) Fe ? from which we obtain Fs(max.)= tt Xe2 Re 13 Rrlke w= 19) If now the coil had originally been defected to Fs8/2, under the aetion of the facorised minimum load Fap to the left of zx, the load change in the coil would be EE nse and by going through the same procedure as above we obtain the minimum factor:— 4 Fs(min.) =) and similarly if we were to assume that the fection was the average value AI (F541) we would artive at an average factor: kr aR Fs(av.) = A “Te Using the average factor equation Fs (av) it was found by suitably modifying the coefficients that the slope produced by it at KriKs=0 could be made ideatical to that of the curve Gtting equation (15) and the following final equation is obtained in which it should be noted that the coeficients suffer only a slight change:—— 140-2513 Kriks 140-7556 KriKe — .. (18) It is interesting to note that the resulting independent variables in Fs (max.), Fs (min.) and Fs (av) are purely in terms of the Kr/Ks ratio which gives added proof of the fact that the factors are dependent wholly upon this Fe(av.) Tatio (within sensible design limits) and independent of ‘shell radius and loading. All three equations (16), (17) and (18) are plotted on Fig. 24 The type of factor just described has been christened by the authors the “Self Induced Factor™ because it is 4 factor, applicable to a layer, which is induced purely by the addition of the remaining coils of the same layer and not by the addition of layers above it. We can at this stage, before terminating this particular article, qualify the three principal factors applicable 10 rope coiling, with two of which we shall be principally concerned. These are:— (a) Self induced factor Fs— ‘That factor which is applied to the coil loads of any particular layer but which is induced purely as a result of coiling more rope under Joad on to that same layer. It is simpler to think of this in terms of the first layer only as it is interrelated from the second layer upwards with the better known factor (b). (b) Mutually induced factor Fx— ‘That factor which is applied to the coil loads of a layer by virtue of the mutual action of additional layers wound over it. In the ultimate it is not possible to isolate this factor as it is interrelated with (a). (©) Auxiliary factor F— - That factor which is applied to the coil loads of any layer by virtue of the change in coil diameter due to the physical yield- ing or crushing of the ropes in layers beneath it, We are principally concemed with (a) and (b) above because too litte is known about the behaviour of rope crushing, and the transverse modulus of elasticity related 10 it, to be able to assess any factor such as in (¢). Rope manufacturers know of no physical research in this Gitection, Such a problem is very complex when one considers the structure.of all the various types of rope use today and this particular aspect is not being pursued. A little thought will show that, could this be allowed’ for, it would only serve to reduce the shell loads stil further, thus by ignoring its action the results will err on the safe side. 4.1 Further thoughts on rope load factors At the end of section (.8) of this series we defined the three basic factors affecting the loads in the coils of any layer and it was further stated that the “Auxiliary Factor", due to rope crushing, was not being considered i view of the scarcity of data from the rope manufa turers relating to this particular aspect of wire rope behaviour. We shall however continue to concern our: selves with both the “self indvced factor” and the “mutually induced factor", both of which play an im- portant part in the determination of shell loading and 0 apologies can be given for dwelling on this subject. Tt has so far been decided that, as the original curves prodiced by equation (14) are in error. the more prob Able factor equation (18) will give results in Keeping with what we should expect. Equation (18) i repro- duced here for convenient reference and is applicable only to the fist layer. 1402513 KriKe T$O-TSS6 RrIKe asy Foun 2ng teen al aie ses we Paambeners How can the numerical values obtained from this equa- tion be proved? Further thoughts on the matter suggested that a com- puler programme be prepared so as to emulate the manual approach (3.6) but with no limitation on the 4 back terms, By using the compute, feedback was i longer any problem. The programme was duly pre- pared, based upon the same infinite sbell analogy, and Several successful runs were made to produce the results shown graphically in Fig. 26. For these trials the coil loads were initially constant ‘and were assumed acting at one inch increments along the infinite shell in exactly the same way as in the abortive manual assessment. Five plots are shown with the first layer self induced factor “Fs" as ordinate against coil pitching as abscissa, the curves being lettered from (a) to (e) inclusively. The data relating to each curve is as follows in Table Il. ‘Various total lengths of coil coverage were tried in order to observe trend behaviours. [TABLE TLaPrnent Dave Rewarng ro Ras Grn Fie 26 ROPE eee) Se oR See? The values of the ratio of KriKs relating to each, case is shown in Fig. 26. Ii should be pointed out and emphasised that, despite using factual ropes for assesing the rope metallic area per inch of shell length, the ropes are themscives all fumed to be acting at one inch pitch such that each additional coil adds one inch to the layer length. In practice, unless using a one inch rope, this is never so, fs each coil Joad acts at increments of one rope “pitch” ‘along the shell. This approach was adopted, at the time, as being more convenient but later in the series factors will be assessed using the true rope pitching. - 26 Fig. 26. Co load factors based upon coiling, Under uniform rope pull, over “infinite” shell At no point is the shell artticilly ‘estreined in the radial erection Fig, 27. Dellection diagram relating to cure (0) of Fig. 26. Note the deflection ‘Under the fist and last colts The curves clearly show that the “factor” values are not cossian along the whole length of-coll coversge, particularly when approaching the last coil. ‘The last Ebi is assumed to have suffered 90 load change and thus has a factor of 10 ‘We are not interested, for the moment, in what hap- pens at ‘ends but only in what happens over the EEatal areas of that part of the shell which is covered by cols ‘Examining curve (a) fist it can be assumed that the factor is "sealing at about 0617 and, if we look at the graph of Fig. 24 or Fig. 28 of this article 2 factor of 0807 is observed corresponding to a Kr/Ks ratio of 0°538. If curve (a) of Fig. 26 is assumed correct then the factor curve Fs(av) of Fig. 24 is 1-2% in error which, in view of the approximations assumed in its derivation, is not too great an eor. ‘Curve (c) of Fig. 26 is “settling” at 0-930 and the Fe(av) curve of Fig. 24 gives a factor of 0-926 for'a Kr/Ke ratio of 0165 which, ia the same sense as above, gives an error of under 05%! In view of the limited number of coils involved, ‘curves (b) and (d) must be discounted as the factor has Sot “teuled” in s0 short a coverage length, but they are of the same order. In fact the error, using the minimum Salue of curve (2, from the value given by the Fata”) curve of Fig, 24 is less than that ofthe Gnal value taken from curve (c) as can be seen. Curve (e) exhibits the fame close agreement as the others with an error of only about 0°25%. ‘These Sgures indicate a high degree of agreement between the values obtained fom the approximats anual method given by equation (18) and the more fract method based upon accurate computer analysis, burt shoulé be pointed our that a reasonable degree of factor “settlement” is only obtained when te total length of coil coverage is somewbat greater than the appro- priate semi critical length of approximately 1835V7t For each shell ‘This fact can be observed by examining curves (b) (4) and (e), The coverage lengths for curves (b) and (d)are just about equal 10 this semi critical length and the factor value is not setling, but for curve (e) the semi critical length is approx. 24'S ins. whereas the coverage is 30 ins. and the factor is beginning to sete but will, jn the ultimate when many more coils have been ap- plied, exhibit similar properties to that of curves (a) and (©). ‘The reason behind the slight increase in the factor at the left hand end, as mote coils are wound onto the layer. lies in the influence of the previously mentioned “sup: plementary” deflections creating defiections in the nega- tive sense. When the coil coverage exceeds the semi critical length by a reasonable amount the effect of such supplementary deflections becomes less evident and the factor seitles down. This cross check should by now have convinced the reader of the validity of the self induced factor equation (18), as applied to the first layer only, when considering the central areas of very long shells, but we must go much further than this. It is no good finding out what the loads are only over the central area, as this will not tell us what is happening at the boundaries and this must bbe known before any assessment can be made of the boundary stress. Quite clearly the results just given in Fig. 26, being based upon an inGnite shell, are not truly Fepresentative in as much as the first and last coils are allowed to deflect radially. Fig. 27 shows what the defiection diagram is like under the action of the loads relating to curve (a) of Fig. 26. Before determining the effect of boundary deflection restraint on the resulting factors it would be advisable to investigate the type of factor expressions that would ‘be obtained due to two or more layers. The resulting expression is in the form of a combination of both the self induced and mutually induced factors already de- fined and is only identical to the self induced factor when dealing with the first layer. The expression which ‘we are about to find will be called a “Combined factor”, 4,2 The combined factori"Fn” Referring back, for the moment, to equation (18) ie. Fotav)—if the shell thickness was to increase, leaving the rope size the same, the resulting value of Fatav) would be greater, which is what we should expect (3.7) If instead of increasing the shell thickness a layer of tensioned rope coils were to be wound onto the shell and then the value of Fs(av) for an additional ot second layer was to be examined separately we should find that the results would be very simliar. Why is this so? Tt should be remembered that shell deflection is pro- Portional to the nett applied load, in the infinite shell analogy, and for a given load, an increased shell thick- tess would produce a smaller deflection thus resulting in a smaller rope load change and a higher factor. We have, of course, decreased the Kris ratio result- ing in the higher factor In the ease of a shell already containing one layer of coils the situation is similar, i.e. as the second layer loads are apolied there is a reduction in the loads of the first layer coils thereby giving rise to a smaller additional nett deflection than if the first layer coils were absent, Tesulting in a higher self induced factor for the second layer. It is hoped that the reader is not being confused by the factors increasing as the layers increase when all the 2 time he is expecting them to decrease: it should be ‘emphasised that we are only examining the self induced factor for each top layer and, if the reasoning above has been grasped, it should be seen that this factor does in- crease. The combined factor, on the other hand, decreases as will be seen later. When a layer of coils is applied to the shell its efect is to increase the numerical value of Ks by an amount equivalent to the degree of “help” which this layer can sive 10 the shell when assessing the combined effect upon a subsequent layer. If the wire ropé comprising the first layer could be considered as a solid metal band across, and of equivalent thickness to, the shell then Ks would bbe doubled, i.e. the shell thickness would bave been doubled. In the case of actual rope coils. however, the true addition to Ks would be Kr and layers beneath the layer being examit 2 Kr ot Kr (n-1) where nis the number of the top layer. If there was only one layer then Kr (n-1)=0 and nothing would be added to Ks thus leaving equation (18) as it stands, Having established this fact we can say that the “self induced factor” for any top layer is given by: 14 0251 Kr Keel) kr (07556 Kr ) Rr as) assuming, of course, that this factor is being con- sidered separately from that ot the combined factor. The similarity of form between this and Dolan’s Factor equation should be observed. 4.3 Mathematical evaluation of Fr The following nomenclature will apply:— total number of layers equivalent unit pressure created by final layer coils as they are wound on. combined factor Load change (reduction) in final layer Load change (reduction) in under layers per layer. d= nett deflection due to final additional nett load on shell. Pd = equivalent unit pressure on shell produced by final layer after load reductions. Fon) = self induced factor for final layer (a.thayer). We are here dealing with changes in loads and de- flections over an infinite shell where each layer of coils is wound on under a uniform tension. Let us imagine an infinite shell. in radial equilibrium, having (n-1) layers already wound upon its surface under tension. The radial deflection is everywhere similar and the shell remains purely cylindrical. Imagine now that one more layer is wound on, under the same tension, making a otal of “'n™ layers: this last layer will be called the “final layer”. ‘As this final layer is completed the shell will deBect a further amount “a” radially but in doing so the under layers will be deflected an additional amount “d" there- bby creating a load reduction Lu in each layer. We have already seen that the final layer itself will sufer a load eduction due to its inherent self induced factor because ‘of the progressive nature of the coiling. This load redue- tion is LF.If PF is the load in the coils (converted to unit pres- sure) as they are actually being wound on then it can be clearly seen that the final additional load on the shell, after the final layer has been wound on, and equilibrium obtained, Pd=PF—LF- Lu (rl) z je general expression for the deflection of an shell under uniform load p is, ae aP (since Ei Ks) Fn ince but p=Pd=PF—LF - Lu (nel) land the additional shell defection due 10 the nex effect of the final layer load can be written as ds 2 [PF-LF-Lu : nr @-0) an ‘The chone in unit load Zu ean be expressed as and by inserting equation (21) in place of “d” in the expres: sion just given results in:— wae treater tan) i Kr ks tan. pp-X. rr-S Fe ig ee ss [+ Ze-n])-Zirr-oa (22) ) -- (23) Jsing equation (20) the combined factor for the nth. layer ean be written as: paw TE EP Lu (n=l) PF PF_LF Lu “See PF PF PF ‘The load change in 20 Substituting equations (23) and ( respectively gives: ) for LF and Lu [PF —(PF~PF Fs(n)) & Soon on 7F Cancelling PFs gives: Sf u-u-r Z-Ro) Frm -[1=Fe(n]— “t=1) 142-0 Pa which Sally gives:— kr (rE Fen) Fam Fs(n)\-—> eZ o-0) or simplified ay one Fam Fs (n) | 1 Zo] For the first layer only when 1: comes: the expression be- Fr=Fo{l) (1-0) =Fx(1) of as in equation (19) Finally we wi 1g 0513 Kr 4 h *Ke@-) Og no) OsS6 Ke reonn *KeG-D& se (24) Which will give valves of the combined factor for any layer, related to the mid points of long shells. ‘This factor equation is shown plotted on the graph of Fig. 28 with Fn as ordinate and the ratio Kr/Ks as abscissa and gives values for up to 10 layers. In use the combined factor Fn can be Used directly ig. knowing th in an it is simply a ‘matter of mul loa ue of Fn