Improvements to laser forming through process control refinements

Optics & Laser Technology 30 (1998) 141±146 Improvements to laser forming through process control re®nements Gareth Thomson *, Mark Pridham Department of Applied Physics, Electronic and Mechanical Engineering, University of Dundee, Dundee DD1 4HN, UK Received 6 April 1998; accepted 14 May 1998 Abstract Laser forming is a process that uses the energy of relatively high powered lasers to cause permanent deformation to components by inducing localised thermal stresses. It is envisaged that this material processing technique will ®nd a number of commercial applications. This paper brie¯y discusses laser forming and the development of a basic process monitoring and control system used to overcome variability problems due to the complex nature of the lasers themselves and the manner in which they interact with material. It then goes on to show how the basic control system was modi®ed, using increased feedback data sampling, time delays and a modi®ed control algorithm which takes account of the forming rate in addition to the error. The e€ect of these developments is then illustrated by a series of tests which show the modi®cations signi®cantly improve process tolerances. # 1998 Elsevier Science Ltd. All rights reserved. Keywords: Laser processing; Process control; Laser forming; Prototyping techniques; System optimization 1. Introduction Laser forming is a process that has been developed over the last ®ve years or so as a novel material processing technique. Essentially a defocused laser beam, normally from a CO2 device with a power rating of 100±200 W, acts on the surface of a sheet of metallic material. This very localised heating causes the region under the beam to attempt to expand while at the same time this region is constrained by cooler surrounding material. This con¯ict creates compressive stresses within the material and depending on various process parameters, can lead to through thickness buckling or plastic compression of the upper surface. It is the latter process, more formally known as the `temperature gradient mechanism' that will be used in the tests carried out in this paper. A typical example of the components which can be produced by laser forming is shown in Fig. 1. More detail on existing work and a fuller explanation of the laser forming mechanism may be found elsewhere [1±7]. * Corresponding author. Tel.: +44-1382-344-908; Fax: +44-1382345-509; E-mail: [email protected]. 0030-3992/98/$19.00 # 1998 Elsevier Science Ltd. All rights reserved. PII: S 0 0 3 0 - 3 9 9 2 ( 9 8 ) 0 0 0 3 7 - 1 Laser material processing of any form is often a highly complex process. High powered lasers are devices which rely on optical, electrical, gas ¯ow and cooling systems to operate e€ectively, all of which can be subject to ¯uctuation. This may take the form of the degradation of optics through accumulation of debris or oscillation of the laser gas through minor system leaks and automatic replenishment systems. Such variation will alter the nature of the beam produced and so have an e€ect on processing [8±12]. Likewise, nominally similar materials will feature variations in surface ®nish, composition, thickness and residual stresses set up during production, which will show up as process variability [13]. While some of this process variability can be reduced through regular laser servicing and careful material selection and preparation, in practice it can be very dicult to produce repeatable results using purely historical or theoretical information. These diculties have led to the development of a wide range of dynamic process monitoring systems, particularly in the case of laser welding [14± 19], to ensure optimum conditions are maintained at all times. It has been seen that laser forming can be prone to process variability with nominally identical test pieces and processing conditions exhibiting di€er- 142 G. Thomson, M. Pridham / Optics & Laser Technology 30 (1998) 141±146 Fig. 1. Laser formed part made to approximately 1:8 scale of real door panel. ent responses to the forming process [20±22]. It is evident that the process will require a similar feedback approach if it is to achieve the degree of repeatability required for commercial exploitation. the top surface together as it cools resulting in the material always bending toward the beam. The process is also incremental, with the de¯ection increasing with each pass beneath the beam. If the forming increments are too large there is a risk target de¯ections may be subjected to excessive overshoot beyond these targets. 2. Basic control system A simple laser forming control system was developed to investigate how easily the process could be governed and a schematic representation of this system is shown in Fig. 2. Essentially a strip specimen of 1 mm thick mild steel, 10 mm broad and 80 mm long was clamped to a small CNC table. This allowed the specimen to be moved beneath a pulsed CO2 laser beam producing an average power of around 200 W and with the beam defocused to give a beam diameter at the material surface of 3 mm. The specimen was laser formed using repeated traverses beneath the laser along a track 30 mm in from the free end of the strip, while the tip de¯ection was monitored using a linearly variable displacement transformer (LVDT). The signal from the LVDT was fed into a PC running software which compared de¯ections to predetermined targets and responded with appropriate signals for the CNC controller. Laser forming using the temperature gradient mechanism is essentially a non-reversible process. The plastically compressed upper region of the specimen draws Fig. 