ae GD Audi Linearisation of stiffness and damping characteristics of rubber bushes in vehicle dynamics by using frequency response analysis of MSC/NASTRAN. Authors: Dr. Stefan Démok P+Z Enginooring GmbH Anton Ditt Bogon 3 D-80939 Munich Dr. Eckhard Plank, Dr. .Jiirgen Merk AUDI AG VET-112 D-85045 Ingolstadt Abstract: For vehicle dynamics analysis detailed modelling of bushes is necessary. The following paper describes different possibilities to use measurement data of rubber bushes by linearisation of stiffness and damping characteristics in frequency response analysis. 41st MSC Worldwide Autamativa Confarance 1999 Pana 1 of 141 Introduction Analyses in tho field of driving comfort aro of great importance today in the development of motor vehicles. Driving comfort is influenced decisively by all vibrations excited during the operation of the vehicle. The causes of these vibrations are imbalance, uneven road surfaces and engine shake. In the case of uneven road surfaces, for example, vibrations ara transferrad from the tires via the suspension, the body, seats and even the steering wheel to the driver. MSC/NASTRAN is a recognised tool for analysing driving comfort and hence the calculation of driving dynamics. For this purpose MSC/NASTRAN offers the solution SOL111 "modal frequency response", This frequency response analysis describes the harmonic response of a system in steady state caused by harmonic and frequency-dependent excitation. In such case, the forces of excitation are of the same frequency but may have differing phase angles at various degrees of freedom. ‘The Solution 111 only takes Into account geometric linearity. This Is valld as tong as the amplitudes remain small, which is the case in most driving situations. It is only under patticuks driving conditions such as on very uneven surfaces or for the analysis of the structure under abuse loading that geometric linearity is no longer guaranteed. In a vehicle system analysis, material non-linear effects occur particularly in suspension bushes. This is because the dynamic stiffness of tha bushas is a function of the amplitude of the vibrations which occur. These material non-linear effects can not be taken into account with MSC/MASTRAN SOL111. But those bushes with non- linear material behaviour are exactly the ones which have a significant effect on the results of the calculation of driving dynamics. That is why their exact representation is crucial to obtaining calculation results which closely reflect the reality, The following are necessary in order to calculate a good vehicle dynamic response: 1. The dynamic stiffness of the bushes must be known 2. The material non-linear behaviour of the rubber bushes must be taken into account in a material linear calculation (MSC/NASTRAN SOL111) ‘Ast MSC Worldwide Automotive Canfarance 1999 Pana 2 of 14The following solution is offered: On servo-hydraulic test machines it is possible to measure the characteristics tor non-linear stitness and damping of each individual bush depending on preload, direction, frequency and amplitude, The non-linear materlal behaviour can then be taken Into account by means of an iterative procedure by establishing the displacement of each bush in an initial calculation. In a second step the stiffness of the bush is updated according to amplitudes based on measurements and a new analysis is performed. This methodology is repeated until the stiffness of the bush ceases to change or until the change is only minimal. Figure 1 shows the result of a frequency response analysis for a vehicle system model in the form of a fraquency responsa at a point which affects driving comfort Linearisation should result for the frequency at which maximum acceleration occurs. Acceleration at the seat rail acceleration frequency Figure 1 This iterative method of determining the dynamic stiffness of rubber bushes has been applied within the PELIT application by P+Z Engineering for AUDI AG and is explained in greater detail in this article. Yet MSC: Worldwide Aritamative Crnfaranea 1990, Pana Qaf1d2 Characteristics and modelling of bushes 2.1. Physical description ot the rubber bushes In regard to damping, a distinction is made between material damping and viscous damping. material damping viscous damping k k [> xx |» xx LS Li d h Figure 2: Material damping and viscous damping The force-displacement relationship is: Fa(kti-d)-x Fokt+hx With the dieptacoment formulation xexy The force-displacement relationship = (k+i-d) xe F = (b+ iho): age! The result for the complex stiffness Is therefore: Kjole = (45-4) Kamps =k +40 The dynamic stittness is the magnitude ot the complex stiffness, Ast MSC Worldwide Autamativa Confaranca 1999, Pana 4 of 14Tests have shown that dynamic stiffness depends on the following factors: * Preload on the bush under static load + Excitation frequency © Size of amplitude of the vibration These factors are explained in greater detail below. 