ihat yer and using he tepuas-a shell load, “Great Gare Should be iaken in the use of equation (24) as it will only provide factors which are applicable to the central areas of plain shells and therefore has limited use, especially so when compression rings are introduced. In view of the ability to determine loads far more accurately by means of the computer pro- grammes shortly 10 be discussed the complications of arriving at factor expressions for shells containing com: pression rings will not be introduced. Furthermore, such computer programmes will enable us to arrive at the loads over the whole length of the shell from boundary to boundary, even on shells containing compression, rings. Before proceeding to examine computer aided factor expressions for ropes coiled onto shells having radially restrained boundaries it would be profitable to digress for a moment and examine the behaviour of shell having elastic boundary supports. Further data relating to factor will be given in part 5 of this series.4.4 Flexible or elastic boundary supports When initially dealing with this most interesting aspect of drum design we must revert. once again, t0 the simpler load conditions of pans 2 and 3 wherein the resulting shell loads were considered uniform and constant all over. This makes for an easier approach ang a simpler understanding of the problem, 1 has previously been stated that if the ends of a loaded shell were allowed to rotate in the axial planes then the end fixing moment would be reduced. until Such time as simply supported end conditions were arrived at with @ consequent zero moment. If the boundaries of 2 shell were attached to flexible members, then, as the shell became loaded, these boundaries would rotate and ultimately sete down to an equilibrium condition where the boundary moment would, by itself, produce identical slopes in both shell and support members. Fig. (298) shows such support members which are seen to be in the form of annular fat plates securely anchored, or built into central bosses whereas the outer peripheries are rigidly attached to the shell at its boundaries, Fig. (29b) shows the deflected form of both shell and end support Quite clearly itis possible, by the use of circular fat plate theory, 10 determine the amount of slope at the periphery of the end support for any given value of applied peripheral bending moment. If this can be saute nome yt sila sop © and boundary moment for the sbell then the value of moment so found from this equality is the resulting moment from which bending suresses can be derived, both in the shell material and in that of the end support Having derived an equation, too complex to be ex- plained here, from which the’ shape of the deflection ‘curve for the end support plate could be obtained, when o) ig. 28. For this anaiysis the drum it considered as comprising ‘homogeneous shell forming an Integral part of both and suppor Diates and shiving bosses. The dellected form (b) ilustates the 2imianity of slope for both shel and end plate bounderies subjected 10 peripheral bending moments and shear forces, it was a simple matter to produce a computer programme to calculate these defections and so arrive It is also found that the slope = is directly propor- tional to the applied moment thus the slope/moment re- lationship can be represented graphically by a straight line passing through the origin. What is now needed is an expression from which the slope/moment relationship of the shell boundary can be constructed. If this latter relationship be plotied upon the same graph as that of the end support then the resulting moment at the boundary junction is given by the value at which the two graphical plots intersect. 4.5 Slope/moment relationship of shell boundary ‘Take the case of a simply supported long shell Fig. 30) subjected to a constant uniform load all over. At the ends where x=0 the following relationship exists: Bending moment sStope 2 OPS ie Er (25) To simulate a fixing moment at the end an anti-clock- wise, or negative moment must be applied. In the case of a free shell with no superimposed load or end sup- ports such an applied moment would create a deflection as shown in Fig. 31, where:-— , Tp vom 22 Manze EMOTE. (note flexural rigidity) AS the ends of the shell being examined are supported there cannot be any deflection Wy, at x=0 therefore to simulate this a shear force must be applied to bring ure riber cart, Matheny esr a set ara Fig. 30. This simply supported shel 1 provided with » unitorm Toad ail over in order to arive at simple mathematical boundery exoressionsWo Mo —. EL staan Fig. 31. & uniform moment applied to the thee end of an unves- luained shel! will produce # postive delection at shown Wy. back to zero. The shear force on its own must produce a deflection equal to Wy. ip the opposite sense and such a deffection is shown in Fig. 32, where: Qo IPD Equating these two deflections gives: Mo Qo "TRD IPD Qo=- 8 Mo = (8) The slope produced at the end of this simply supported, unloaded shell under the action of Mo and Qo above Woon o= 7) from which:— ae 1 He Tp Ah Mor Oo) . (9) In is now necessary to equate all the above equations (25), (26) and (27), ete. to arrive at the nett slope of } fully loaded shell having a moment applied to its end. Referring back to Fig. 30. dw oP ee OA oo QS) M, Now apply a moment and corresponding corrective shear thus— Applied moment=M, Applied shear =Q,=- @Mo w+ (28) Slope due to applied moment and additional correct- ing shear 1 2FD (2 f Mo~B Mo) from Eq. (29) TAD : = 30) The final nett slope is equal to the mathematical sum of the slope produced by the superimposed load on a simply supported shell (equation (25)) and that pro- duced by the applied moment and correcting shear force (equation (30) ). Referring to Fig. 33 the slope produced by the super- or Ea: 00) poduon + 6 a Wop +. 5 ut, Shall ni Fig. 32. A free Qo acting in the negative sense produces & ‘negative deflection atthe tree end of an unrestrained shel! imposed load (equation (25)) is represented by angle 'A” and in mathematical convention is —ve, The slope produced by equation (30) is +ve and represented by angle B.* The nett slope is represented by the sum of angles A and B as shown and since the slope, generally, is given as aw Fen) tan 6 then, in this particular case tan A+tan B tan Atan B de Fy net mtan ((— A+ B= but since. aw pF tan 4a aya 9 tan B= (5)= — MO rs 28D the nett slope becomes:— pe Mo SM (nett) mF 24D 1p he. Me BD ng substituting ———. for D the expression becomes:— oer Ps zt Mo+p fs) Gn 2p Mo EX It is significant to note that, when dealing with very small angles, the denominator in the last expression 6 very small and for the ease of the practical example to follow its aumerical value, for unit moment, would be 1=51x10" In view of this the errors would not be very large if the denominator was considered az having 1 numerical value of unity and the final expression for the nett slope becomes:— ae 2 © ete 2 pp Mo ee Om ERM Mo+7 Al = GD Equation (31) is of the form x=—Ky -C’ which produces a siraight line plot but not passing through the origin. ‘AS a simple check, if equation (31) is equated to zero the resulting expression would be the moment at the boundary of a fixed ended shell. The expression becomes Mo= - Which is correct (see equation (8) of part 2), Fz (vet 1F Mo becomes zero the resting expressions —p 2 which is, of course, the slope for zero moment on the shell of Fig. 30. (Equation (25) ). ‘All that is necessary is to plot both slope/moment relationships for the end support and the shell to obtain the resulting boundary moment. 4.6 Practical example of elastic ends By way of an example let the shell shown in Fig. 34 be analysed. This is basically the same shell as al- ready used in the previous examples, except for the = 30 -boundary conditions. AS the slopermoment relationship for the shell boundaries produces a straight line it is only necessary to calculate the slope for zero moment and the moment for zero slope and join the two values on the graph, ‘The slope for zero moment is 374% 3,600 x0-1173 30x 1082 . fixed end conditions) is 2 374 IR” Irons 13,580 Ib ins/in of circumference ‘The straight line joining these two values has been drawn on the graph of Fig. 35 where the moment is plotted as ordinate against the slope as abscissa, To make the problem more interesting the’ slope/ moment relationships for a succession of end support plates of varying thickness have been drawn from the origin, the thinnest being 2in varying in jin increments up 10 2in thick. This will provide us with a range of stresses, in both shell and end support, from which to choose. Each line of the individual end support plots inter- sects that for the shell at specific points which each indicate the slope and the moment’ applicable at the boundary junction for each particular plate thickness. From each of these moments we can determine the siress in both shell and end support material. Such stresses can be put into the form of a continu- ous graph by plotting both as ordinate against the end support plate thickness as abscissa. This is shown in Fig. 36. It can be seen from this graph that for a given ‘drum geometry and shell thickness we can choose what ever plate thickness required such that the resulting stress is below the fatigue limit for the type of welded joint used. Wis possible that the reader may be rather confused concerning the inter-relationship existing between both the boundary slopes and bending moments. Quite ob- Viously, for a shell mounted on flexible boundary sup. Ports, an increase in the supericial Toad will create a larger boundary slope because the end support boundary has been rotated to a greater extent thus requiring, and Producing, a larger bending moment at this point. This is, of course, shown by the siope/moment lines relating to each of the various end support plates. How then can Wwe reconcile the inverse relationship of the shell bound- ary slope/moment line in Fig. 35? ‘The full ine drawn for the shell boundary only, represents the value of the moment exiting for any given slope at that point for one particular value of super- ficial load. For example, if the end plate was infinitely thick then its slope/moment line would pass vertically through the origin and no matter what the valve of the load the slope would remain zero i. the end plate is inelastic. Similarly, if the end plate was infinitely thin its slope'moment line would pass horizontally through the origin and therefore no moment could be produced at any time, Let us select the 2in side plate as a further example. an be seen that, ifthe loaded shell is in such a state shat boundary 8h sm ——— Fig. 33, Rotations of the shell about » boundary produced by ‘he individual forces mentioned inthe text S Catia Lr cvaies) a Fig. 34. Physical data relating to the results given in Figs. 25 tnd 28 that its boundary slope is 0-00173 the moment produced would be 4,650 Ibs ins/in of circumference thus equating itself exactly with the 2in plate under similar circam- stances, What now happens if the load is changed whilst still retaining the 2in side plate? Assume that the original load of 374 Ibsfin* is halved then both the slope for zero moment and the moment for zero slope of the shell boundary would be halved as each is directly proportional to the load. What we should expect is that the flexible boundary moment and slope have both been halved. Tis is evident from Fig. 35 where the dotted line is the new slope/moment line for the boundary of the shell having the reduced load. ‘The intersection with the 2in plate line is now af 000086 and 2,325 respectively which are half the pre- vious values. For this particular case, using a 2in plate, the true slope/moment relationship of the shell boundary is identical with that of the 2in plate ie. zero slope and zero moment representing an unloaded shell. 4.7 Analysis of results Ithas been mentioned previously that it is not always possible to arrange in the design for a free ended sbell connection and the only alternative is to rigidly attach the shell boundaries 10 the supports. If the supports Were inelastic it has been shown that for this particular Shell geometry the boundary moment produced had a modulus of 13,580 Ib ins/in of circumference which repre- sented a stress of 20.400 Tbyia®. far too high for safety when it is remembered that this is only the frst layer. Examination of the graph of Fig. 35 shows how drastically the end moment is reduced by supporting the shell on flexible members, In using this type of construction itis found that the design criterion is not the shell boundary stress but the nd plate boundary stress as can be seen from Fig. 36. ae4gure The end plate stress is only numerically equal to the Jell stress when both nave identical thicknesses ie) 2in but even so the stress is very conservative at about 7,000 Ibn’. A 2in sideplate would be considered far ‘100 thick as a structural feature, therefore a thinner plate would probably be used,.The Lin plate produces a very modest stress of 5,000"lbjin* which, from the shell aspect. is, virtually a simply supported condition produc ing only'{,250 tb/in® stress in the shell material ‘The reader may feel that we are at last getting some where towards a solution to the problem of overcoming the possible high boundary stresses purely by introduce ing @ flexible end suport of the type just illustrated This is true, up to @ point, but cenain other aspects must be considered. Normally each end support plate is stiffened by radial arms to provide some stability to the structure as a whole. It would be clearly inadvisable to omit the stiffeners altogether unless some other means could be Provided 10 prevent the drumshell from swaying from Side to side along its axis, or unless the sideplates were thick enough or had sufficiently limited annular depth 0 overcome such a tendency. It will be found in prac- tice that if a lin or din thick sideplate was to be used for purposes of economic construction then a much smaller annular depth would be provided which would uutomatieally stiffen the flexible portion of the sideplate such a5 10 produce a slope/moment line approaching, for even exceeding, the line drawn on Fig. 35 for a 2in “ite, This would ‘automatically alter the stress diagram Fig. 36 and produce higher sideplate stresses all round. 