ODS Analysis and Vibration Solving PDF
ODS Analysis and Vibration Solving PDF
ODS Analysis and Vibration Solving PDF
Report international internship Startdate: Monday, June 5th 2006 Enddate: Thursday, January 18th 2007 Rob Jenneskens (R.J.), studentnr. 501016 Supervisor TU/e: Prof. Maarten Steinbuch Supervisor SNU: Prof. Jongwon Kim Section Control Systems Technology, TU/e
Contents
1 Introduction 2 Eclipse-II 2.1 Eclipse-II Platform . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.2 One-man Ride Machine . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.3 Vibration Issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 Vibration Analysis 3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2 Modal Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.3 Operational Deection Shape (ODS) Analysis . . . . . . . . . . . . . . . . . . 4 ODS on One-man Ride 4.1 Introduction . . . . . 4.2 Set-up . . . . . . . . 4.3 Measurements . . . . Machine . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7 9 9 9 10 11 11 11 12 15 15 15 16 19 19 19 21 23 23 23 23 24 24 29 31 33 35 35 35
5 ODS Results 5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.2 Time Signal . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.3 Crosspower Spectrum . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6 Design Modications 6.1 Introduction . . . . . . . . . . . . . . . 6.2 Changing Mass, Stiness or Damping 6.3 Dynamic Vibration Absorber (DVA) . 6.4 DVA Requirements . . . . . . . . . . . 6.5 DVA Design . . . . . . . . . . . . . . . 6.6 DVA Implementation and Tuning . . . 7 Conclusions and Recommendations References A Equipment, Procedure and Test Experiment A.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A.2 ODS Equipment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
. . . . . .
. . . . . .
. . . . . .
. . . . . .
. . . . . .
. . . . . .
. . . . . .
. . . . . .
. . . . . .
. . . . . .
. . . . . .
. . . . . .
. . . . . .
. . . . . .
. . . . . .
. . . . . .
. . . . . .
. . . . . .
. . . . . .
. . . . . .
. . . . . .
. . . . . .
CONTENTS A.3 Measurement Procedure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A.4 Test Experiment: Modal Analysis of a Simple Plate . . . . . . . . . . . . . 37 38 41 43 45 47
B Measurement Points C Time Signal D Design changes next prototype E Technical drawings
List of Figures
2.1 2.2 3.1 4.1 4.2 5.1 5.2 5.3 6.1 6.2 6.3 6.4 A.1 A.2 A.3 A.4 A.5 One-man ride machine . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Major vibrations in the circular and vertical columns . . . . . . . . . . . . . . Animation in LMS Test.Lab . . . . . . . . . . . . . . . . . . . . . . . . . . . . Schematic presentation of the hardware set-up . . . . . . . . . . . . . . . . . Denition of the coordinate system . . . . . . . . . . . . . . . . . . . . . . . . Time signal of operating area (start-stop) . . . . . . . . . . . . . . . . . . . . Non-simultaneously stopping of the motors . . . . . . . . . . . . . . . . . . . The crosspower spectrum . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Simplied DVA . . . . . . . . . . . . SN-curve for aluminium . . . . . . . CAD drawing of the assembled DVA DVA mounted to vertical column . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10 10 12 15 16 19 20 21 25 27 28 29 35 36 38 39 39 41 42 43 48 49
The LMS Scadas III 310 data acquisition system B&K 4294 exciter with B&K 4506 accelerometer The test experiment set-up . . . . . . . . . . . . Test experiment: attached accelerometers . . . . Test experiment: FRF . . . . . . . . . . . . . . .
B.1 Measurement points for rst sequence . . . . . . . . . . . . . . . . . . . . . . B.2 Measurement points for second sequence . . . . . . . . . . . . . . . . . . . . . C.1 Time signal plot . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . E.1 Technical drawings block . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . E.2 Technical drawings spring . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
LIST OF FIGURES
Chapter 1
Introduction
The Robust Design Engineering Laboratory (RoDEL) at Seoul National University (SNU), South-Korea, has developed a new parallel actuated platform, called Eclipse-II [1]. The platform is used for generating six degree of freedom motions in space. The Eclipse-II platform is very exclusive because it can generate continuous 360 degree rotations around all axes and translations in all directions. A real size motion simulator, based on the Eclipse-II platform, has been designed and manufactured. During tests, serious vibration issues were found. To be able to give an advice for solving these vibrations, the vibration shapes of the structure under operating conditions are determined. This is done by performing an Operational Deection Shape (ODS) analysis. When the vibration patterns of the structure are known, it is analyzed which components are critical. Possible design changes are proposed, including adding dynamic vibration absorbers. This report is organized as follows. Chapter 2 gives an introduction to the Eclipse-II platform. The motion-simulator is presented and the vibration problems are discussed. In chapter 3 the theory on vibration analysis and ODS is explained. Chapter 4 handles the ODS analysis of the motion-simulator. The results of these experiments are presented and discussed in chapter 5. In chapter 6 possible design modications are proposed. The design of the dynamic vibration absorbers is explained. Moreover, advice for implementing and testing the absorbers on the motion-simulator is given. Finally, some concluding remarks and recommendations follow in chapter 7.