3 Factors affecting the dynamic stiffness of bushes in a Frequency Response Analysis Preload These are the forces acting on the bush which result from the weight of the vehicle in a static state, The effect of preload is shown in figure 3. Factors are measured on a test rig. The dynamic stiffness is represented dependent upon the amplitude for two different preloads for a constant excitation frequency. The non-linear dependence can be clearly recognised. Dynamic sttiness vs. bration amplitude for 2 bush preloads ‘Moscuroment value fora constant frequency ‘rs ange Figure 3: Dependence of dynamic stifiness on proload 1st MSC Worldwide Autamativa Conteranca 1999 Pana Sof 14Excitation Frequency ‘At this moment frequency dependence is not taken into account in the procedure presented here. This simplification, however, must be examined from case to case. Measurements have shown that the dynamic characteristic values increase with an increase in frequency and a decrease In distortion. Vibration amplitude Figure 4 shows the dynamic stiffness as a function of the vibration amplitude. Also shown ere the real and imaginary parte. The illustrated curve is based on measurement results. ‘Real and imaginary par end dynamic efi dependent upon ampltude Measurement values fora constant requency == te me magury tt once ‘Arete Figure 4: Dependence of dynamic stiffness on vibration amplitude The non-linear effect dependent upon amplitude can be seen clearly here. These results show that the measurement data for various preloads and amplitudes must be available before it is possible to perform an accurate frequency response analysis. The dynamic stiffness can then be determined separately from the measurement results for each excitation frequency. tet MSC Worldwide Autamntive Canfarance 1000 Pana Ant 144 Modelling of Bushes with MSC/NASTRAN MSC/NASTRAN offers CELAS1 elements for the modelling of bushes. The damping used by these elements is material hysteretic damping. CBUSH elements are used for viscous damping. CELAS1 and CBUSH clements consist of two nodes and refer to a property card (PELAS or PBUSH). One ot more degraes of freedom for stiffness and damping can be given on both of these cards. In a frequency response analysis, acceleration, for example, is calculated by means of the frequency. As previously shown, stifness and damping are not identical for each amplitude and are therefore largely dependent upon the vibrational behaviour of the system. That means stiffness and damping can only be given exactly for one excitation frequency with use of the PELAS or PBUSH card. From version 70, MSC/NASTRAN offers additional options through the introduction of PELAST and PBUSHT cards. Both of these cards refer to a TABELDI entry in Which stiffness and damping are detined separately over a trequency range. How the non-linear material behaviour of such spring/damping systems can be dealt with in a frequency response analysis is explained in more detail in the following section. The PELIT application (PELAS LINEAR ITERATION) was developed especially for this purpose. This automatically generates and modifies the corresponding NASTRAN cards, 5 The PELIT Application PELIT is an application for the linearisation of rubber bush stiffness and damping around a defined operating point or for a larger frequency range in vehicle system analysis with MSC/NASTRAN SOL111. In order to use the application, the characteristic values of amplitudes and preloads for all relevant rubber bushes in the vehicle system model must be available, Using a database containing the characteristic values, PELIT determines the accompanying values for the real and imaginary part of the complex stiftness for the displacement amplitudes in a frequency response analysis which occur in each rubber bush direction. The corresponding NASTRAN cards are then modified automatically with these new calculated values. The calculation procedury is divided inlu twy different steps: 1. Calculation of preload for each