1\ should be appreciated from the graph of Fig. 35 that as the sideplates get thinner and thinner the effect upon the shell boundaries approaches that of a simply supported condition ie. the boundary moment ap- roaches zero. Conversely as the sideplate thickness approaches infinity the effect upon the shell boundaries approaches a rigidly clamped condition which allows the evelopment of the full shell boundary fixation moment with zero sideplate stress. No hard and fast rule can be given here with regards to the best thickness and annular depth of the flexible portion of the sideplates as the variations are infnite and each design must be prepared with a view to the final requirements in respect of economy, duty and 20 ponte + Fy fatigue life stress limits Many readers will have appreciated, by now, that the fundamental conception bebind the analyses so far pre: Sented lies in the term “symmetrical deflection”. At no point in this work bave we mentioned any structural member which, when applied to the drum, will prevent symmetrical defection of either the shell or the exible portions of the sideplates. The classical error of con- Struetion, especially with respect to fabricated structures, is to provide radial stiffening arms to the sideplates and to run each arm up to the under surface of the shell where it is rigidly attached by welding to the shell plate, Nothing could be worse than this as it prevents the shell from deflecting to a circular form thus induc- ing circumferential bending in addition to the already familiar axial bending: it also provides the shell boundaries with local rigid fixation thus inducing high local axial bending moments, the very’ thing we are tying to overcome. Under these circumstances it will bbe quite clear that stress analysis would be highly come plex indeed and may lead to uneconomic structures of ‘otherwise dangerously high local fatigue stresses within an economic structure. The simpler the structure becomes the easier it is to arrive at the principal stress levels and generally it is far more economic. This is what we have arrived at by the introduction of flexible end supports. Much more has to be done, however, before the true solution is obtained as this has been only a glimpse of the simpler aspect of design. Because we were interested in treating this problem mathematically the loads over the shell have not been factorised and structural stability hhas not been allowed for in the selection of the side- plate geometry. except in so far as to outline the importance of it 8 Restrained boundaries and the effect upon load factors Mauch has so far been written concerning rope load factors and the computer results have already cor roborated the first layer self induced factors, as cal- cculated using equation (18). This corroboration was, hhowever, based upon results obtained using an infinite shell having no radial restraint whatsoever (See Fig. 26) ‘Will the factors remain the same over the central areas of a long shell when radial restraint is applied to the Fig. 35 (let). Showing slopelmonvent inter- {20tr for shell and end plate Boundaries, for lend plate of constant annular depth Fig. 36 (right). Shell and end plate boundary ‘Stresses Tor variations in end plete thicknese 4nd for end piste of constant annular depthFig. 37. Detlection diagram for shell having facially resvained laste boundaries. 3 ‘adius=60in. Shell thicknestm2in. Shell length=85in nom. = 30% 10% Ropem Thin dia, stranded. Pitch coetficient= 1-025. Initial ‘ope pull at coil Wo. 1=33.130 1b, Single layer onty. Ke IKsm 0-165 Fig. 38. Bending moment and shear force Giagrams for shel date of Fig. 7. The shear curve is an average value. The tue shear {force diagram aernates a8 shown. in par, atthe leftshand and a) Fig. 29, Rope load factors for shel of Fig. 37, ‘AS this 1s only the frst layer these values 27a the sel induced factors F5 Shes ace 1/ino cramtounce boundaries? In order to investigate the factor behaviour over the first layer of a practical shell the authors have produced 4 computer programme which does not depend upon the same method of analysis as the very early programmes, This later work does, of course, use fundamental shell theory but takes into account the progressive nature of the coiling, in that the coils are applied one at a time in the same way as was used in the programme pro ducing the factor values in the graph of Fig. 26 This later programme is far more sophisticated in as ‘uch as the loads are applied at the true coil pitching across the shell and values of slope, bending moment and shear forces can be obiained at any point, Variable boundary fixing moments are also catered for. The major contribution made by this programme is for the allow nce of up 10,6 layers of rope on the rum and a maxi mum of 70 eéils per layer. We now have available a programme that, despite being based upon statical theory, will behave in a dynamic fashion in that it is behaving, in time, exactly a8 2 true drum would behave. The time scale is, how. ever, considerably extended as the computer running time is very much greater than the time normally taken to complete one winding cycle. The authors feel that this programme provides a ‘major advance in the field of winder drum design and virtually all the work which now follows will be based 33 upon the use of this programme and variations of it, This programme has been called “Research Programme No. 14." In order to discover whether restrained boundaries will affect the factors over the central areas the para- meters relating to the shell, which has featured through- ut this work, were used'as data and the number of layers restricted to one only. The rope used was Ijin dia, triangular strand with an assumed stretch modulus of 15 x 10* Ibjin’. The coil pitching was taken as 1-035 x fin, i.e, the rope pitch coefficient =1-035. and the coil loads were arranged to act at this pitch ‘across the shell. The full set of results obtained from the computer, using “Research Programme No. 14,” are resented graphically in Figs. 37, 38 and 39. Confining ourselves, for the moment, to Fig, 39 which shows the factor values, it can be seen that comparisons between the value of factor given here and that of Fis. 26 shows quite remarkable agreement in that Fig. 39 Shows a settlement at around 09353 and Fig. 26 shows 0-930 whereas equation (18) provides a value of 0-926, an overall error of about 1%, Despite radial restraint at the boundaries there has been no significant change in the value of the frst layer Self induced factor over the central area of this Jong shell. Before proceeding to give proof of the combined factors it would be interesting, at this stage, to analyse 03k meaTne : at ne Fig. 40. Showing col lond factors foreach layer 24 its completed, KLIKS for this shell= 0-172, Twenty coils of fin de. strended rope the results given in Figs. 37 and 38. The shell deflection is shown in Fig. 37 on which. for the first time, can be seen the acival values of the slopes did for each boundary. These will be useful later when considering the end support members The maximum defection occurs at coil 18 and is 2220 thou. whereas previously (Fig. 19, Part 3) a cen- tral deflection of 23°5 thou. was obtained for a uniform load of 374 Ibjin’. Multiplying 23:5 by the factor just found, ie. 09353, we obtain 220 thou, which is al- most correct. It should be remembered from previous work, however, that variations io end restraint produce marginal changes in central deflections. Fig. 38, showiag the bending moment and shear force umes for this shell, produces some very interesting re- lus, As will be seen from the defection diagram the boundaries are neither fixed nor free but have been Provided with an end restraint which produces a bend- ing moment of ~6.582 ib infin of circumference which is approximately aif that for a fixed ended shell (=13520). This has of course reduced the boundary stress” in ‘the shell material from 20400 Ibjin® 10 9.900 Nb/in® which is a far safer igure. The shear force has also dropped from 3186 to 2.121 Ib/in of eircum- ference. This is of course the natural outcome of bav- Fig. 41. This graph itustrates the continuity ofthe factor exares: {ons om the computer results. Values relate 10 the central areas (Of the shell a indicated in Fig. 40 3 ing a reduced bending moment. By providing flexible boundary supports of a given flexibility it is possible to obiain maximum bending and compressive stresses in the shell and maximum bending stresses in the end support members, all having roughly the same magni- tude and all well within the safe limit The shear force diagram is rather interesting in that the true shape has been shown at the left hand end of the graph. It consists of alternating vertical lines, the length of each being equal in magnitude to the value of the coil load at that point. Each vertical line is joined to its neighbour by a diagonal line. This is exactly analogous to the bending moment diagram exhibited by a plain beam subjected to a multitude of concentrated Joads but with one exception and that is, bn a beam the joining lines are, of course, all horizontal. Shear force diagrams will ‘not generally be plotted in this way as it is tedious and can be confusing but it may become necessary later on when dealing with shear force diagrams for shells containing compression rings. 4.9 Proving the combined factor equation Fig. 28 in the previous article has already shown how the combined factors Fn vary for variations in Kr/Ks, the relationship being dictated by the computation of equation (24) over a wide range of parameters. The first layer factor has already been proved but f the second and subsequent layers? Research programme No. 14 can only accept a maxi- mum of 6 layers of rope and so it is only possible to obtain comparisons up to this number of layers. An entirely random choice was made for the shell used as a checking case but care was taken to see that it was a “long” shell to allow enough length for factor settlement. Details of this shell are given in Fig. 40. Programme No. 14 was run so as to “wind on” 6 complete layers and the resulis are shown in Fig. 40 where the combined coil load factors Fn are plotted as ordinate against coil numbers as abscissa. In all six cases the value of the factor has been chosen relating to the “high points” over the central areas (as has been done in the previous analyses) of each of the plotted curves. AA further graph has been drawn (Fig. 41) in which the factor value Fn is plotted as ordinate against layer num- bbers as abscissa. The full line shows the variation in factor values obtained from the plotted computer re- sults of Fig. 40, whereas the short dotted line shows the equivalent values caloulated from equation (24). A re- markable agreement is immediately noticed, As a further point of interest Dolan's factors have been introduced (long dotted line) as a comparison and it is clear to see that his factors are in error to the ex- tent of roughly 649% on the first layer. As his first layer factor is considered as being unity this is not surprising but in any case the error is not too great. ‘The main error in Dolan’s factors lies in the absence of “self induced factor” for each layer: in other words, if each of Dolan’s factors was to be multiplied by the author's own self induced factor for the appropriate layer then the results would compare very favourably with the results in Fig. 40 obtained from the computer, It should be quite obvious now that equation (24) is as near accurate as is required for checking the value Of the factors over the central areas of long shells. but equation (24) can only produce one value per layer, Ttcannot tell us what value of load exists near the bound. ary. it cannot tell us the boundary support reactions of the bending moments or shear forces. nor can it decide for us what type of end support is required. Only “Re- search Programme No. 14” can give us the complete answer, 5.1 The introduction of compression rings So far throughout this series compression rings, or stiffening rings as they may sometimes be called, have only been mentioned briefly in passing, If the reader would refer back to the photograph (Fig. 5, Part I in the December, 1966 issue) he will see the exact nature of, and the form taken by, compression rings in a practical shell. Why ate compression rings used? We have already seen that a sbell, in order that it may safely withstand ‘any superimposed load, must have a minimum thickness dictated by limiting design stress limits. In view of this then, what is the purpose of introducing such rings when a plain shell of appropriate thickness could be used? There are, of course, various reasons and these will now be outlined as follows: Supposing, for example, that preliminary investiga tions showed that @ heavy duty winder drum, such as those typical of S. Africa, demanded a plain shell Sin thick to comply with the safe compressive stress limit (ignoring other stresses for the moment). It would be extremely Unlikely that such thick plate, in the sizes required for such a shell, would be obtainable and, furthermore, there would be a problem in rolling it 10 the required radius. Faced with such circumstances there are two alternatives firsly, resort to a cast steel Grumshell and secondly, resort’ to thinner plate and supplement it with compression rings. As this series of articles is aimed primarily at fabricated drums then the second alternative is the only one with which we should bbe concerned, Some readers may well ask, what is wrong with a Plain cast steel shell Sin in thickness? The answer is, of course, nothing at all except that the designer may bbe faced with rapid changes of material thickness at the ends of the shell, which is not at all desirable in steel casting, and also that the design may tend to appear disproportionate or in other words it will aot “look right” Let us return to the second alternative above. It will bbe ‘necessary to choose a shell thinner than Sin but how thin? This is not an easy question to answer in 4 few words as there are so very many factors which can influence the decision. L- Bop ip dete ae SB 16% vw br 984 Ou 6 AL & uM ¢ 35 Figure 42 will illustrate this point admirably, On this sraph there are six curves plotied with radial deflection as ordinate against coil number as abscissa. The results are obtained from “Research Programme No. 14” the data for which was related to the shell used throughout this series. Only one layer of rope has been “wound on” although this shell is destined for 3 layers as mene tioned before. In producing this graph the main object was to outline the variations in maximum deflection Produced by various combinations of ting sizes and sitions such that the soral cross sectional area of each Sheli/ring combination was the same, AS a reference, curve “{” is the deflection applicable 0 a plain 2in thick shell with no compression rings Whatsoever, the total cross sectional” area being 698 x 2=1396in*. The next step was to introduce fone ring only of 29-72in* situated under coil number 24. Curve “a shows the resulting deflection which, from the compressive stress point of view, shows mar- inal gain only over that of curve “f."The total cross sec: Honal area in this case being 139-6+29-72= 169:32in: Two tings were next introduced each having a cross sectional area of 14-86in* (curve “b") then 3 rings with an area each of 9:91in? (curve “¢"), There is, upon examination, little to choose between curves “b” and “c” unless a difference of one tenth of thou. is going to worry anyone, in which case, curve “>” with only two rings:is the marginal winner, Results for four tings of 7-43in? were computed but aot plotted as it would have created a confusion of lines in this area Dut the absolute maximum deflection was 17-994 thou. If now a little thought is given to these results it should be clear that there must be an optimum condi- tion which will produce safe stress limits for the most economical iayout. on pepsin hh en es dt Fig. 42. This graph ilustates the variations in deflections for various compression ring/shell combinations.” The combingg {ota/ cress sectional area of ring/shell arangemants “s° tava ‘inclusive we each identical. Shell "T" ie the basic 2in thick shel ‘wih 0 comoression rings. Shell “f” (dotted) is « plain shel ‘ving ihe same total cross sectional sree at shel “oe “pe, “e" and “a”. Exch arrow represents the poshion of « compression ‘ing. Only ane layer of coils hes Been noplioeSuppose it were possible to introduce an extremely large number of very tiny compression rings evenly paced across the shell. This would be almost, but not guite, analogous to having a plain shell whose total cross Sectional area was equal to that used in curves “a", bv and “c", etc. i.e. the plain shell would be 169-32) 69:8=2-427in thick. From a compressive point of view the muhitudinous ring analogy is quite sound but from a bending stress point of view it is not as the values of # and the Mexural rigidity D must be related to a 2in thick shell. Curve “d” indicates such a deflection, the maximum being 18°58 thou. Imagine now that it was only possible to obtain one- inch thick plate for the shell. We should not despair as it