CHAPTER 1. INTRODUCTION
Chapter 2
Eclipse-II
2.1 Eclipse-II Platform
Motion simulators are virtual reality systems. They assume the appearance of a real situation by using movements of a motion-base and audio-visual eects. Most current motion simulators use a hexapod motion base. A big disadvantage of such a base is that it has limited tilt angels (20 30 deg.). Therefore, the Robust Design Engineering Laboratory (RoDEL) has developed a mechanism capable of 360-degree tilting motion as well as translational motion. This parallel platform is called Eclipse-II. For more information about the kinematics of the Eclipse-II platform you are referred to Jongwon Kim et al. ([1] and [2]).
2.2
After extensive testing of scaled prototypes, a real motion simulator has been designed and manufactured. This machine is called the one-man ride machine1 and is shown in gure 2.1. Its specications are given in table 2.1. Table 2.1: Specications of the one-man ride machine Machine dimensions (L x W x H): Machine weight: Kinematic workspace ( x H): Max. linear speed: Max. linear acceleration: Max. angular speed: Max. angular acceleration: Number of axes: 4200 x 4200 x 4700 mm 12000 kg 236.6 x 491.3 mm 600 mm/sec 4900 mm/sec2 (0.5 g) 120 deg/sec 500 deg/sec2 9
In the hereafter the Eclipse-II one-man ride motion-simulator will be referred to as one-man ride machine
10
CHAPTER 2. ECLIPSE-II
Vertical Column
Circular Column
Electronics Box
Cabin Motor
2.3
Vibration Issues
During tests, some unexpected vibrations were found. Especially, when the vertical and circular columns are rotating around the vertical axis and high decelerations are applied. In that case the columns are vibrating heavily (see gure 2.2), causing the cabin to vibrate. It should be noted that the vibrations shown in the gure are rst assumptions, based on visual observations. The eects of vibrations in machines can be numerous; noise and annoyance, but also, wear, fatigue, position inaccuracy or even failure (irreversible damage) can occur. Of course, in a virtual reality machine no unwanted vibrations should be present.
Figure 2.2: Major vibrations in the circular and vertical columns (scaled)
Chapter 3
Vibration Analysis
3.1 Introduction
Dierent experimental techniques can be used to analyze vibrations. The most common techniques are experimental modal analysis and operational deection shape analysis. In experimental modal analysis the exact dynamic behavior of a structure can be determined. This is explained in section 3.2. In operational deection shape analysis the vibration shapes under real operating conditions can be determined. This is explained in section 3.3.
3.2
Modal Analysis
The dynamic properties of a structure can be determined by FEM modal-simulations (Finite Element Method), or by experimental modal analysis. As the machine already has been built, an experimental approach is chosen. In experimental modal analysis the FRF (Frequency Response Function) of a system is determined. The FRF is a model of a linear system. It is the relationship between the measured output (e.g. displacement) and input (e.g. force), as a function of frequency. When both the applied force and the response to it are measured simultaneously, the FRF can be calculated. From this FRF the natural frequencies, mode shapes and modal damping can be obtained [3]. In modal analysis an important assumption is that the measured structure is isolated from its surroundings (e.g. by suspending it by elastic springs), so that no external forces are acting on the system. Considering the size of the one-man ride machine, it is very dicult to isolate it from its environment. Moreover, there is no equipment available to simultaneously apply and measure input forces of this (high) level. Therefore, it is chosen to determine the vibration shapes under operating conditions, using operational deection shape analysis.
11
12
3.3
Introduction In ODS analysis the vibration shapes of a structure, under operating conditions, are determined. The output of the system can be any number of things (displacement, acceleration, etc.). In contrary to modal analysis, the forces acting on the system (inputs) are not measured and, thus, unknown. The advantages of ODS with respect to modal analysis are [4]: + there is no assumption of a linear model + the structure experiences actual operating forces + true boundary conditions apply The disadvantages are [5]: no complete dynamic model is obtained, so no natural frequencies, mode shapes and damping properties can be determined operational deection shapes only reect the cyclic motion at a specic frequency, but no conclusions can be drawn for the behavior at dierent frequencies In the case of the one-man ride machine, the measured output is acceleration. With these acceleration signals the vibration shapes of the structure can be animated as follows. In specic ODS software the structure is represented by a simple wire-frame. The measurement points are positioned on this frame. The signals are used as inputs for the points on the wire-frame, and shown as (scaled) displacements. For each measured frequency dierent movements can be animated. This allows the designer to study the vibrations and to identify the critical components. An example of such an animation is given in gure 3.1.
Figure 3.1: Animation in LMS Test.Lab DOFs For each measurement point accelerations can be measured in three perpendicular directions. One combination of position and one direction is a Degree Of Freedom, or DOF.