2. Laser forming control system schematic. G. Thomson, M. Pridham / Optics & Laser Technology 30 (1998) 141±146 Fig. 3. Relationship between the part de¯ection and the number of passes by the laser for an ideal laser forming control. If the steps are too small however, processing time could be very long. A system is therefore required which could switch from rapid forming in the early stages of processing a given part to smaller forming increments as the target deformation nears, so minimising overshoot while keeping processing times short. This ideal is illustrated in Fig. 3. This concept is analogous to the roughing and ®nishing cuts used when turning a bar to a diameter. For this approach to succeed two key factors had to operate e€ectively. First, the control algorithm needed to be able to determine correctly at which stage the process parameters should be adjusted to produce the slowest forming rate. This would avoid both the risk of overshoot by switching too late or excessive processing time through changing too early. Secondly, the adjustment of the process parameters must be able to provide the control system with a range of suitable and predictable forming rates. For simplicity, the traverse rate at which the specimen was passed beneath the beam was used as the control parameter since this could easily be accessed through the CNC controller's interface. A high traverse rate for example, would impart less energy to the specimen than if this traverse rate was lower and so, in general, the rate of fold associated with this setting would be small. Other process parameters, such as focal position or laser pulsing could have been used but had certain practical diculties, though these were related more to the speci®cs of the equipment used rather than fundamental problems. Fig. 4. De¯ection versus number of passes for basic control system (5 mm target de¯ection). within, the appropriate signal would be transferred to the CNC controller. For example, a small error indicated the target was close to being achieved, that a slow forming rate was therefore required and so a high feed rate would be used for the next pass. A very large number of tests were carried out using this basic system albeit with some variations to the error thresholds at which the process parameters were switched and also to the traverse feed rates triggered by each of the thresholds [20±22]. The results of a typical test are shown in Fig. 4. Examination of the results of this test and others in the same series show that the control system does not appear to be functioning as had been hoped. Comparing the example of Fig. 4 with the ideal in Fig. 3 it can be seen that while the process parameters were changed at the appropriate points they seem to have had little e€ect in reducing the laser forming rate. This results in the potential for quite large overshoots of the target de¯ection. Why was this happening? Tests were carried out using constant traverse feed rates for the whole of the forming operation and these tests were repeated for a variety of feed rates. Results are shown in Fig. 5. These tests show that while on average the laser forming fold rate does reduce with increasing feed rate, the rate of fold is not constant at high feeds but increases at a rising rate with respect to de¯ection. 3. Control system trials A control algorithm was written which allowed the user to enter a target minimum and maximum value. This algorithm then monitored the de¯ection of the strip as it was formed, taking a reading after each forward±reverse traverse of the specimen beneath the laser beam. After each reading the PC would compare the error between the current de¯ection and the target minimum value to a series of four ®xed thresholds. Depending on which threshold the error value fell 143 Fig. 5. Constant feed tests for a variety of traverse feed rates. 144 G. Thomson, M. Pridham / Optics & Laser Technology 30 (1998) 141±146 This then highlights the source of the problem with the initial tests since at high de¯ections, close to the target values, all traverse feeds appear to produce similarly high rates of fold (i.e. steep gradients on the graph). It does not however address why this is happening or how it can be dealt with to achieve a more satisfactory control system. Two possible reasons for the rising rate are suggested: temperature build up in the specimen leading to a reduction in material strength and sti€ness and/or the formation of a plastic hinge. Laser forming is unusual in comparison to most other forming operations in that it is non-contact, relying on internal stresses rather than any external pressure to create the deformation. In conventional folding a specimen will be subjected to external forces from clamps and presses. When these clamps are removed the specimen relaxes as the external forces and moments are removed from the specimen and springback occurs. In laser forming, because the moments are internal they remain present after the forming operation. Each subsequent forming operation incrementally increases the applied moment. This has certain advantages in allowing the process to produce accurate geometries without the need to overbend to compensate for springback as the part is removed from the formers. This feature however may also cause certain diculties. When a strip of steel, for example, is folded by normal mechanical means, the relationship between the moment applied and the degree of fold is not linear due to the elastic±plastic behaviour of the material. Instead, the angle of fold, while rising with the applied moment, will do so at a reducing rate. The internal stresses opposing the applied moment