individual bush 2. Iterative linearieation process for calculating stiffness and damping for an ‘excitation frequency. ‘1st MSC Worldwide Autamativa Confaranca 1999 Pane 7 af 145.1. Preload Calculation The procedure is shown in Figure 5. For the preload calculation, a vehicle system model is fixed at the wheel contact points and analysed under its own weight. A static, calculation with MSC/NASTRAN SOL101 delivers the displacements of all CELAS1/CBUSH nodes and issues these in a punch fle, PELIT then calculates the preload for each individual bush from the relative displacements of a CELAS1/CBUSH element and from the stiffness of the bush. At this moment the static stiffness is already linearised. It can be a result from a Multiple Body System. All results are summarised in a documentation filo. MSC/NASTARN input file for SOL101 (static) ¥ MSC/NASTARN SOL101 (static) MSC/NASTRAN punch file Including the displacements of nodes, related to all rubber bushes y PELIT Calculation of preload for each individual bush from displacements and the static stiffness of the bush Figure 5: Preload calculation ‘Int MSC: Worldwide Automotive Confaranre 1000 Pana Raf 145.2 _Linearisation of the stiffness in a Frequcnoy Response Analysis In order to use PELIT, all relevant rubber bushes must be measured on the servo- hydraulic test rigs. The measured factors, stiffness and damping are available as a function of preload, amplitude and direction and stored in a database. In regard to direction, it is important to note that the measured direction need not necessarily correspond to the degree of freedom in the FE model. Furthermore, data measurements for a bush on various directions (i.e. degree of freedom) may already be available in the bush database. The measurement values can be allocated to a bush by means of a bush niame. The principle procedure is shown in Figure 6. The access to a bush of the FE model occurs by means of a property card. Depending on the application, these are PELAS, PELAST, PBUSH or PBUSHT entries. The following factors are allocated to each property card in an allocation file: = Property number of the bush. This also means that the property may not refer to several elements ‘Type of property card (PELAS, PELAST, PBUSH, PBUSHT) Bush name ~ i.e. allocation to measurement data from the database Degree of freedom of the bush Preload Determining the dynamic stiffness for the first calculation This is termed 0. Iteration. Because no measurement values are usually available in tha database for the exact calculated preload, linear interpolation is performed between the individual preloads. The starting value for stiffness and damping is the mean of all amplitudes. PELIT then modifies the NASTRAN record according to these newly calculated factors. Ina SOL111 calculation, MSC/NASTRAN determines the frequency response for the selected excitation spectrum. The results - ie. the displacements of the nodes of the bushes — which should be linearised are written into a punch file. PELIT then reads in the complete NASTHAN input tile as well as the results trom the punch file. Stitmess and damping are linear interpolated trom the preload curves and te accompanying amplitudes. 1s important here ty note thal the displacements of bolt bush nodes are usually not in phase. The maximum displacement between both nodes is therefore determined over a 360° phase angle. This maximum displacement corresponds to the amplitude which is used as the basis for the new calculation of stifiness and damping. PELIT subsoquently modifics the bush values in the input record as woll. ‘1st MSC Worldwide Autamativa Confaranca 1999 Pana 9 of 14Which ocours in tho form of a MSC/NASTRAN calculation followed by tho modification of the bush value using PELIT, is repeated until there are only minimal changes in the frequency response curve (Figure 6). It is important here to note that linearisation is performed only for one excitation frequency. Between these iterative steps it is therefore necessary to represent the curve graphically in the form of a 2D plot, and to determine the frequency for which the bush valuos should be linearised. Because as a result of the modification of the bush stiffness, frequency displacement can occur while the individual iteration steps are being performed. P+Z uses the FE-Graph [1], a commercial tool the company itself has developed. For a driving comfort calculation, the examination of this frequency response takes place at