3.3. OPERATIONAL DEFLECTION SHAPE (ODS) ANALYSIS Ideal situation In the ideal situation, all DOFs for ODS analysis are measured [5]: 1. simultaneously, 2. under constant operating conditions, 3. with signals having a high signal to noise ratio, so that no averaging is required.
13
Item 1, obviously, depends on the number of DOFs to be measured and on the number of acquisition channels on the measurement system. Item 2 depends on the complexity of the system and whether all DOFs are measured simultaneous or not. When not all DOFs are measured simultaneously, the operating conditions at dierent measurements can be dierent. Item 3 depends mostly on the equipment and measurement techniques used. It is very important that the measurement equipment does not inuence the dynamic behavior of the machine. Therefore as little mass, stiness and damping as possible should be added to the structure. Time ODS There are three types of ODS [4]: 1. Time ODS 2. Spectral ODS 3. Run-up/down ODS Time ODS is used to investigate the vibration pattern of a structure as a function of time. Spectral ODS and Run-up/down ODS, in contrary, investigate the vibration pattern at discrete operating frequencies. This is typically done in continuous rotating equipment. Because the vibrations in the one-man ride machine occur at movements with high accelerations and decelerations (e.g. run-stop), time ODS is chosen to be the best method. The measured time signal can be analyzed to see what happens to the machine over time. Reference DOF In ODS one specic DOF j is used as reference station. If operating conditions between dierent measurements change, they hopefully change in the same way for all DOFs, so the change can be canceled out due to the relative measurement. The reference DOF also provides a xed phase relationship between the dierent response signals. Crosspower spectrum When it is assumable that the operating conditions will not change between the dierent measurement sequences it is possible to just measure the crosspower spectra between each DOF i and the reference DOF j. The crosspower spectrum is dened as
Gij () = Xi ()Xj ()
(3.1)
where X denotes the complex conjugate [5]. Crosspower functions have the advantage that peaks clearly indicate high response levels. Obviously, crosspower functions are very useful when all DOFs are measured simultaneously.
14
Chapter 4
The measurement procedure and the measurement equipment are described in appendix A. To become familiar with equipment, software and the measurement procedure, a test experiment on a simple plate has been performed. This test experiment is also described in appendix A. The experiments on the machine were repeated once, because the results of the rst set of experiments were not satisfying. The major reason was found to be the presence of too much electrical noise. This chapter describes the procedure. Furthermore, the changes of the second experiments with respect to the rst experiments are mentioned throughout this chapter to show what was undertaken to improve results.
4.2
Set-up
Setup hardware and dene geometry Considering the high level of electrical noise on the rst experiments, the LMS Scadas III 310 measurement system was electrically ground during the second experiments. This reduced noise signicantly. Moreover, to enable the simultaneous measurement over 54 channels, an extra data-acquisition system was added. The LMS Scadas III 310 (described in appendix A.2) was used as master system during both sets of experiments. The LMS Scadas III 316 was added as slave system during the second set of experiments. It is similar to the LMS Scadas III 310, only it has 18 channels in stead of 36. The hardware set-up is shown in gure 4.1. Furthermore, a global coordinate system is dened, see gure 4.2.
Slave Scadas316 Laptop Triaxial accelerometers (6x)
Scadas310 Master
16
Figure 4.2: Denition of the coordinate system Dene DOFs To dene the DOFs to be measured, rst, appropriate measurement points are selected. To be able to suciently animate the vibrations, measurement points are distributed on the main components. Six points are dened on the circular column and ve on each vertical column. Because all movements of the cabin are in direct contact with the user, these should be identied well. In the rst set of experiments, these movements were not clearly visible. Therefore, ten measurement points (in stead of four) are dened on the platform. Four of them are placed on the frame, containing the cabin. Three points are dened on both the top and bottom surface of the cabin. Furthermore, measurement points were dened on the main joints. In gure B.1 (appendix B) the measurement points on the circular column and on the platform are shown. Also a closer look on some main points is presented. In gure B.2 the measurement points for both vertical columns are shown. Considering the complexity of the machine, it is chosen to measure the acceleration at each measurement point in all three directions (x, y, z). Dene measurement sequence Considering the number of DOFs to be measured, the experiment is split up in two parts: 1. First, the vibrations of the platform and the circular column are measured (18 points simultaneously, see gure B.1). 2. Second, the vibrations of the two vertical columns are measured (12 points simultaneously, see gure B.2). The 2 sets of data are combined using the reference DOF (accelerometer number 1).