cannot increase inde®nitely but are limited by the yield strength of the material. Inverting this relationship shows that if the applied moment is increased at a steady rate, as it would be by laser forming, the de¯ection will increase at a rising rate. This can be seen in Fig. 6 and the shape of this curve is similar to that achieved experimentally for de¯ection versus number of passes at high speed traverse rates such as those shown in Fig. 5. Formation of a plastic hinge may therefore help to explain the high forming increments at large de¯ections. The plastic hinge analysis, while appearing to explain the nature of the folding process does not provide a full explanation. As can be seen in Fig. 6, the analysis was based on an assumption that the material exhibited elastic behaviour up to the yield point (s y) at which point it became perfectly plastic. In reality a degree of work hardening would occur which would allow the stress induced to increase beyond the yield stress as the material was distorted, albeit at a reducing rate. The e€ect of this would be to make the rising rate relationship shown in Fig. 6 less marked. Both the yield point and the degree of work hardening possible can drop with increasing material temperature however [23]. The e€ect of this would be to make the stress relationship closer to the assumption used in the earlier analysis. In other words, the material would become easier to bend as temperature rose. It was felt in practice that heat was building up in the specimen as it deformed and this increased with the number of passes beneath the laser. This would tend to reduce the yield point and the degree of work hardening of the material so making the deformation rate at high de¯ections greater than may have been hoped. By allowing cooling time between passes at high de¯ections, the temperature of the specimen could be kept relatively low thereby encouraging work hardening. This would allow for a larger internal moment in the lower regions of the specimen to help oppose that moment associated with the plastic deformation of the upper surface, so helping to control the e€ects of the laser forming process. 4. Process control development Fig. 6. Results of the theoretical model of de¯ection versus number of passes assuming plastic hinge formation and that each laser forming pass produces a constant increment to the applied moment (insets show schematics of the assumed through thickness stress distribution). A number of measures were taken to try and improve the control process. These included using a control algorithm which took account of the forming rate, allowing cooling delays and doubling the process sampling frequency. The control algorithm was modi®ed to allow decisions on selecting the correct feed rate to take account of the forming rate and not just the error. The basic control system made use of ®xed thresholds to determine when process parameters should be adjusted. This made it dicult to produce consistent results from day to day since non-controlled process variability such as subtle variation in material thickness or ®nish or laser power ¯uctuation produced G. Thomson, M. Pridham / Optics & Laser Technology 30 (1998) 141±146 undesirable variations in the forming rate and so compromised the e€ectiveness of the ®xed thresholds. The control method was therefore modi®ed to allow process changes to be triggered when the error fell within a ®xed multiple of the previous forming increment. The algorithm therefore was able to take account, to a certain degree, of how the uncontrolled process parameters were altering the forming rate. This modi®cation would not necessarily show itself as a process improvement immediately. However, over a large number of tests, the control based on the forming rate would be able to cope with variations such as the gradual degradation of optics or variations in di€erent component batches more readily than one based on ®xed thresholds. A second and quite obvious improvement was to sample the deformation after each pass of the specimen beneath the laser beam rather than after each forward± reverse double pass. This added some minor practical diculties but ensured that the process parameters were changed at a more optimum time and also ensured the process could be stopped as soon as the target had been met without the need for a return pass. The ®nal process improvement involved adding in cooling time delays as discussed earlier. This step was taken in an e€ort to help keep the forming increments at high de¯ections to a minimum by allowing a degree of work hardening. Fig. 7 compares the results of a series of tests using the original control system, with results when using the individual process improvements in isolation and ®nally a set of tests were performed with all the process improvements incorporated. It can be seen that the process improvements signi®cantly reduce the spread and mean overshoot above the target. The mean ®nal de¯ection overshoot Fig. 7. E€ect of various control re®nements on overshoot above 5 mm target de¯ection. Key: 1, original ®xed control, measuring on double passes; 2, relative rate control, measuring on double passes; 3, original ®xed control, measuring on double passes, cooling time delays added; 4, original ®xed control, measuring on single passes; 5, new control, (relative rate, measuring on single passes, cooling delays added). 