is only a matter of adding sufficient area of ring to make up the same total cross sectional area as before, of is it? We saw from curve “b" that two rings were Adequate so let us apply two rings 10 this thin shell, each having an area of 49-75in*. Curve “e” indicates the devastating result. There is a maximum deflection of 37°85 thou. which, in this shell, produces a compressive jo, _StteS5.of, 18,925 Ib/in* for only the first layer! wes This unfortunately is not the worst, The maximum "Bending moment occurs over the first ring and is ~ 5,188 © Ib iniin of circumference, a stress of 31,200 Ibs/in®. This is clearly ridiculous and amply ilustrates The fact that it is not just the total cross sectional area that marters but the way in which it is arranged. If we try to analyse what is happening it is seen that the all_ important semiccritical length is playing a large part, The semi-critical length for the one-inch thick prell at 60in radius is 14-2in thus the critical length is ‘vice this i.e, 28-4in and the rings are spaced at 22-7in This situation is almost identical to having a fxed ended shell of length equal to the critical length above where the end fixing moments are ~5,900 Ib infin of cit- cumference and a maximum central compressive stress of 19,450 Ibjin* for a uniform load of 374 Ibjin? sub- jected to the appropriate first layer facior. The actual central compressive stress from curve “e” is 17,200 Iofin* which is lower than that just calculated as the two Tings ate in fact closer together than the critical length by about éin, What, then, is the best combination of ring layout that can be chosen from Fig. 42? It should be remem- bered that the critical length for the basic two-ineb thick shell is very nearly 40in and in thé ease of a single ring (curve “a") it is situated some 35in from the boundary point, almost equal to the critical length. Crawford states in his article that there is nothing to be gained, in the compressive sense, by inserting rings at a spacing greater than 4drt inches apart and this statement is quite correct as curve “a” illustrates. The most economical layout, both in terms of the number of rings and hence the amount of welding necessary for their installation, would be that shown in curve “b" ie. two rings only, each 1486in* of cross Sectional area. The spacing here is 2275in which is just about equal to the semi-critical length of 20in and the central deflection is one thou. less than the maximum V curve “a”, 5.2 Bending moments at the rings ‘The less the number of compression rings in a shell of given toral cross sectional area the greater will be - 36 Fig 43, Plain equivalent thickness shell section the bending moment in the shell material adjacent 10 the ring. This is particularly wue near to the central areas and depends upon the ratio of sbell metal to ring metal. Curve “a”, Fig. 42, produces a bending moment of ~ 6,320 Ib injin of circumference over the single central ring, almost identical to that at each boundary. The maximum bending moments over the rings of curves ’b” and “c” are ~3,150 and -2,079 Ib in/in of cir- cumference respectively. The moment, in curve “b” being approximately balf that of the boundaries. We have of course already seen the drastic results for curve “e™. From the point of view of the boundary moments these are of course influenced by the number of rings in the shell. For curves “a”, “b" and “e” these momeats are ~6,748, ~6,413 and "6,018 Ib in/in of circum- ference respectively and the shell containing four rings produced ~5,900 Ib in/in of circumference. The plain shell, curve “f", is subjected to a bending moment of ~ 6,583 Ib in/in of circumference at the boundary. The last five values of bending moment show only nominal changes and it would thus appear that the number of rings introduced, providing the total cross sectional area is the same, have only a negligible effect upon the boundary moments. We should now be in a position to consolidate the facts so fat presented and thus create a basic law for ring spacing, which can be stated as follows:— @ Having ascertained the total cross sectional area of ring metal required it is best to ensure that the spacing between the rings is no greater than the semi-critical length for the shell portion re- maining. Gi) The total cross sectional area of ring metal must be so related to the cross sectional area of the shell metal such that, with the ring spacing chosen, the bending stresses in the shell material Adjacent to the rings and the compressive stresses in the shell material between the tings are within the safe limits required of the design. Rings positioned nearer to the boundaries than approximately 1°5 rt. ins, serve no useful pur- pose in relieving boundary stresses or in reducing central area deflections unless they are very large in which case they may develop large axial bend- ing moments in the shell material. The above three simple rules rely upon the fact that, whatever the number of rings placed in a shell the total ‘ross sectional area of the ring/shell combination re- mains constant. It is possible, purely by introducing extra large rings near to the boundary, to reduce the boundary stresses. This aspect has not been dealt with as it is uneconomical to do s0 and completely unnecessary if the boundary stress can be kept below the safe limit by the methods already described. Although the above three simple rules are quite valid it is not correct to base the ring/shell area relationship fon the equivalence of total cross sectional area. It ismore correct to base it upon what is termed the “equiva- lent thickness” of 2 shell which will be explained more dealing with the practical design aspects. ‘d” illustrated that the maximum defection was 188 thou, whereas curves ‘e” and the une plotied results for four rings showed that the maximum deflections were less than this. The equivalent thick- nesses for the shells in curves “b” and “c” and for the four ring shell are:— 2653in, 2-S8in and 2-522in re- spectively whereas the plain ‘shell, whose equivalent thickness is its own thickness, is 2-427in. The area of the rings are all a little 100 great in order to create, with the spacing given, an equivalent thickness equal to the plain shell therefore, in a properly designed drum, there is a slight economy to be gained by using the equivalent thickness technique. 5.3 Rings as an aid to circular stability The second reason for introducing rings can really be divided into two parts, (a) The improvement in stability of a “free” shell during machining. (b) Circular stability in service under load. This really introduces the problem of the elastic collapse stability of the shell as a whole and will be dealt with under a separate main heading. Let us first of all deal with (a) under this main bead- ing, Jn some of the older type constructions, and indeed in one method already outlined by the authors’, the shell does not form an integral part of the end supports in that it is subsequently secured to the end supports, by some means, on final erection. Very often such a shell is subjected to machining after rolling and one of the difficulties lies in adequately sup- porting such a relatively thin structure as a plain shell such that accurate machining can be carried out. A farther difficulty arises out of the need for stress relieving the finished fabrication which presupposes extensive ‘welding operations, having been carried out on such items as compression rings and flanges, etc. Almost without exception, on free ended shells, it. becomes necessary to introduce auxiliary bracing, of the type shown in Fig. $ to prevent undue distortion of the shell during heat treatment but in the shell illustrated this bracing is relatively simple and serves mainly to contain the shell joint diameter at its correct value. The first machining operation, after fabrication and stress relieving, is to plane the joint faces of each half shell then each half can be bolted together to form a whole. For thin shells containing no compression rings, or Ranges, quite elaborate bracing may be required to ‘maintain it in its circular form but the inclusion of rings overcomes this requirement and may mean that no additional bracing is requited over and above that which ‘as originally introduced to prevent heat treatment distortion. 5.4 Circular stability under load It is a well known fact that some winder shells, when deflected under load, are no longer circular ie. they tend. to go “out of round”. One case has been reported where this has happened on a shell in which rings were actu- ally fitted but they were of such a shallow depth, in the aT radial sense, as to contribute virtually nothing to the stability. In order to be of any use compresion rings must be fof sufficient depth 10 maintain the shell in a circular form and this is even more important in a machined cast shell due to the possibility’ of variation of metal thickness resulting from the casting of such large mem- bers; such variations can induce out of roundness under load. This is, however, less important in rolled steel shells, where the original plate thickness is reasonably constant, but if poor machining is carried out this effect could still be obtained. For a ring to be of any use in this sense its intrddue- tion must be made to materially increase the cross sec~ tional moment of inertia of the completed shell. A few simple figures will lustrate this. Take the case of a shell 2in thick and having. for the moment, no rings. When rings are introduced they will be placed at say 20in centres. This length of plain shell will be compared with a similar length of thinner shell containing one ring such that the total area of metal is the same in both cases. Fig. 43 shows the plain shell. ‘The moment of inertia of this section about the neutral 20x23 axis N.A. is =13-333in¢ If the shell is now reduced in thickness to 14in and a ring added to make up a total of 40in* the following is obtained Fig. 44,” The moment of inertia of this particular section approx. 106in‘, which is a big increase. It can be seen that the effect of rearranging the metal of a shell in the form of rings improves its circular stability as it is far more difficult for it to be bent out of round than the equivalent plain shell due to the large increase in the moment of inertia of the shell section. This is, of course, quite elementary but has been included to ilustrate the great advantages that can be gained from fitting rings. It also demonstrates the “equivalent thickness” technique, ‘One question yet remains and that is, knowing the cross sectional area of a compression ring what should be the relationship of width to depth? ‘Very thin deep rings, are not considered good practice unless the free edge is stiffened to avoid the possibility of edge buckling. This immediately adds cost by virtue of having more welding to do on each compression ring, unless some special section can be found which would comply with the requirements, On the other hand very wide rings, with a radially shallow section, are not very satisfactory as they con- tribute fttle to circular stability. There is also a danger of overstraining the ring to shell connection welds by virtue of the variation in deflection of the shell across the face of the ring. Such a connection should be fairly narrow to avoid this possibility —————- Tyas were LL ren Fig. 44. Eauivaa shail with ingBoth authors favour some compromise whereby the proportion of depth 10 width is approximately 3 to 1. This gives a sturdy ring which is very simple and easy to instal! with the minimum of welding. Knowing the area, how can the ring proportion be obiained? By now the reader will be aware of the fact that, after being deflected, the material in the ring will bbe subjected to circumferential compression and each ring will be exerting an outward force upon the shell in the same way, but in the opposite sense, to that force created by a tensioned coil of rope. Let us imagine the ring removed from the shell bui subjected to the same radial deffection. This deflection can only be obtained on the free ring if a positive um- form load was to be applied to its outer surtace and of value equal to the interactive force between ring and shell whea the ring was acting as part of the shell. Such load is given in the output data produced by “Re- search Programme No. 14” and it is this load which must be used to determine the ratio of depth to width As a practical example it will be necessary to use results relating to our standard shell but containing three compression rings each 9-91in* cress sectional area and having three layers of rope wound on. The maxi- ‘mum ting reaction is ~3,477 Ib per inch of circum. ference. Expressed as a maximum compressive force over the ring section this is ~ 3,477 x60= ~ 208,420 Ib and as a compressive stress in the ring material = 208,420/9-91 = 21,000 Ibyin’. ‘The minimum compressive stress which would cause {BS cellapse into wo nodes on an unrestrained plain 8 is given Where:— or crmcritical stress in the ring (compressive) E=Young’s modulus Poisson's ratio Teradial depth of ring r=mean radius of ring (usually taken as mean shell radius) Let us suppose the ring to be Sin deep then:— 30x108 ( $ 7 For a ring lin deep cr=229,000 Ibjin*; but yield point for this material is 38,000 Ib/in* therefore a ring Sin deep would be perfectly stable up to yield point thus, the ring in the practical example, having an area of 9-91in? would have a factor of safety of $7.250/ 21,000=273 10 1 on the theoretical collapse pressure but 38/21=1-81 10 1 on yield, The proportion for this ring would be Sin radial depth by Zin thick and would be perfectly safe and elastically stable in its own right; it would also be arrow enough to preclude the possibility of any undue YR strain during shell defection. Equation (32) can be rewritten thus:— APO fEE Joana 120) inches but if, for mild steel, »=03, £=30%10* and oer is assumed 10 be 63,000 Ibyin* (UTS. for mild steal) the above equation becomes:— T=0-0874 inches radial depth... (33) Tae reason for choosing 63,000 Ib/in® for © or gives the designer a good factor of safety on the theoretical col. lapse stress and ensures that yield stress is reached well before this, it is then only @ question of ensuring that the compressive stress in the Ting, as found from Re- search Programme No. 14, gives an adequate factor of safety based upon the yield stress for this material. The compressive stress in any ring is a measure also of the compressive stress in the shell at that point thus the two become compatible. * If equation+(33) had been used to find the critical ring depth, in the practical example, it would have Produced: — T=0-0874 x 60=5244 inches which does of course allow for the development of 63,000 b/in* critical stress instead of $7,250 lb/in® as found for a $ inch ring using equation (32). It should be clear that the ring size chosea has en- abled us to choose a thickness which is equal to that of the basic shell, namely 2in. This shows that the ring size and proportion has worked out very economically as the ring segments can be cut from similar plates to that of the shell. If shell machining is being catried out then obviously the basic shell plate thickness would be at least 2pin and possible 24in thick and it would be wise to investigate making rings trom this basic material thickness but not at the expeuse of sacrificing radial depth because of the increased thickness that is obtained. It must be remembered that it is the radial depth which determines the critical collapse stress and ot the thickness, A great deal more could be written about compression rings but as space does not permit, it is felt that the Preceding text has at least outlined the major points arising out of their inclusion in a shell. The very keynote is simplicity, as it has been throughout this work, and the less complicated a compression ring becomes the cheaper it is to introduce. 