4.3
Measurements
Setup analyzer The analyzer software is set-up with the following conguration (There was no reason to change these settings after the rst set of experiments). The sample frequency is 2048 Hz,
4.3. MEASUREMENTS
17
within a frequency range of 1-1024 Hz. This frequency range is sucient as low-frequent vibrations are expected. Operating the machine and starting the measurement is done by dierent persons. A time of 5 seconds ensures overlapping of measurement and operating, so a measurement time of 5 seconds is chosen. Calibration and attachment of accelerometers After setting up all hardware and making all connections, the accelerometers are calibrated (at the measurement location). Thereby, there is no temperature and humidity dierence between the calibration location and the measurement location. The accelerometers are attached to the structure, and local and global coordinate systems are written down. Make measurements and validate measurements To cause the machine to vibrate, all three motors (two on the vertical columns and one on the circular column) are actuated simultaneously. When the platform is rotated 3.6 degree around they are stopped. Several trial measurements are done. The crosspower spectrum is calculated for each measurement. When the measurement data seems repeatable and noise seems acceptable, three measurements are done. All data is saved. During postprocessing it is decided which data set is best. It should be noted that, as the technique used is Time ODS, no averaging can be done and no coherence can be calculated to ensure good measurement data. The accelerometers are relocated to measure the second sequence (i.e. measuring the vibrations of the vertical columns). One accelerometer is kept in place (number one). This reference accelerometer is used to couple the two sets of data. Results are presented in next chapter.
18
Chapter 5
ODS Results
5.1 Introduction
The time signal and crosspower spectrum plots of all six measured sequences have been analyzed thoroughly. The best sequences were combined to one set after these analysis. This chapter is based on that data set. The animation of the geometry, vibrating in time, has been studied to determine the structures behavior in time during the experiment. The animation at certain frequencies have been given a closer look to determine problematic frequencies. The most pronounced results are presented in this chapter.
5.2
Time Signal
A plot of the time signal is shown in appendix C. Dierent peaks can be identied in the time signal. A close up on the rst three peaks is presented in gure 5.1. In this gure, the rst peak (at t = -1.60 s.) is the starting point (when all three motors are started simultaneously). The start-up results in no signicant vibrations of the structure. The second and third peaks (at t = -1.24 and -1.21 s.) are caused by the non-simultaneously stopping of the motors. This can be seen when the animation of the time-signal is run.
19
20
In gure 5.2a the scaled deformation of the structure (in this time animation) at t = -1.24 s. is shown. The whole structure seems in normal shape, except the right vertical column, which is deformed heavily. This deformation is caused by the high deceleration applied by the motor attached on that corner. In gure 5.2b the scaled deformation of the structure at t = -1.21 s. is shown. Now, the left vertical column and the circular column are deformed heavily. This is caused by stopping the two concerning motors. As the time signal shows, there is a dierence in stopping time of 0.03 second. In the time signal between start and stop peaks, there are relatively low vibration levels. As all motors are started simultaneously, it is expected that stopping all motors simultaneously will decrease the vibration level too. It should be noted that it is not sure wether the vibrations will increase or decrease when all motors are stopped simultaneously. The rest of the time signal seems very chaotic, the structure is vibrating heavily. To analyze if there are repeating movements in the structures, the crosspower spectrum is studied.
21
5.3
Crosspower Spectrum
The crosspower spectrum is shown in gure 5.3. Four (low-frequent) peaks can be identied; at 3.65, 5.06, 5.72 and 7.35 Hz. The moving animations of the structure show the vibration shapes at these frequencies. These vibration shapes will be explained here. 1. 3.65 Hertz At this frequency a combination of motions can be identied: Both rectangular columns are vibrating in, what seems to be, their rst bending mode. The circular column is vibrating heavily. As a result of these vibrations, the cabin is rotating around . This peak is clearly the highest peak in the crosspower spectrum. 2. 5.06 Hertz and 5.72 Hertz At these frequencies, especially at 5.06 Hz, the platform is vibrating heavily. It is performing a translational motion in x-direction. The vibrations at both frequencies are comparable. It is noticeable that the columns are hardly moving. That is why it is assumed that this movement is a mode of the platform at 5.06 Hz. These peaks are signicantly lower than the peak at 3.65 Hz, though. 3. 7.35 Hertz At this frequency the platform translates in z-direction. It is noticeable that this frequency is almost exactly twice as high as frequency number 1 (3.65 Hz). Probably it is a harmonic frequency which is excited by the vibrating vertical columns. All other peaks in the crosspower spectrum have lower amplitudes and are less distinct. Lowfrequent vibrations were assumed to form the biggest problem before doing the experiments. These assumptions are conrmed by the results of the experiments. Phase lag between DOFs located at joints has been searched for to detect looseness of joints. No looseness of joints was detected.
3.65 7.35 5.06 and 5.72
22
Chapter 6
Design Modications
6.1 Introduction
Chapter 5 made clear that there is a dierence in stopping time of the three motors driving the columns. It is unclear wether changing the control algorithms (so that all motors are actuated simultaneously) will result in an increased or decreased vibration level. This should be tested rst. It will be done by members of the RoDEL laboratory. When changing the controls does not give the desired eect, this chapter provides other solutions to the vibration problems. To eliminate or decrease the level of the vibrations, several changes can be considered. The constructions stiness and inertial parameters can be changed, to tune them against resonances. Damping properties can be increased by using other materials. Finally, dynamic vibration absorbers can be used to reduce the vibration level [6].