comfort points which are selected by the user. Automatic procedure PELIT offers an automatic procedure in which the MSC/NASTRAN calculation and the modification of bush stiffness are performed automatically after each other. PELAST/PBUSHT cards in MSC/NASTRAN offer frequency dependent stiffness and damping. PELIT automatically updatco these eardo with cach iteration. Tho oritoria for torminating tho procodure ie oithor 2 maximum numbor of itorativo stope or a minimum chango in bush values. Bocause theso changes must bo determined over a patticular frequency range, the area below the curve of the TABLEDI card is used. PELIT supports this entire calculation procedure with the following characteristics: ‘+ Control over the complete Nastran input record in terms of input data format and how the individual cards relate to each other ‘* Automatic generation of the SET definition for all nodes of the bushes * Documentation of all calculated factors by PELIT in a documentation file These are: - Preloads - Stitfness and damping for all iterations * Reading in the bushes from the bush database and a table of all bushes containing the name of the bush, direction and preload, ‘1st MSC Worldwide Automotive Canfaranca 1999 Pana 10 of 14Bush allocation: # Property number and type of property © Bush name Direction « Preload PELIT 0. iteration Calculation of stiffness and damping for all bushes, based on measurement data Modification of bush data of NASTRAN input record MSC/NASTRAN SOL111 MSC/NASTRAN punch file PELIT 4h iteration Determination of stiffness and damping for all bushes from the measured bush data i MSC/NASTRAN SOL111 MSC/NASTRAN punch file Checking the iteration results Database of bushes Figure 6: PELIT iteration process 1st MSC Worldwide Automotive Canfaranca 1999 Pane 11 of 146 Examples of Application The results of a linearisation of dynamic stiffness are illustrated in an example. A FE vehicle model has been used. The FE model is shown in Figure 7. The performed calculation is only an example for demonstration of the procedure of PELIT. Ten bushes (symmetric, 5 on avery side) were modelled by means of rubber bushes in the region of the suspension. Alll of these were taken into account by PELIT in the iterations. Figure 7 shows only 5 bushes, 2 on the right side and 3 on the left side. Bushes used forth nearsation SX Figure 7: FE vehicle system model 1s! MSC Worldwide Automotiva Conference 1999 Pana 12 of 14This test load is a calculation ot excitation with the engine idling. The vibration of the engine caused by excitation in idle mode leads to body acceleration, lke shown In Figure 8. The bushes are modelled using CELAS1/PELAS cards in NASTRAN. A point at the seat rail was used as the comfort point for the evaluation of acceleration. The maximum of the acceleration is shown in Figure 8. At these maximum a linearisation of all relevant bushes were performed. The calculations were performed {as desoribod above with MSC/NASTRAN SOL111. The results aro illustrated below for iterations 0, 1, and 2. The effect of the modified bush stiffness can be clearly seen. Two iteration steps were necessary. Acceleration at the seat rail acceleration frequency Figure 8: Acceleration at the seat rail ‘1st MSC Worldwide Auitamative Canfaranaa 1999, Pana 18 of 147 Conclusion With the procedure described, it is possible to take into account material non-linear effects of rubber bushes in frequency response analyses. This requires an iterative process for determining amplitude-dependent stiffness and damping values on the basis of measured values. As a result it becomes possible to model the bush stiffness of rubber bushes in a frequency response analysis more accurate and therefore to evaluate the driving dynamics of a vehicle system model effectively on the basis of calculations. The PELIT software described above was developed to support the user in the iteration process. From version 70, MSC/NASTRAN offers PELAST and PBUSHT cards. These allow stiffness and damping values to be given variably over the entire frequency range. This procedure is also supported by PELIT. Tha primary disadvantage relating mainly to the CBUSH element is the significant increase in caloulation time. No results are currently available regarding calculation time when using PBUSHT cards. This will become apparent with use in praxis. 8 Bibliography [1] FE-GRAPH - Users manual. P+Z Engineering GmbH. tet MSC Warldwidy Auitamativa Confaranca 1999 Pane 14 of 14