5.5 Critical collapse pressure for plain shells For a very long. or infinite, plain shell, the critical Pressure to cause collapse into’ two nodes (i. the shell would collapse into an elliptic form initially, with » very rapid ultimate collapse of the minor exis. The ultimate collapse would produce two lobes) can easily ‘be found from a revised form of equation (32) i.e. Pee + lbiet rere using the parameters for our usual shell gives:— 301082 "FeO With three layers on this shell the approximate pressure, as found from the Research Programme, was 860 Ib/ia! Producing a compressive stress of 26,560 Ib/in* in the 1BU mee, Ge Per $305 Ibjin? penshell material (This was a plain 2 inch thick shell with ‘no compression rings). This does of course look disastrous but two points’ should be made, (a) Equation (34) gives the minimum critical collapse pressure for a very Jong shell but, in the practical Shell, the end supporis have a marked influence in increasing the critical pressure. True collapse, as we know it, by subjecting the shell 10 2 uniform gas pressure which does not change in magnitude throughout collapse displace- ‘ment, could not possibly take place. In the case fof a shell containing tensioned coils of rope, as soon as collapse started to take place the rope tensions would change thus alleviating the collapse condition, Despite what has been said in (b) above it is still ood practice 10 ensure that the shell is constructed such 8s (0 preclude the possibility of collapse ie. we should imagine the pressure produced by the rope coils as remaining constant over the whole surface when co:- lapse conditions are reached. How do boundary conditions affect the collapse pres- sure found from equation (34)? The analysis of the theory of elastic stability is very involved and will not bbe undertaken here but sufficient information can be Gerived from sources of published works to enable us to understand this aspect of shell theory a little more clearly, ‘The critical collapse pressure for a shell of finite length is given by:— (o Et eT e 2nni-r oe te (SS) n= 1) (Lea? Bj? Py 12? Len? Pia? 73 bj? G5) Ih can be seen that the criti Wepends not only upon the us met but also the shell length “I” between boundaries and the number of nodes “n” into which the shell col- lapses. By ringing the changes between parameters it is possible to construct a graphical plot consisting of a series of lines each for various ratios of t/2r with the critical pressure as ordinate against the ratio I/2r as abscissae. One important feature about equation (35) is that it ‘can be used to calculate the critical collapse pressure of 4 shell irrespective of whether the ends are built in or simply supported as it is found that the mode of end restraint has little effect upon the value of the critical pressure. If we introduce into equation (35) the para- ‘meters relating to the shell featured throughout this work it is found that the minimum, critical collapse pressure occurs when n=7 but has a value of 8.950 Ib/in? over the outer surface! This pressure would produce a com- Pressive stress im the shell of 268,500 Ib/ia® which is well above yield stress! In general this is typical of most winding engine drum sbells of a practical nature where- in the compressive yield point is reached long before there is even the remotest chance of reaching critical collapse conditions. For this particular shell 1/2r would have to be 3 or more before the critical collapse pressure approaches 1 collapse pressure parameters already 39 the pressure produced by 3 layers of rope i.e. it would have to be 30 ft long instead of the existing length of 5 ft Sia, : It is fairly safe to say that for all normal sized winder drums there is absolutely no danger of elastic instability occurring before yield point has been reached. 5.6 Brake paths Unfortunately space does not permit as full a dis- ‘cussion as one would like concerning the effect of brake paths upon the sbell and boundary support members. It is assumed in what follows that each brake path is a pure extension of the cylindrical shell beyond the ‘normal boundary point and that for ease of explanation there are two such extensions on each drum shell, one at each end, and of identical length. How are we going to introduce these extensions into the general shell programme? It should be pointed out that the nett restraint at the shell ends is made up of the combined effect of the end support plate stiffness plus that of the brake path overhang. A suitable mathe: matical expression could possibly be found for this combined stiffness but it would be virtually impossible to introduce it in such a way as to be applied in the direct computation of shell defections of Research Pro- gramme No. 14, The way out of this dilemma is to consider separately all elements extending outside the normal shell boundaries thereby leaving the shell boundaries free of complexity. This enables simple boundary expressions to be used in the normal com- utation process. It has already been demonstrated that it is possible, by means of moment/slope relationships, to integrate the shell boundary and the end support members such that some equilibrium condition is reached (see 4.4). The same method can be used in the case of shells computed by means of Research Programme No. 14 because, if the shell boundaries are attached 10 exible members, which themselves exhibit proportional moment/slope relationships, then the boundary point of the shell must behave in a similar manner. 5.7 Moment / slope relationships of brake path section The boundary point at the sheli/brake path/end sup- ort junction can be “exploded” and providing each in-dividual element is provided with a similar value of bending moment slope and appropriate linear force to )that existing when it was a homogeneous part of the ‘whole. it is perfectly proper to treat it as an individual item. If each brake path extension is thus treated it can be assumed 10 take the form of a very short cylindrical shell as shown in Fig. 48. The only external forces on this member are the bend- ing moment and shear force, components at the right hand end. There is no known theory existing for providing a direct answer to the question of what is the slope for a given terminal moment and what is the deflection? , To obtain these answers a further computer programme was developed by Mr. B. M. Scott of the Computer Unit. Various trial runs were made and it was proved that the slope was directly proportional to the applied terminal moment. A further programme was developed to include a circumferential stiffening ring near the free end and a similar relationship was observed between the slope and the moment. This does, of course, make the problem delightfully simple, if one allows for the complexities along the way, because each brake path extension is behaving in the same fashion as the end support plates with respect to ‘moment and slope. If the end support plate is now “joined” to the brake path plate at the boundary then each will be subjected to the same slope but the total external moment necessary to deflect the combination, will be made up of the sum of the moment My necessary deflect the brake path alone and that necessary to Geflect the end support plate alone fe. Mr. Such an external moment will be provided by the shell, at its boundary, when it becomes “joined” to the above com- bination, The moment/slope relationship for both the brake path and the end support pate will be known in ad- vance as they will usually be dictated by design con- siderations. The sum of these two will be equal to the moment/slope relationships required at the shell bound- ary when utilising Research Programme No. 14 and it is simply a case of entering this relationship as input data so that the appropriate boundary correction can be made thus allowing the coil loads to be correctly computed. : ho Ix Tus amticle the authors will try to present to the reader the procedute adopted in obtaining the working stresses for a practical winder drum. As will be seea later the drum chosen is identical in almost every re spect to that which has featured so largely throughout the later parts of this work, This particular selection was made for the prime purpose of affording a compari son with the multifarious treatments given to this parti- cular geometry in previous articles and not merely as a matter of convenience. It is hoped the reader is not ining the impression that the methods used through- ‘out this series, and those shortly to be described, are only capable of designing one particular drum because this is not so, even though no other design is being investigated in this work. Far better to “ring the changes” on one particular shell than to confuse the issue by presenting a lot of disconnected results. Before commencing with the design analysis it will be necessary to amplify upon the short description of brake paths which was presented in the last article 66). 6.1. The constant “C” There is nothing magical about “C”. It is simply the ratio of boundary moment divided by the slope pro- duced by the moment i. Mol Slope which, for any specific shell geometry, or brake path ‘geometry or end support plate geometry. is a constant. Let us refer back to section (4.5) in the April issue and examine Figs. 31 and 32. It was found that, if the free end of a very long shell was given a terminal moment Mo then the corresponding shear force necessary to maintain this free end at zero deflection was —2Mo ie. Qo=—B Mo (equation 28). and that the slope produced at this free end by sub- stituting for Qo in the basic equation was:— ae _—Mo Eo app easton 30).From the definition of “C” above we obtain for the shell boundary:— Mo t=. -28D. Cerge7F 29D 2 & But Da or we 12 (=v) 4 ptr Substituting the first expression giv. 2pEe 12s cre. but pa . 2d vit 1285 £1 Crm VS EY 6 vra-) and cancelling the constants (assuming Ee vit 66) This is tne “C" value for the shell boundary and is constant for this shell as the slope produced is direcuy proportional to te applied moment, The reader must not be confused by this however be- cause, in the above context of proportionality, it is assumed that the shell boundary 1s connected 10 some elastic medium such that, as the shell load increases the boundary slope increases and the resteaining momeat in the elastic medium builds up in proportion to the slope. ‘The above value of “Cs” in equation (36) presumes that the shell extends beyond the boundary to infinity thus the boundary restraining moment/siope relationship is equivalent 10 the conditions obtaining in Figs. 31 and 32 as already described. This particular value of Cs will be called its “natural” value, 1 cam be shown that under these conditions of simu- lated end restraint the boundary moment is approxi- ‘mately half that produced at the boundary of a rigidly fixed ended shell ‘The value of Cs found from equation (36) is not the only admissible value as it can be changed within Re- search Programme No. 14 to assume a different, or ap- arent value such that the boundary conditions can exhibit behaviours ranging from rigidly fixed to simply supported by applying suitable corrections. This will be dealt with a litle more fully later on, Allied to the Cs value we can obtain similar values of Cb for the brake path boundary and Ce for the end plate boundary by applying suitable parameters to the appropriate computer programmes. ‘The manner in which Cb and Ce are found is by extracting the appro- priate bending moment and slope from the results ob- tained from the computer. go% We 13) gives: Crm 0.235. 6.2 The correlation of Cs,"Cb and Co In the ultimate, in order to simulate a practical shell, its boundary is rigidly connected to both the brake path ba. and the end plate; therefore, for equilibrium, all three ‘components are subjected to the same slope. Under these conditions we find that the moment demanded by the shell boundary must equal the sum of the moments produced at the boundaries of the other two components ue to their being strained to the same slope. In other words: — Cs=Cb+Ce+C for any additional component for the moment we are only concerned with two external components thus. using the modulus of the “C” values, weet Cs=Cb+Ce - G7) This simple expression is only made possible because of the direct proportionality existing between the applied moment and the slope produced on any of the three ‘components. ‘The apparent value of Cs can range between zero, for a simply supported shell boundary, through to infinity for @ rigidly fixed boundary, but the useful range will extend from zero up to some value around the natural Cs found from equation (36). 6.3 _A practical design ‘Where does the design begin? Quite obviously we shall know the basic geometry ie, dram diameter. width between flanges, rope diameter and type, shaft depth and loads, etc., etc., but the things that must be Known before commencing the analysis are the shell thickness, the number and area of compression rings and the type of end support/orake path combination. It is fully and rightly expected that many readers will a: “what on earth is the use, to the designer, of all this theory if the poor chap has got to design it first in any case?” This is absolutely no different from designing 2 bridge in which a rough idea must be obtained before its own self weight can be determined in order that a More accurate stress analysis can be made. The designer of a bridge cannot proceed om the basis of weichtlese members otherwise he is in for a shock when the nal ‘weight is taken into account. ‘Any drum desimm must be started with this in mind and, providing this is accepted, the design can proceed. It will be quite obvious that the authors have fully investigated the particular layout now to be described and, by virtue of this, know what the answers are goine to be. Generally speaking a little guess work is needed in order to arrive at a oreliminary design ‘that will be suitable for investivation and from the authors’ noint of view this has already been done. The reader's time will not. therefore. be wasted by too much preliminary “juoaline” and so it will be possible to go right to the final answers. The drum to be desizned will be one of a vair suitable for a double drum machine. Each drum will be fited with nwo narrow brake paths, instead of the usual sinele wide one, the ‘diameter being the same as that of the drum, i.e.. 10 ft. The width between the drum flanges willbe 5 fi in and this wil also be the distance between the supporting end plates. Shaft loads are such that the effective pull of the rope, measured at the drum, pro- Vides an equivalent unit surface pressure of 388 Tb/in? varying lineally across the drum to 360 Ib/in* at the ‘opposite end of the first layer. The equivalent pressures over the second and third lavers are 360-332 Tbfin® and 332-304 Ibin* respectively. The winding rope is a Ijin oedia. triangular strand construction assumed as having a stretch modulus of 15% 10° Ibjin?, +9: |The choice of stretch modulus for the rope is rather important as the value changes after the rope has been in use for some time and will gradually get larger to a degree dependent upon the method of manufacture and the type of rope used. A larger modulus will mean a higher KriKs ratio with resultant lower factors. In view of this itis imperstive that the manufacturer's “as new” modulus value should be chosen so as to enable the Worst conditions of loading to be applied to the shell in the design stages, a et 6.4 Preliminary analysis—the shel) ” Determine the mean load of each layer“ Layer No.1 Mean load 374 Tofin? 