6.2
The best solution to solve a vibration problem is to separate the components natural frequency from the exciting force frequency. In this case, the high decelerations are required to simulate real-life situations in ghter-aircrafts. Therefore, the exciting forces cannot be changed. The natural frequency can be changed by increasing or decreasing the components mass or stiness. Furthermore, the vibration level can be decreased by increasing the constructions damping properties, by using dierent materials. The structural changes to bring the problematic frequencies to another level are expected to be comprehensive. It is expected that the budget of RoDEL laboratory is not sucient for realizing the required changes. Therefore, it is decided to decrease the vibration problem using Dynamic Vibration Absorbers (DVAs). Nonetheless, some design changes for the next prototype are suggested in appendix D. The design of the DVAs is explained in the next sections.
6.3
A DVA is a device designed to have the same natural frequency as the (problematic) component it is mounted to. The DVA will vibrate with the same frequency but with a phase dierence of 180 degrees, reducing the initial exciting force. The theory behind the working principle of a DVA is explained in Den Hartog [7]. In general, a DVA consists of a simple 23
24
mass-spring system; an additional mass, connected by an elastic element to the source of vibration [6]. DVAs have several advantages over changing the construction geometry: + DVAs are relatively cheap. + While changing geometry can be a matter of trial and error, designing a DVA is a straightforward procedure. + Temporary DVAs can be attached. If the temporary absorber does eectively reduce the vibration, it can be left in place till a more permanent solution is designed and installed. If the absorber is not eective, it can be removed with the structure being unchanged. There is one disadvantage of DVAs: If the problem is not resonance, adding a DVA will create a resonance problem. Therefore, if adding a DVA does not give the desired eect, it should be removed again. Broadly speaking there are two types of DVAs; undamped and damped DVAs. Undamped DVAs work on a narrow frequency band. Damped DVAs work on a wider frequency band, but their eect is less. Moreover, a damped DVA inuences the structure at dierent frequencies [6].
6.4
DVA Requirements
The vertical columns cause the biggest peak in the crosspower spectrum. It is chosen to design a DVA for these columns rst. Because the vibrations at 3.65 Hz show a very narrow peak, it is chosen to design an undamped DVA. Before designing the DVA for the vertical columns, clear requirements are set: 3.65 Hz The natural frequency should be around 3.65 Hz Tunable The DVA should be tunable, i.e. the natural frequency can be varied in a range ( 10 percent) Safe The DVA should be safe, a lifetime calculation should be made Costs The DVA should be low-cost, so easy to manufacture (drilling, milling), no exotic materials will be used
6.5
DVA Design
To keep the design easy to manufacture a simple design is desired. The natural frequency n of the DVA is determined by its mass and stiness. By simplifying the design, these two parameters are separated. The stiness is variable by changing the length l of a cantilever beam, dened as spring (see gure 6.1). The mass, dened as block, is not variable. To achieve a tunable natural frequency the block can slide on the spring. In the following the dimensions in lowercase are dimensions for the spring and in uppercase for the block; t and T for thickness, w and W for width and l and L for length. The thickness
25
Vibration
Structure Sliding l
Figure 6.1: Simplied DVA T of the block is chosen to be 20 times higher than the thickness t of the elastic beam. The resulting stiness is 203 (see equation 6.4) or order 8103 times higher and can be assumed innite. Consequently, it can be assumed that the stiness (read: exibility) of the system is determined by the spring. From Korenev et al. [6], it is found that the mass ratio of DVA to structure can lie between 1:5 and 1:20 to give the desired eect. It is chosen to design a DVA with a mass ratio of DVA to structure of 1:10. For the vertical column (including electronics box and part weight of cabin, about 700 kg) this means that the total mass of the DVA should be 70 kg. Den Hartog [7] gives analytical equations for the simplied model of the DVA. A cantilever beam with a mass M attached to the end has a natural frequency of k (6.1) M + 0.23m with stiness k and mass m of the beam. A natural frequency of 3.65 Hz and a total mass of 70 kg is required. The stiness of the spring should be n =
2 k = n (M + 0.23m) = (3.65 2)2 70 = 36.8 103 N/m
(6.2)
The stiness k of a cantilever beam can be calculated by 3EI (6.3) l3 with Youngs Modulus E and second moment of inertia I of a rectangular cross section k= wt3 12 Looking at equation 6.3 the stiness is determined by the parameters E, I and l. I= (6.4)
Material choice To achieve a low natural frequency, low stiness is required. The allowable strain for a material is dened by y /E. It gives an indication of how much deformation the material can tolerate. The properties of carbon steel, spring steel and aluminium are given in table 6.1. Aluminium has the lowest Youngs Modulus. Moreover, it has the highest y /E-value. Spring steel also has good properties, but it is not easy to manufacture. Therefore, aluminium is chosen. An extra advantage of aluminium is that it hardly corodies.