145 dropped from a mean of 0.334 mm for the original control system to only 0.023 mm for the completely modi®ed system. Each of the individual improvements helps in this respect, though it appears that the total reduction in the mean value of overshoot seems greater than the sum of the individual improvements. This may be due to the test with the cooling delay improvement only having had these applied delays added after each double pass, while a dwell was added after each single pass when this feature was incorporated into the ®nal control system. 5. Conclusions These tests have shown that a feedback control system can be used to produce simple laser formed parts to a tolerance comparable with more conventional techniques. However, for the system to be fully e€ective care must be taken in the implementation of the control method. In particular, care should be taken to avoid heat build up within the specimen, which can give rise to large fold rates even if low bending moments are applied to the part. It is also important to include consideration of the folding rate in any control algorithm and not merely to rely on calculations based on deformation errors. Inclusion of the folding rate in the control algorithm helps the system with the natural irregularity in the fold rate caused by variations in process parameters such as the material ®nish. These parameters would otherwise be dicult and expensive to regulate suciently well to allow an open loop system to be used. From the work presented in this paper it is envisaged that a laser forming system could be used to create automatically simple components from sheet. The system could also be used to modify locally more complex devices. For example such modi®cations could be used as a mechanical `trimming' technique to introduce subtle deformations to assemblies to eliminate the build up of tolerance between the individual components. Alternatively electro-mechanical sensor devices could be mechanically adjusted to alter their behaviour subtly. This again would allow a reduction in tolerances or allow sensor customisation from a limited range of stock parts. While these suggested applications are more complex than the trials reported in this paper and clearly more work is required, the results presented here increase the possibility of using laser forming as a niche technology. References [1] Vaccari JA. The promise of laser forming. Am Mach 1993;June:36±8. 146 G. Thomson, M. Pridham / Optics & Laser Technology 30 (1998) 141±146 [2] Vollertsen F. Mechanisms and models for laser forming. Proc. Laser Assisted Net Shape Engineering 94. Erlangen, 1994. p. 345±60. [3] Vollertsen F, Rodle M. Model for the temperature gradient mechanism of laser bending. Proc. Laser Assisted Net Shape Engineering 94. Erlangen, Oct. 1994. p. 371±8. [4] Sprenger A, Vollertsen F, Steen WM, Watkins K. In¯uence of strain hardening on laser bending. Proc. Laser Assisted Net Shape Engineering 94. Erlangen, Oct. 1994. p. 361±70. [5] Vollertsen F, Holzer S. 3D-Thermomechanical simulation of laser forming. Proc. NUMIFORM 95: Simulation of Materials Processing: Theory, Methods and Applications. New York, 1995. p. 785±91. [6] Alberti N, Fratini L, Micari F. Numerical simulation of the laser bending process by a coupled thermomechanical analysis. Proc. Laser Assisted Net Shape Engineering 94. Erlangen, Oct. 1994. p. 327±36. [7] Fratini L, Micari F. The in¯uence of the technological and geometrical parameters in the laser bending process. Proc. International Symposium for ElectroMachining. Switzerland: Lausanne, 1995. p. 827±36. [8] Schuocker D. Dynamic phenomena in laser cutting and cut quality. Appl Phys B 1986;40:9±14. [9] Garcia de Vicuna GE, Beitialarrangoitia JC, Ghosh SK. Defects arising from laser machining of materials. In: Niku-Lari A, Mordike BL, editors. High power lasers. Oxford: Pergamon Press, 1989. p. 227±49. [10] Bondelie K. Sealed carbon dioxide lasers achieve new power levels. Laser Focus World 1996;August:95±100. [11] Hsu CR, Albright CE, Khakhalev A. The in¯uence of laser cavity gaseous impurities on the performance of an industrial CO2 laser. ICALEO 95. San Diego, Nov. 1995, p. 1118±27. [12] Ho€man P, Neubauer N, Deinzer G. Recent developments in laser system technology for cutting and welding applications. [13] [14] [15] [16] [17] [18] [19] [20] [21] [22] [23] International Symposium for Electromachining. Lausanne, April 1995. p. 785±800. Sun Y-S, Weng C-I, Chen T-C, Li W-L. Estimation of surface absorptivity and surface temperature in laser surface hardening process. Jpn J Phys 1996;35 (Part 1)(6A):3658±64. Kinsman G, Duley WW. Fuzzy logic control of CO2 laser welding. ICALEO 95, San Diego, Nov. 1995. p. 161±7. Watanabe M, Okado H, Inoue T. Features of various in-process monitoring methods and their application to laser welding. ICALEO 95. San Diego, Nov. 1995. p. 553±62. Miyamoto I, Mori K. Development of in-process monitoring system for laser welding. ICALEO 95. San Diego, Nov. 1995. p. 759±67. Chang DU. Real-time laser weld quality monitoring system. ICALEO 95. San Diego, Nov. 1995. p. 1128±37. Schwarzenbach AP, LaÈdrach P. Recent progress in laser processing. International Symposium for Electromachining. Lausanne, April 1995. p. 801±7. Garcia de Vicuna GE, Beitialarrangoitia JC, Ghosh SK. Defects arising from laser machining of materials. In: Niku-Lari A, Mordike BL, editors. High power lasers. Oxford: Pergamon Press, 1989. p. 227±49. Thomson G, Pridham M. Prototype and part manufacture using laser forming with feedback control. IMC-14. Trinity College Dublin, Sept. 1997. p. 753±61. Thomson G, Pridham M. A feedback control system for laser forming. Mechatronics 1997;7(5):429±41. Thomson G, Pridham M. Controlled laser forming for rapid prototyping. Rapid Prototyping J 1997;3(4):137±43. Lubhan JD. Deformation phenomena. In: Dorn JE, editor. Mechanical behaviour of materials at elevated temperatures. McGraw-Hill, 1961, ch. 12.