2.5% Layer No.2 Mean Joad = 346 Ibyin’ v.34, Layer No. 3° Mean load = 318 Ibfin? “vya) These loads will now have to be “factorised” in order ermine the factors until a shell thickness is known. is is where a litle judicious guess work is required. ‘A starting point can be made by assuming a limiting compressive stress over the central area of the shell , Let a factor of safety of 2 to 1 be chosen on the yield fe siress of 38,000 Ib/in® thus limiting the compressive stress to something in the order of 20,000 Ibjint ‘determine the nett Jond on the shel but we canzot thi (3° “Assuming now a plain shell with a thickness, based upon experience, of 2}in we are able to assess the Kr/Ks ratio as follows:— )) KrsEr Ar=15 x 10'x0-9075=13-6 x10" where Ar is the metallic cross sectional area of the rope. Ks=Es As=30X10'X25x 1-375 x 1-035=106-8 x 10" the rope pitch across the shell is 1-035 times the rope dia, and:— from the graph of Fig. 28 (in the March issue) the ap- priate layer factors will be: Fn,=094, Fn,=084 and Fn,=076 The net load on the shell will be:— From Ist layer = 374x094=352 From 2nd layer = 346x084=291 From 3rd layer = 318x0-76=242 Total 885 Ibjin* Pr _885x60 ‘Compressive stress in shellm 2m 21,250 Ion? resive srs in shell ma $0 Ibi which is a litle higher than required but in the right order. It was decided that a 24in thick plain shell was too thick and we would settle for a 2 inch shell reinforced ith compression rings. This decision could be. based fpon any number of reasons, e.g, material availability, rolling capacity, etc.. and could apply to any particular project. For the sake of the exercise this decision was ‘ade such that the procedure adopted for the inclusion cof compression rings could be investigated. = 42 How many compression rings are required? In the last article it was stated that the maximum spacing dis. tance should be no greater than 1°835 V/rrin which for a in thick shell is 20-i. With 2 rings the spacing would be 23in if evenly spaced along the shell, and in view of this three rings have been chosen, For simplicity of computation Research Programme No. 14 demands that each compression ring be situated exactly beneath a coil point, therefore the three rings have been placed be- neath coil numbers 12, 24 and 36 which gives a 12, coil 4plich spacing, iz, 17-lia, 8 t0 shells containing compression rings. Imagine an infinite shell containing compression ings, each having the same cross sectional area and all spaced at Zin apart. If the shell portion was fin thick and each ring cross sectional area was Ain? then ap- proximately the same fadial deflection would be ob- tained by using the same load if a plain shell of equiva Tent thickness 1, was used where:— tent Sin Ne o (38) ‘All that has happened here is that the sectional area of metal in the ring has been distributed over a length of shell equal to the ring spacing centre Z and the result- ing thickness added to that of the existing shell portion £. ‘This approach to the ring/shell combination is rather different from that outlined in the previous article wherein we were basically concerned with the effect of constant foral cross sectional area by suitably resrrang- ing the number of, and hence the amount of metal in. the compression rings. Returning to the design, the equivalent thickness has already been decided upon at 2tin and the basic thick- ness 1 is now to be 2in rearranging equation (38) to give A results in: 39) Z (t.—1) 99 in ring area 1x05 in? In view of the fact that the preliminary compressive stress was a little higher than our limit demanded it would be wise to slichtly increase the sectional area of each comoression ring. 10in* suegests itself and this is what they could be (In actual fact the ring areas chosen for use in Programme No. 14 were 9-91in’. This ‘was due entirely to convenience in comparing the effect of 3 layers avainst 1 laver for the same shell geometry as was used in the Dreoaration of the graph in Fig. 42 of the previous issue (curve c)). ‘The true equivalent thickness, using the 9-91in* rine is 991 tend 2s IM im 6.6 Values of the constants ” 1 will be found convenient to let the value of Cs for the shell take on its “natural” value of oss 2Fig. 46, This graph shows the variation in Ce for variations of Internal radius of + flenble end suppor plate having constant outside diameter and thickness 0-235 «30% 10% 2? x2 = 5-166 x 10° therefore the sum of Cb and Ce for the brake path and end support plate respectively must equal 5-166. 10+ ‘The brake paths chosen, from considerations of mecha nical braking requirements, are each 10in wide by 2in thick and each will be provided with a small circum- ferential stiffening ring of Sin? at its outer extremity. This not only effectively stiffens up the free end against heat distortion during braking but also acts as a small heat “sink” which evens up the temperature distribution when the paths get hot. ‘The above data was used in the running of the ap- propriate “Brake Path” computer programme and from the results we ascertained that Cb=32 which leaves $166—382=1-346 for the end support plate constant Ce. 6.7 End support plates There would now appear to be some difficulty in finding a flat annular plate to fit adequately into the structure and having the right proportion of annular depth and thickness to suit our requirements. This is not quite so difficult as may be imagined. It was decided to use 1in thick plate for this member and so all that is required is to run the “Flat Circular Plate” pro- gramme for different values of internal radius and plot 3 graph from the results. Such graph has been plotted in Fig. 46 for constant outside diameter and thickness, the variable inside radius as abscissae against the Ce value as ordinate. The intercept of 1-346 occurs at an inside radius of 44in (approx). This fs quite a convenient figure but if some inconvenient result had occurred then the plate thickness would have to be changed and a new graph drawn. Tt should be men- tioned that higher values of Ce are obtained by increas- ing the plate thickness for a constant outside diameter and annular depth or by decreasing the annular depth for a constant outside diameter and thickness. - 43 We are now in possession of all the data necessary for feeding into the computer so that Research Pro- ‘gramme No. 14 can produce its results. 6.8 Research programme number 14 It is not possible in the space available to: present an ‘analysis of this programme as it is too highly involved. A full study would probably run into about 15,000 to 20,000 words and the authors’ own report on this pro- gramme does not fall far short of this figure. It is clearly impossible to embark on any description of this nature. ‘Most of the theory contained within the programme has already been dealt with throughout the series but the techniques used to adjust coil loads layer by layer and the methods adopted to correct the ‘boundary infiuerices, shear forces, bending moments and slopes, etc., as each new coil and each new layer is applied must unfortunately remain hidden. For those interested in computer analysis we give below the computer input data in its correct order of presentation. Only the numbers are entered on the input data ta Shell radius in i 60 Shell thickness in in 2 Number of layers of rope . 3 Rope stretch modulus 15,000,000 res HP Number of compression tings 3 Possible No. of coils in Ist layer a Rope diameter in in 1375 Radius to Ist layer rope col centre Tine in in. . . 60-6875 Weight per ft of rope 1542 Initial rope pull in 1b 33130” ‘Area of each compression ring in 991 Rope coiling pitch coefficient . 1035, Coil number over Ist compression ring 2 Coil number over 2nd compression ring 4 Coil umber over 3rd compression ri 36 Coil number over 4th compression ring o Rope metallic area coefficient 048, No. of coils in final layer 48 LH. boundary Cs apparent value ... 5.166.000 RH. boundary Cs apparent value... - 5,166,000 ‘One or two points of explanation will be in order. ‘The possible number of coils in the first layer automatic- ally sets up the width between flanges to take 48 coils at the appropriate pitching irrespective of whether the first layer becomes filled or not. To some readers the weight per ft of 1:542 Ib for @ Ijin dia, triangular strand rope may seem low. The explanation is quite simple as the machine in question is winding up an incline, thus the weight given is the apparent weight obtained by multiplying the true weight by the sine of the incline angle. In any catslogue produced by the wire rope manu- facturers the reader will find tabulated data by which the cross sectional area of the wire metal can be found for various rope constructions. The usual equation is:— AZK where K is a constant depending upon the type of rope and its construction and ““d" is the overall diameter of the complete rope. The constant K above is the “Rope metalic area coeficient” in the programme input dat, ve, O48. ~Finally. the apparent value of Cs for both boundaries | has been given its “natural” value as previously men- tioned. 6.9 Programme results Figs. 47 and 48 show the complete system deflection ‘and ‘bending moment diagram respectively. It should bbe noted from Fig. 47 that the absolute maximum de- fection is 44-358 thou. under coil number 17 and the deflection at the centre ring is 41-285 thou. thus the mean deflection will be approximately 42'821 thou. The average resultant force over this same area is approxi- mately 894 Ibjin’. Using the fundamental equation: 3 894% 3,600 ” & gives a defection of Or _matg thou. By BN6S # deflection 30 106 2-58 eu ) hhas always given answers less than the true defiec- tion as will be remembered from the defection diagrams in section 3.1 and in view of this the above is a reasonably accurate comparison of the actual conditions. In the preliminary analysis (6.4) a maximum com- Dressive stress of around 20,000 Ibjin* was desired and the first attempt produced a stress of 21.250 Ib/in? which led to a slight increase in ring area being introduced. ‘The final compressive stress is a maximum of 22.179 Ib/ in? under coil 17 but over the central ring it is only 20.642 Ib/in*. When it is considered that the preliminary Compressive stress is based on using — an equation t which, on normal shells, always gives slightly low values over the central areas then the actual final stresses are notfarou. 6.10 Boundary junction stress analysis This is the most important aspect of the whole design as “bad” stresses at this point would mean a complete revision of the problem. Let this junction now be in- vestigated. In Fig. 49 the “exploded” section is shown, the ends of each individual member being provided with the appropriate forces such as to maintain the joint in equilibrium at the given slope of 0-00303. The appropriate values of the "C” constants, the bending moments and the resulting bending stresses f, in the shell and the brake path are self explanatory but the stresses at the boundary of the end support need some explanation f bi tanned oe ttnnt od Btlee ae Anton, ede ” a nny mene Fig. 47 (top). Detlection diagram of complete shall Brake paths ‘and end plates assembly - Fig. 48 (bottom). Bending moment disgram for complete drum ‘assembly of Fig. 47Fig. 48. Simple analysis of direct forces in sheli/bvakepath|end piste junction for shell of Fig. 47 and 48 Research Programme No. 14 provides us with a sup: ort reaction which, for the R.H. boundary, is —7,465 Ib/in of circumference, This is a “natural” value ‘ob- tained by considering the shell as extending beyond this point to infinity. The support reaction is made up of three elements:— (a) Shear force due to the load only on the shell, as if the ends were simply supported. (b) Shear fore: due to applying the fixing moment at end of shell proper. (©) Shear force due to straining the end of the in- finitive extension to the same slop: The shear forces in (b) and (c) are of identical mag- nitude and are each equal to @ Mo (see equation (28)), For this shell the shear component is 0-1173x 15.610: 1.830 Ib/in of circumference. If this infinitive shell ex- tension be removed and reolaced by a pure moment which needed no shear reaction then the support reac- tion would be 7,465—1.830=5,635 1b. A brake path extension is now to be fitted but its boundary point de- mands a shear reaction when a moment is applied. It was found from the appropriate computer programme that for a brake path boundary moment of 11.540 Tb infin of circumference the shear at this point was 1,685 Ibfin of circumference. The true end plate support re- action is now 5,635+1,685=7,320 Ibjin of circumference which is a compressive force of 7,320/1-25=5,860 Ibjin® ‘over the periphery of the end plate. Combined with this compressive force are the stresses due to bending the end plate which, for a bending moment of 4,070 Ib injin of circumference, are + 15,600 Ib/in®. For the fully deflected state the combined stresses are 21.460 Ib/in? compressive and 9,740 Ibjin® tensile. We now have what appears to be a very satisfactory situation with all the maximum stresses around about the same intensity. i.e., 22.179 Ib/in? max. compressive in the shell. + 23,400 Tb/in* shell bending stress and 21.460 Ib/in® max. compressive over the periphery of the end support plate. Although all these stresses may look 10 be quite in order there is another very important aspect 10 consider, 6.11 Fatigue It is very surprising how litte attention is paid 10 this most important branch of mechanical engineering. = 45 coesaneatet tee2enen’ eet ifn MaetS40 in a6-2604ie eA fez 000) When a designer, in any of the general engineering fields is faced with a problem involving repetitive stress ap- plications how much thought does he give to such things as S°N curves, life rating, stress limits and, most important of all, how much emphasis is laid upon the design of the weld which holds his components together? ‘He may have analysed pretty thoroughly the mechanical strength of the composite whole and is therefore ex- tremely perplexed and frantically searches for his calcu- Jations when, in six months time, his pride and joy Ties, in an untidy heap upon the floor. Upon examination he may discover that certain components have completely broken at points where they were welded to other mem- bers, and to exonerate himself, hurriedly blaims Fred the Welder who was “never any good anyway”. In all pro- ability Fred wae not to blame at all as he was merely carrying out the designer’s instructions. What, then, had one wrong? Tt is only in recent years that the more forward look- ing Companies have given their desianers a chance to answer this question and it has all been made possible by the excelient services given to Industry by the British ‘Welding Research Association. This Association carries ut extensive research into all asvects of fatigue and is also fully eavinved to undertake site testing of com- Ponents and has already been actively enraged in de~ termining the causes of several winder drum failures. ‘Their advice has also been sought in the preparation of this articular article. What is fatieue? So far the research workers in this field are not whollv satisfied amone themselves that the complete answer has been found to explain the mecha~ nics of fatigue. What is more important than this. how- ever. is that an understanding be gained as to the fun: mental causes of fatigue: the effect is of secondary im- Portance to the engineer. ‘There are two things which can determine the fatigue life of a welded structure and these are:-— (a) Magnitude and type of stress fluctuation (>) Type of weld and joint ‘The fatigue Tife is measured in terms of the total number of stress fluctuations to which any parti= cular joint can be subjected before failure occurs. Fromthis the wetual life, on a time basis, can be obtained ‘when the time of one complete fluctuation cycle is Anown, 6.12 Weld types and allowable stresses The biggest factors affecting fatigue life are the type ‘of joint used to fasten components together and the type of weld used in the production of such a joint. B.S.153. dealing with Girder Bridges, allocates certain class letters 10 various specific types of welded joint connections for which it is possible to obtain limiting maximum siresses for a given fatigue life consisting of a given number of cyclic stress fluctuations. Such stresses can range from zero to wholly compressive, or wholly tensile or there can be complete stress re versals. ‘This stress range is usually put in the form of: — Fie ) Fr where Fg minimum stress Fon = maximum stress sith "2" becoming —ve or +ve depending upon the sign of the appropriate stresses. Conventionally, tensile siresses are +ve and compressive stresses are —ve, and in the case of some winder drums Frm/Fan™=Zer0, irrespective of whether we are concerned with tensile f Compressive stresses (i.e. Fay is zero for an unloaded rum), It should be pointed out at this stage that the authors consider the thermal stress relief of fabricated drums an essential part of the manufacturing process as it, will remove a5 much residual welding stress as possible. This leaves the drum in a virtually unstressed state, prior to loading with rope, and ensures that the stresses, {35 found from the previous calculations, are more tealis- tic of the true state as they are not influenced by any unknown residual stress. Machining stability is also im- “ved. of course. a Tiobnien! Compe Soe ag |—Fienie oe ‘oper te ) Fig $0. Pam section through end of shell. The double arrowed lines indiente the directions of the tantle stresses at the Dorn where fetigue failures are most ikely 10 occur Compressive stresses. for an Fin of 2ere, are not being considered in view of the fact thatthe whole stucture i being thermally stress refeved - 46 Winder drums generally (except single drum machines) are subjected to working stresses ranging from zero, in the unloaded state, up to some maximum Value of tensile or compressive stress which exists after all the tensioned rope has been wound onto the drum. In certain cases of drums containing a high proportion of “dead” coils which have been loaded by the “doub- Jing down” technique the above is not true and special attention should be paid to the stress range ratio Frg/Fnax in which Fy, is some stress other than zero). In general, and in view of the fact that stress relief is being carried out, it is possible to ignore the effect cof stresses ranging from zero to purely compressive as itis known that fatigue failure, due to this type of stress in a stress relieved joint, is highly unlikely. ‘The problem now resolves itself simply into assessing the effect of wholly tensile stresses upon the fatigue life of the Welded drum. i.e. Fai/Fye=zero With Fox, being the maximum tensile stress at any particular joint being ted. ‘The limiting stresses given in B.S.153 for each class of welded joint have been based upon life tests carried cout by subjecting each specimen to pure tensile fluctuat- ing forces up to fatieue failure point. In the case of the above drum we are faced with a combination of bending and direct stresses. It will be found sufficient to consider the maximum tensile fibre stresses and calculate the fatirue life on the basis of an equivalent direct force which produces a direct stress equal to the above fibre. stresses. How does all this fit in. with the drum desien being investigated? It is first necessary to determine the clas of joint which is applicable to our case before an analysis can be undertaken, Fig. 50 shows a cross section at the end of the shell and it will be noticed that the joint contains full penetration welds between shell and suo- Fig. 51 (left). Joint in the “as welded” condition. The fully prepared Welds run trensverse to the dlrection of the stress artows, This simulates the conditions for fatigue life calculations related to the maz. tensite fibre tres at point B (Fig. 50) before machining the radius. t i given a Class “F" classiication when the lines ‘of strae are in the drection shown Fig, 52 (middle). The sme joint asin Fi. 51 but sil walds have Deen machined to radivs. Thi simulates the conditions Yor gue lite calculations relstad to the mex. tensile Tove sesses 41 points A and B of Fig. 50. Athough having the seme basic ‘geometry ax the joint of Fig. 51 and wih the stresses in the same Grection Mt has’ the equivalent Yaligue strength of » Class “D° (detail ue to machining the welds Fig, 53 (right). This configuration simulates the condkions for {atigue fife calculations related to the max. tensile fore sess at oint D of Fig. 50. Anhough this weld detail identical to thet of Fig. 1 its given a Class “E™ classification due to the siteration ‘in the dhection of the stressPort plate and between shell and drum flange. In terms of B.S.153 this is a cruciform butt weld which is allocated a Class Letter F in the “as welded” condition (see Fig. 51), The surface at point A is, however. machined and, depending upon the roughness of the machining, will clevate this point 10 a somewhat better class than F. If the machining is reasonably good @ Ciass D joint can be assumed (see Fig. 52). It is now found convenient to use one of two Fatigue Life Diagrams published by the BW.R.A. and based upon B.S.153. One diagram is used when the stresses are predominantly compressive, and the other when they ‘are predominantly tensile and it is the latter which will be used in this instance. The tensile stress at point A, Fig. $0, based upon the sitess obtained at the theoretical boundary point, ranges from zero up to a value of +1045 ton/int and, using the appropriate fatigue diagram, we obtain Point Ref. Weld Class Stress range Cycles to failure A D Oto +1045 4x10" For this machine each stress cycle occurs 66 times per hour. therefore the fatigue life. on a time basis, will be 400,000 56x 24% 365 = 69 years! Let us now examine point B, which, in the “as welded” condition, has a Class F rating (see Fig. 51) Point Ref. Weld Class Stress range Cycles 10 failure B F O10 +77 35x10 (unmachined) which gives a life of six years! Itis a simple matter to machine a radius at point B by extending the machined surface of the brake path along towards the drum flange. If, after machining, it is allocated a Class Letter D then the life will be 69 years. Tt should now be very clear to the reader that the choice of class letter for any particular point around the joint can make a vast difference in the fatigue life ex- Pectancy. This is amply illustrated by the life difference at point B due to elevating its class letter from F to D. By suitably machining this point the life has increased to 1] times its original value. What of point A? For this particular construction the only thing which can be done is to improve the quality of the machined surface at this point. This would involve extra fine machining or grinding which, on a structure of this size, although possible, becomes economically undesirable. Nobody, not even the experts, will say that the assess- ment of fatigue life is an exact science. It cannot pos- sibly be so as no two similar specimens can be manu- factured such that the welding in each is absolutely identical. The B.W.R.A. diagrams give a fatigue life below which failure is not expected and as such is the safe minimum life for any specific joint detail. On the other hand the actual life for a specific joint could be well above the predicted value. ‘The reason for this is that the fatigue life, as found under actual test conditions. for a multitude of sup- Posedly identical specimens produces a “‘seatter band” fon a graphical plot of maximum tensile stress against coytles to failure, for an Fau/Faw ratio of zero, Very broadly speaking the B.W.R.A. diagrams (and B.S.153) are based upon a fatigue life created by select- ing limiting stresses lying near the lower limit of this scatter band thus ensuring that there is a reasonable chance of a greater life than that calculated, but a very much smaller chance of failure before the predicted life. With respect to the drum design being investigated there is a minimum life expectancy of some seven years with a possible extension of this up to some unpredict able value depending on the quality of the machining. This is not at all satisfactory and it would be far better to design for a minimum life expectancy of say 25 years. What alternative is now open to the designer other than costly additional care during machining? One alternative is to have simply supported shell boundaries: but what if the design costs do not allow for this in view of the necessity of separate bolt on brake paths, etc.? Tt becomes ridiculously obvious, when considering only tensile fatigue, that the ideal state of affairs would exist if the drum flange could be repositioned such that its influence was outside that area of the shell sub- jected to large tensile forces, ic.. position it in an area where the shell bending stresses over the outer surface Fig. 48 will show that the negative bending dies away very rapidly and becomes zero at only 316in from the boundary. If this sbell retains its basic geometry but the boundary points, or end support plates, were to be moved outwards towards the ends of the’ brake path then the above effect would be obtained. Having re~ arranged these components in this fashion it will be found that the fatigue life will now be dictated by the relatively low tensile stress existing in the support plates and also by that existing in the under surface of the shell adjacent to the compression rings. So far no mention has been made as to the life ex- pectancy of the joint based upon the stresses at point D in the end support plate. Ignoring the purely compres- sive stress we are left with a maximum tensile stress ranging from zero up to +4°35 tonfin’. At this particu- lar point we can allocate a Class Letter E (see Fig. 53) and this will give a minimum fatigue life of 260 years which is clearly ample! This final point has been left until this late stage in order to show that, providing the maximum static com- pressive stress in the outer periphery of the support plate has an ample factor of safety on the yield stress than the repositioning of the drum flange. as already outlined, virtually eliminates the possibility of fatigue failures associated with the shell and its attachments within a very long useful life, It is:hoped to show that this is so when the data appropriate to a redesigned end has been presented>}1_ Increasing the fatigue life "in the previous article it was shown that in certain types of end construction, which may result out of pure Revessity or out of peor design, the minimum fatigue life was very low and was produced, not by a load bearing member, but by a flange whose sole purpose is Yo retain the rope cails within defined limits. The flange {o shell connecting welds were positioned at. point on the shell subjected to high tensile bending stresses and it should be a sirict rule, when designing any welded structure. which is subjected to regular stress fluctua- tions, to avoid welding in highly tensile areas. This is not ulways possible, in which case great care should be taken 10 limit such tensile stresses to safe limits dictated by the fatigue life required of the structure. 11 is not possible 10 outline in a few words the “do's” and “don'ts” related to fatigue as so much depends upon the type and magnitude of stress and the way in which it fluctuates coupled, of course, with the type of welded upon which the stresses are acting, this much sould be apparent from what has already been written, ‘A further computer aided study was made on the shell shown in Fig. 47 (previous article) using “Research Pro- ramme No. 14" with a view to rearranging the end sup- Ports $0 as 10 increase the minimum fatigue life from seven years up to some more realistic figure. In order to assess the maximum tensile design stress for uny specified machine it is first necessary to know its eyelic duty. As a typical example let us take a winder _Pefating a shallow depth which has a tua eel time 2 min (ie. from bank to bottom and to bank again), Ii is assumed that, at bank, the tensile stresses some- Where in the drum shell are’ at a maximum whereas at Pit bottom they are sensibly zerot thus, in this particular tase, the stress fluctuation cycle time is 2 min, ie, 30 times per hour. Having selecied a suitable “life” for the machine the stress cycles 10 failure can be found from:— No. of cycles to failure =52 LH.D.W. where:— selected life in years stress cycles per hour hours usage per day. Wesdays usage per week, For the above machine running 16 hours per day, 5 days a wesk for 25 years, the cycles to failure=3-12 x 10* and for class EB and F weld details the maximum tensile stresses must be Kept below 58 and 43 ton/in® respectively (13,000 and 9,650 Ibfin’). If the above machine was run for 24 hours per day, 7 days a week the failure cycles would be 6552 X 10" and the stresses for the same weld classes would then have to be kept below $-0 and 38 ton/in? respectively (11.200 and 8.500 bin’), ye 4B Perhaps many readers would consider 25 years too low a life and consider say 50 years more appropriate, in which case the limiting stresses would be 10.000 and 6.720 Ib/in*. If the maximum tensile shell stresses can be Kept below 5.000 Ib‘in? then fatigue failure is not likely, although the accurate assessment of the limiting stresses in the upper life range for these poorer-classes of weld detail is extremely difficult when working to the BW.R.A. fatigue graphs, Figs. 54 and 55 show the complete system defiection and bending moment diagrams respectively. This shell hhas now been rearranged so as to contain only one wide brake path at the L-H. end, whereas at the RH. end the shell boundary is connected only to the support plate. The L.H. support plate has been moved outwards a istance of 4 coil pitches, ie. approx. Sdin and is of similar construction to that previously used. The RH. support plate, on the other hand. is completely unstif- fened by radial arms, all the longitudinal stability coming from the LH. end, Several things become immediately apparent and these, related to the shell of Fig. 47, are:— (a) LH. boundary moment has’ been reduced from = 15.500 to 8,860 1b in/in of circumference. (b) R.H. boundary moment shows a drastic reduction down to - 2,237 Ib in/in of circumference. (©) The moment under the L.H. drum flange is now only 2,368 Ib infin ‘of circumference. A tre- mendous reduction, although this design has stil not placed this flange in a compressive area. (a) Both support reactions have been reduced. (©) Central area compressive stresses remain virtually unaltered, Considering the minimal alterations, this has achieved a remarkable reduction in the stresses at the “critical” points in the drum which has obviously improved the fatigue life beyond all expectations. 7.2 Fatigue life aspect ‘The shortest fatigue life that can be measured is 3.460 years. as outlined in the diagram of Fig. 56, and relates to the tensile stress in the support_plate ranging from zero, for an unloaded shell, up to 7.335 Ib/in’. This is 4 highly theoretical figure and can be taken to mean that, for this stress range, fatigue failure is quite unlikely in the Class E weld. The same applies to the points of attachment at the drum flanges where the tensile stresses are so low as to preclude fatigue failure altogether. There will be no trouble at the compression rings providing the shell has been correctly stress relieved and so it can be con- fidently expected that this design will have a useful life far exceeding its opzrational requirements Figs. 56 and 57 show the exploded view of each end giving all the relevant details appropriate to these points.