26
Table 6.1: Material properties [kg/m3 ] 7830 7830 2800 E [GP a] 207 207 70 y [M P a] 500 1000 500 y /E 2.4 103 4.8 103 7.1 103
Dimensions block 70 To realize a mass of 70 kg a volume of m = 2800 = 0.25 m3 of aluminium is required. To achieve low stiness of the cantilever, the length l should be maximized. Smart placement of the DVA is necessary. The total length of the absorber Ltot = l + L can not exceed 750 mm, as that is the maximum space between the vertical column and the frame. This place is chosen for mounting the DVA (which will be shown in section 6.6). To leave enough space for the length of the spring, the length L of the block is set to 200 mm. As mentioned, the total thickness T is 20 times the thickness of the spring, or 300 mm. The resulting width W to achieve a volume of 0.25 m3 is 417 mm. Dimensions spring Setting the initial length l of the spring to 500 mm leaves 20 mm outward movement of the block for decreasing the natural frequency and 30 mm space between DVA and frame. As will be shown later, the fatigue stress depends on the surface properties of the material. A smooth surfaces will fail at a higher number of cycles than a rough surface with notches and cracks. Therefore, it is chosen to use aluminium of a standard thickness. This standard rolled plate has good surface properties which should not be altered by using cutting techniques. The sides of the plate should be nished using peripheral milling. It is very important that the sides are machined using down milling (or climb milling), so the chips will not damage and roughen the nished surface. It is found that aluminium plate is available in standard thicknesses of 10 and 15 mm. Now, the dimensions of the spring for a thickness of 10 and 15 mm are determined. The following parameters are known; k, E and l. Equation 6.3 written explicit for I gives 36.8 103 0.53 kl3 = = 2.19 108 (6.5) 3E 3 70 109 The maximum stress in the spring can be calculated when the maximum force at the end of the spring is known. From tabel 2.1 we know that the maximum angular acceleration of the platform is 500 deg/sec2 or 2/360 500 = 8.73 rad/sec2 . The distance (radius) between mass and axis of rotation (z-axis) is approximately 2.4 m. This gives a maximum linear (tangential) acceleration of 2.4 8.73 = 20.95 m/sec2 . The bending force F on the spring is then F = m a = 70 20.95 = 1467N . I= From Fenner [8], we nd that the maximum stress in a cantilever beam of length l with a concentrated force F at the end is max = F Lt 2I (6.6)
6.5. DVA DESIGN For a thickness of 10 mm we nd max = For a thickness of 15 mm we nd max = 1467 0.6 0.015 = 301M pa 2 2.19 108 1467 0.6 0.010 = 201M pa 2 2.19 108
27
(6.7)
(6.8)
Fatigue life of spring The fatigue stress is a value for the allowed stress in a material to withstand a number of cycles until failure. It depends on material properties, the surface properties and on the number of load cycles. To be able to guarantee safety over a sucient test period, the number of load cycles is determined: It is expected that the DVA vibrates over a maximum time of ve seconds per stop. This means a maximum of 20 vibrations (at 3.65 Hz) or load cycles per stop. The DVA will be designed for 1000 start/stop-runs. This is sucient to test the DVA extensively. 1,000 runs equals 20,000 cycles. A safety factor of ve is chosen to guarantee safety over the test period. Failure should not occur before completing 100,000 cycles. It should be noted that the spring should be removed after 1000 start/stop-runs, though. The most common way to express fatigue stress is with SN-curves. In SN-curves the number of cycles at which failure occurs is plotted against the maximum occurring stress in these load cycles. In gure 6.2 a SN-curve for aluminium is shown. As can be seen in the gure, aluminium parts with a smooth surface have a longer lifetime then parts with a rough surface. Plate of standard thickness and peripheral milled sides is used. The gure shows that the maximum allowed stress for a lifetime of 105 cycles, with a normal to smooth surface is 30 ksi, or 207 N/m2 . Comparing this value with the maximum stress calculated in 6.7 and 6.8 shows that only the spring with a thickness of 10 mm satises the lifetime calculation.
28
Realization Using equation 6.4 we can nd that for a thickness t of 10 mm the width w of the spring becomes 263 mm. The width w of the spring is much higher than the thickness t. This ensures that the mass will vibrate in the desired (exibel) direction. The total length Ltot extending from the column is 720 mm. It is composed as follows; working length spring l 500 mm, extension for clamping the block 200 mm, extension for tunability 20 mm. The plate is extended for 200 mm inwards so it can be mounted to the column. The total length becomes 920 mm. Four 12 mm holes are drilled in the plate, which can contain 12 mm threads to mount it to the column. The block is divided into two parts. Each block contains four 12 mm holes, and is clamped on the spring with 12 mm threads. The blocks can slide on the spring to change the natural frequency of the DVA. The only manufacturing techniques that should be used are drilling and milling. In table 6.2 the dimensions for spring and block are shown. A CAD drawing of the DVA is shown in gure 6.3. Technical drawings can be found in appendix E. It is advised that the blocks are connected to the column with a steel cable, for extra safety. This should be included in the design. Table 6.2: Dimensions for DVA t/T [mm] 10 200 w/W [mm] 263 417 l/L [mm] 920 200
Spring Block
29
6.6
For safety, it is very important that the DVAs are xed well. It is advised that the blocks are connected to the column with a steel cable. Damping decreases the eect of the DVA [6]. Therefore, a hysteresis-free connection should be achieved. This can be achieved by bolting the DVA to the vertical column. Four 12 mm holes should be drilled in each vertical column. In gure 6.4 an example for the implementation of the DVA for the vertical column is shown.
Figure 6.4: DVA mounted to vertical column When the DVAs are attached, a rst run/stop can be operated. Visual inspection should be done to check if the initial vibration problem has decreased or increased. Iterative test procedure Because the intensity of the exciting force is not known, some trial-and-error experiments might be needed. The DVA should be tuned, such that its natural frequency exactly matches the natural frequency of the vertical column. This can be done by repeating this sequence: 1. run/stop the machine 2. check the eect of the DVA 3. move the DVA on the spring 4. start at number 1. If improvements are visible, ODS can be performed again to verify the observations. If no improvements are seen, the DVA should be removed, as it can introduce a resonance in the structure.
30
Chapter 7
31
32
Recommendations
The control algorithms should be changed, such that all three motors are actuated simultaneously. If changing the control algorithms does not decrease the vibrations (at high decelerations), dynamic vibration absorbers can be attached to the vertical columns. It is very important that the production methods described in this report are used. The sides of the spring should be nished using peripheral (or climb) milling. An extra connection between blocks and column is advised. A steel cable can be used. This is not included in the design. If the DVAs are eective, new ODS experiments should be performed to analyze the behavior of the construction. Before building a second prototype, a thorough dynamic (modal) analysis should be performed. In appendix D several design changes are proposed to improve the dynamic behavior of the machine. This report might be used as a reference for performing ODS experiments. This report might be used as a reference for designing DVAs.
Bibliography
[1] Jongwon Kim, Jae Chul Hwang, Jin Sung Kim, Cornel C. Iurascu, Frank Chongwoo Park, and Young Man Cho. Eclipse-II: A new parallel mechanism enabling continuous 360-degree spinning plus three-axis translational motions. IEEE Trans. Robot. Autom., 18(3):367373, June 2002. [2] Jongwon Kim and Sun Ho Kim. A new ghter simulator based on a full spinning six degree-of-freedom parallel mechanism platform. In I/ITSEC, 2005. [3] D.J. Ewins. Modal Testing: theory, practice and application. Research Studies Press, 2000. [4] Brel & Kjaer. ODS tutorial: Introduction, denitions and terminology, measurements u and validation, applications, 2004. [5] LMS International. Tutorial: Operational deection shape analysis, 2005. [6] Boris G. Korenev and Leonid M. Reznikov. Dynamic Vibration Absorbers. Wiley, 1996. [7] J.P. den Hartog. Mechanical Vibrations. McGraw-Hill Book Company, Inc., 1947. [8] Roger T. Fenner. Mechanics of Solids. CRC Press, 1999.
33
34
BIBLIOGRAPHY
Appendix A
In this appendix the equipment used for ODS experiments is presented. The equipment has been made available by the Acoustics & Vibrations Laboratory of the Advanced Automotive Research Center at SNU. The measurement procedure is divided in dierent steps, providing a clear plan to do the experiments. To become familiar with equipment, software and the measurement procedure, a test experiment on a simple plate has been performed.
A.2
ODS Equipment
Data acquisition front-end The LMS Scadas III 310 signal conditioning and data acquisition system is a 36-channel portable lab-device. It is shown in gure A.1. The system enables simultaneous measurement of 36 DOFs. It is connected to a PC through a PCMCIA-card.
Figure A.1: The LMS Scadas III 310 data acquisition system
35
36
Software The software for data acquisition, analysis and visualization of results is LMS Test.Lab. This software forms a perfect combination with the LMS Scadas III data acquisition system. The whole procedure of calibration, dening the measurement range, doing measurements, computing results and displaying animations is integrated in one software package. Accelerometers The accelerometers are Brel & Kjaer type 4506. These are piezoelectric triaxial acceleromeu ters with a low mass compared to the structure of the machine. The specications are given in table A.1. Each accelerometer is calibrated in all three directions before they are attached. They are mounted on the structure using Endevco accelerometer mounting wax or Henkel 401 bond. Table A.1: Specications of B&K 5406 accelerometer Sensitivity: Frequency range: Weight: about 10.00 mV/ms2 (is calibrated before each measurement) 1 - 6000 Hz 15 g
Vibration calibrator The device used for calibrating the accelerometers is a Brel & Kjaer 4294 exciter. It has u 2 . The operating frequency is xed at 159.15 Hertz (= a reference vibration level of 10 ms 1000 rads1 ). It is shown in gure A.2. The measured vibration level is used as a reference for the accelerometer signal.