| oer H 04 hin a i a Ee al mot Fig. 54. Deflection of drum components for revised boundary and end conditions. Nate that this drum contains only one brekepath ending mement under flange — 2.268 Ib inj or circ. producing an upper shell surface tensile stress of 3.550 lbjin® This sess it 100 low to produce fatigue in any measurable time In these wo diagrams it can be seen that the stresses are all very conservative and in fact the largest stresses anywhere occur as a shell bending stress of + 17.550 Ib/in* under coil number 44 and as a shell compressive stress of 21,826 lbjin* under coil number 17. In view of the above, and if one cares to allow the two shell stresses to go’ up slightly, there is no reason why a thinner shell could not be used, providing of course that the remaining stresses are within the li demanded by 2 reasonable fatigue life related to the critical points in the design. Quite certainly a thinner shell, backed up by slightly larger compression rings, would not produce prob shell stresses across the rings. (The max. bending stress ‘over uny ting in Fig. $4 is only 7,650 ibjin’.) 7.3 Modifications to boundary stresses Stress analysis throughout this work has been based ‘upon the assumption that there is no radial shell defiec- tion at the boundaries. It has never been suggested ‘that it is entirely absent but it has been stated that it is Virtually absent (see 1.4). Dr. Crawford was perfectly correct in assuming that deflection did take place but he allowed for 100 much. ‘As an exercise let us investigate what sort of radial UeRection it is possible to attain by analysing the end plate in the design just discussed. Let the outside radius he Sin and the inside radius be say I4in where it is rigidly attached to the bosses. (The inside radius of 44in, in Fig. 54 is not the physical limit of the end plate but the assumed limit of flexibility.) If we think about what happens we shall come to realise that the behaviour of this end plate is like a” section of infinitely long thick cylinder subjected 10 ex- ternal and internal fluid pressures. The external “pres- sure" is provided by the shear reactions from the shell and the internal “pressure” is the radial reaction from the boss. In this analysis it is assumed that the bosses ‘are massive enough to preclude any radial deflection and in this case there will be no circumferential stress exi ing at the side plate to boss interface, By using Lame's equation we shall be able to find an equivalent internal Pressure” such that, in combination with the external ‘pressure”, produces zero circumferential stress at the boss interface. It will then be possible, having found this internal pressure, to obtain the value of the circum- ferential stress at the outside radius such as to deter- mine the amount of circumferential strain at this point and hence the reduction in diameter o radial deflection. ‘As an example take the case of an infinite shell louded only with a negative band of load representing a support reaction. If this load had initially been developed by an un- yielding support, and the support was then allowed 10 yield, a state of equilibrium would be reached whereby the load on the shell equals the load over the support Plate, and would be less than the original. The amount ugby which the final load differs from the original load is measure of the deflection produced to obtain equili- veium. Let original load on shell be F. and the final lead be F. Ip in of circumference. By suitably rearranging Lame's equation we obtain: eee (40) Where:-—r=shell radius T=support plate thickness E=Youngs modulus W=support defection Rar 4h FoR PR with R being the boss radius, We know from basic theory that a single band of load produces a shell deflection of: ) eke ) 2E and to find the load, representing the above shell de- ection change, the equation becomes: — 2EW Pe PaiFe i) Ibfin of circumference change bur which produces:— PBF. Fm ; epee Ks 41) In the shell being discussed (Fig. $4) the support re- action is 1,905 lb/in of circumference at the LH. boundary. This can be considered as F. Inserting the appropriate parameters in equation (41) ives. 3.600 x0-1173 x 1,906 Fe = (6.600 0-1173)+ (26008982) 1.355 bjin of cie Inserting this answer in eq, (40) we get:— 1,355 «60% 0898 x 1,000 a EEO O00 05 thoy, 30x 10x 125 The support reaction has been reduced by 551 Ibjin tooo | i “Ty ¥ Joe Melasma | An\ It love. I > : = i io “| a | se ; i | dee Se | aaa * i rr a ) ig, 55. Bending moment alagram for drum components of dium ange producing & tensile sites 0f aniy 2.580 Ibin’. The bending shown in Fig. 47. Note the low value of bending moment at drum 1g. Moment at right-hand flange is only —7,237. producing an even ower tensile sess. Neither of these points need now be considered at points ot which fatigue Velare will Becur 50of circumference which is very marginal. So then, in the ultimate, support defections ate not too severe and this feature has not been allowed for in the results produced by “Research Programme No. 15" as itis felt that the values it does give will be slightly on_the high side with respect to bending moments3. ‘The above analysis. when applied to a practical de- sign case. can only be approximate because support settlement depends to a very large extent on the ‘Ype of boundary with which we are dealing and has only been included as a rough guide as to the variations produced. 7.4 Stiffness of end support plates ‘The property of the “stiffened” support plate is that, although a limit of flexibility has been introduced, which is assumed as being completely flexible, not only does this flex, but the “stiffened” portion does also to much smaller extent. 11 can be shown that a spoked wheel, subjected to uniform peripheral bending moments around its rim, de- cis very little due to its stfines, relative 10 flat plate. especially when the radial arms. exhi fairly high section inertia. An example is, of course. 10 be found in the difference between the cross sectional inertia of a rectangular fat plate and a tee section both in bending, In the former an inertia of say 2-0 would be compared with probably 100 or more for the tee section although this is not an exact parallel. Nevertheless, the stiffened portion does defect and this ean only help to still further reduce the boundary moment slighty. ‘As the nature of the combined deforma.‘on of a par- tialy stifened annular flat plate involves complex rmathemates its effect upon the boundary moments has not been allowed for, thus the boundary moments found from “Research Programme No. 14” will be slightly higher than is actually the case which, once again, pro- ides a further safely factor in design. Tt should be pointed out, in this respect, that the boundary moments given by the research programme are only ever theoreti- cally exact when the support plate is unstifened and similar to that shown as an example at the right-hand end of the drum in Fig. 54. 7.5 Other stresses Having satisied ourselves that the shell has been correctly designed and is adequate from the point of view of fatigue failure, are there any other eritical points in the drum as a whole at which troubles can occur? In the design being investigated the only other critical Points are the support plate to boss connections, especi- ally the left-hand end where the radial stiffening arms are welded to the boss. At this point the trouble usually arises. not from stresses produced by coiling on the rope, but from the defection of the shaft under the ‘dead weight of the whole drum plus the complete com- plement of rope wound onto i, When the shaft deflects, the slope, adjacent to the drum bosses, causes the bosses to rotate thus inducing a bending moment in the radial stiffening arms propor tonal to the degree of rotation. As a very simple case Jet us examine @ single drum machine whose shaft is simply supporied between two drumshaft bearings as shown in Fig. 58, where, D= 18in (1 for shaft: a= in b=132in 1=168in W= 40 tons (89,600 Ib) The slope of the drumshaft at under the load W is given by: 5,160in') an Wa a 2ET sas (42) For the parameters given here the slope becormes:— 89,600 x 36 on. 132- 2x 30x 10%x 5.160 (32-360) 0:06 radians From this we may obtain the bending moment at the boss end of the radial stiffener which 3El @ bjin i wee (43) he When 1, and 1, are the cross-sectional moment. of inertia and length of the radial stiffener respectively. For the stiflener/end plate combination shown in Fig. 59 the section inertia about the neutral axis is 190in* and its effective length is assumed as being 30in 330% 10% 190 x 0-001 = 30 My=570,000 tojin, ‘The min. section modulus of stifener is Z=28Sin’ i 5 570.000 _ * (8: and the bending stress is “190°? = 20.000 tbjint (89 tonlin’), ‘When considering fatigue in this respect it should be remembered that the stress just found in the arm, adja~ Cent to the boss, will be subjected to a complete reversal ‘once every revolution, and this particular weld joint can only be as good as a Class F detail which, for the above stress, allows an extremely limited number of revolutions. to failure. What can be done to ease this situation? Let the drumshaft diameter be increased to 20in and tuck the bearings in closer to the drum, ie. let a=30in, then we find that the shaft slope at the boss is 0-000548) radians. By reducing the moment of inertia of the sti fenet/end plate combination to 63:Sin‘ we find the bend- ing moment has fallen to 104,500 Ib/in and the bending. stfess to 9,200 Ib/in® (4-1 ton/in*) which gives this joint a life of 4,800,000 cycles which, on the basis of winding say 42°5 rev per wind in 1 min with a 16-hour day and a S-day week, produces a life of only about 24 weeks as a minimum, ‘Supposing the stiffeners were to be removed altogether, ‘The stress can now be easily found fr are, 7 (44) Where is a constant depending on the ratio of boss radius to inner shell radius Rr. If we use the same une stifened plate as in the shell of Fig. 54 then Rir=14/59 }237 and a=7 approx. Tx 1-25 x30 x 10% «0.000548 © (at boss) =: <8 (at boss)=: 3 = £2,440 Ibjin# (1-09°ton/in?) ‘This virtually eliminaies fatigue within the useful life of the machine, as a stress of only 1-2 ton/in? provides a cyclic life of 100,000,000 which on a time basis is nearly 10 years. Although this may appear rather shortgompared with the known life of many old machines it Should be remembered that this is a minimum life which could easily be 50 years, which is reinforced, of course. by the fact that it is very difficult to obtain exact, lives at this exiremely high end of the life range. If the stresses at these points are kept at 2.500 Ib/in? of below then we can eliminate fatigue failure altogether. From the above it is-easy to see how important it is to reduce the arm stiffness. ee 41525 to tt td Fig. 56. “Exploded” view of LH. boundary junction showing ‘relevant forces and stresses. Min. fatigue ie baced on sttest of 7,335 Iblin' tensile in end plate ie 2X10" cycles (class £ weld) ‘producing « ite time of 3.460 years for this machine! ST = ) 49200! a ot a er pe Fig. 57. “Exploded” view of RH. boundary junction showing ‘levant forces and stresses. Min. fatigue ite forth joint cannot ‘be measured as stress is t00 low Once again it is difficult to lay down any hard and fast rules related to design around this area except 10 say that undue stiffness should be avoided at all costs. So much depends on the overall configuration of the whole machine. The example given above was only & simple one of a drumshaft resting on two bearings, whereas normally the shaft is continuous ovet three or more bearings, which automatically controls the shaft slope under the drum bosses for a given shaft size. The effect of gear loads (when applicable) should be carefully considered, as the additional defections and slopes produced when acting in the upward direction an seriously affect the fatigue life even though, when acting in the downwards direction, the slopes can be less than under dead load conditions. In this respect it ould once agsin be stressed that fatigue ile is nor «hversely proportional to stress, so that the bonus gained with the gear load in the downward direction is more than destroyed when it is acting in the reverse manner. These remarks really relate to a single drum three bearing machine. Fatigue life analysis is made increasingly difficult, - 52 when considering the effects of gear loads, due to the variation of transmitted torque throughout the winding cycle. It should be sufficient, however, to maintain & stress limit, at the arms, of 2.500 Ib‘in? related to the slope produced by a gear load equivalent to the R.M.S. horsepower of the driving motor. nally, in respect of arm fatigue, it should be pointed fut that the simple equation (42) can only be used to determine the slope at the drum bosses for the drum configuration of Fig. 58. For continuous shafts, and other configurations, | more , sophisticated analytical methods must be used which, unfortunately, cannot be discussed herein. Equation (43) can be used, however. for the determination of the “arm” bending moment and it should be mentioned that, however many arms there may be around the end support plate, each one is sub- jected to the slope produced at the drum boss, as it asses bottom dead centre, once every revolution. In the frst part of this series (December 1966) in para- graph 1.2 we outlined nine types of load which could be found on any typical winder drum. Of these (d), (g), (%) and (j) have already been dealt with. Of the remainder (a), (6) and (f) can be discounted as not creating too much cause for concern except perhaps in certain cit- cumstances. The two remaining forces are:— (©) Torsional load applied via main driving motor and (6) Forces due to thermal expansion of brake path. Dealing with (b) first. The only point which needs consideration from this aspect is the strength of the Welds connecting the end support plates to the bosses, which must be strong enough to amply resist the driving torsion. Usually this is no problem and the minimum weld area on one boss can be assessed by assuming that ‘one boss only is taking the peak torque from the driving ‘motor. The final weld used in the design must, of course, bear some relation to the support plate thickness even though it may be far too strong to just take the driving torque. ‘These torsional stresses will not have any undue effect upon the stiffener weld fatigue life as the stifener arm itself can play no part in resisting torsional forces at the critical point where its outer edge joins the boss. Generally speaking the torsional shear forces measured in the metal of the support plate are so low as to be not worthy of consideration in the general context. Turning now to (e) above. It is unfortunate that, as yet, very little research has been carried out into thi extremely complex subject and the Authors cannot, at this stage, present any evidence as to the stresses in- volved. This subject is equally as complex as the analyses presented iit this series of articles. and many months of work are needed before any results could be forthcoming. It is known that, under certain conditions, very high stresses can be setup, and of sufficient magnitude to produce failures, which have occurred on some drums constructed yn end supports. With fabricated drums, of the type being discussed herein, it is not known how such stresses would affect the overall struc- ture, particularly the critical points where shell, brake- path and support plates meet. ‘Generally, however, on modern winding engines. due to the more sophisticated electrical control which i currently available, the brakes are rarely used as a re-