Figure A.2: The B&K 4294 exciter with the B&K 4506 accelerometer mounted to it
37
A.3
Measurement Procedure
This section provides a clear measurement procedure which was followed narrowly to prevent mistakes. The stepwise measurement procedure is divided in three main parts and is advised by Brel & Kjaer [4]: u 1. Setup: Setup hardware First, all hardware should be set up. All connections should be made and tested. Dene geometry The geometry of the test object should be well-dened, so that the DOFs to be measured can be chosen. Dene DOFs Before attaching the accelerometers to the structure, suitable DOFs should be determined. A rule of thumb is that there should be something (acceleration) to measure on that specic position in that specic direction. To detect looseness of joints, either sides of joints are also good measurement points. Dene measurement sequence In the case that there are less input channels (on the data acquisition system) then DOFs to be measured, a measurement sequence should be chosen. It is very important that operating conditions do not change between the dierent measurements. As Time ODS is used, averaging is not possible. 2. Measurements: Setup analyzer First step of setting up the analyzer consists of choosing the frequency range and resolution. Time signals and crosspower spectra should be measured. All DOFs should be linked to a dierent channel in the ODS software. Calibration The transducers are calibrated using the exciter described in section A.2. Therefore the accelerator is placed on the exciter. The ODS software has an auto-calibration function, which uses the vibration level of the exciter as a reference. Attach accelerometers and make trial measurement The accelerometers are attached using Henkel bond. This provides a reliable, light and sti connection. The wires should not be entirely xed to the structure. When the wires hang loose, they vibrate less, thereby generating less noise in the signal. Make measurements When all transducers are calibrated and well-xed and the measurement range is chosen, the measurement can start. The machine is operated at normal, but problematic, accelerations. Validate measurement After measuring, the data should be veried. When the results seem not satisfying, both the equipment and the set-up should be checked and the measurement should be repeated.
38
APPENDIX A. EQUIPMENT, PROCEDURE AND TEST EXPERIMENT 3. Results and documentation: Transfer measurements to animation program When the measurement is successful the data can be transferred to an animation program. In this case that is another module in the same software package. The DOFs can be linked to the corresponding directions in the nodes of the dened geometry. For dierent frequencies the response of the structure can be visualized. Geometry animation Now, the designer can analyze the vibration animation and can decide which components are critical in what direction. Depending on this information, modications in the design of the critical components can be done to reduce vibrations.
A.4
Introduction Before going to Incheon (location of the one-man ride machine) to do the ODS experiments, a test experiment is performed. In this way one becomes familiar with the software and equipment before the real experiment is performed. Moreover, the experimental procedure can be practiced. Modal analysis of a simple plate To keep the practice simple and veriable it is chosen to do the test experiment on a simple plate. Because a simple plate does not have operating conditions it is chosen to do a modal analysis experiment. However, the same equipment and software is used. The plate is isolated from its environment by attaching it to a frame through weak springs (see gure A.3).
39
When the analyzer is set, the geometry is dened. A grid of six points on the plates surface is dened. Measurement points are chosen on locations away from nodes of modes that are expected. Each of the six accelerometers is calibrated in all three directions (x, y, z) and positioned on the specied points, with the corresponding orientation (see gure A.4).
Figure A.4: All accelerometers attached in the right orientations Now, the trigger for hammer-impact can be set. The impact frequency range is chosen by impacting several times. When all ranges are well-dened, the roving hammer experiment can begin. To reduce noise, for each point an average of ve impacts is taken. When all six points are excited, the results are analyzed. The accelerations can be transformed to displacement signals by integrating twice. These displacements can be linked with points on the wireframe, thereby animating the vibrations of the plate. The rst six natural frequencies can easily be recognized in the FRF (see gure A.5). The rst six modes (three bending and three rotation modes) can be visualized in the animation. After performing the test experiment, the equipment is known well and the measurement procedure is clear. 50
dB
-40 0 Frequency 1024 Figure A.5: In the FRF of the test experiment the rst six natural frequencies can easily be recognized
40
Appendix B
Measurement Points
12 18 1 (Ref.) 2 17
13
12
11
14 9
16 3 4 15 18
41
42
29 19 19
1 (Ref.) 28 20 25 26 27 21 22 25 26 28
23 23 24
Appendix C
Time Signal
Figure C.1: Plot of the time signal that was selected best
43
44
Appendix D
45
46
Appendix E
Technical drawings
On the next two pages, the technical drawings for blocks and spring can be found. These drawings are downscaled from A3 size. The block is on a scale of 1:2 and the spring on a scale of 1:5.
47
48
49