Fatigue Strength Prediction of Truck Cab by CAE

Shinichi CHIBA* Kimihiko AOYAMA* Kenji YANABU** Hideo TACHIBANA** Katsushi MATSUDA** Masashi UCHIKURA** * **

Abstract
In order to accurately predict the fatigue strength of a truck cab, it is necessary first to estimate the input load history from cab mounts, then with the estimated value, to accumulate the damages from the stress time history generated by the input load, and finally to calculate the fatigue life. This paper introduces the method of combining cab input load estimated by multi body simulation and fatigue life estimation by FEM analysis and fatigue life analysis, and one example of the analysis of the change mechanism of input mode and fatigue life according to the changes of test condition. Key words: Fatigue, Body, Numerical Analysis, CAE

1. Introduction
In the conventional vehicle development, the fatigue strength of a truck cab must be evaluated by conducting physical tests. This is because in the conventional analysis, forces acting on the vehicle body are represented mainly by such static inputs as torsional and vertical bending forces, and the analysis results must be compared with the results of physical tests conducted on prototypes and actual vehicles before the fatigue strength can be assessed. In order to reduce the development period and number of prototypes, a technique is required that depends on so-called virtual prototypes at the drawing stage of development, not on physical prototypes. Thanks to recent improvements in computer performance and the appearance of multi-purpose fatigue life analysis applications on the market, engineers can conduct fatigue strength analyses that take into account even the stress history and multiaxial stress field, thus making predictions more accurate(1). The truck cab is usually fitted on the frame via four front and rear mounts, through which it receives random inputs, suggesting that in order to predict the fatigue strength, the stress history of composite inputs in a multiaxial stress field must be considered. As the first step toward the ultimate goal of fatigue strength prediction based on computer aided engineering (CAE) assuming rough-road durability test situations, the fatigue strength evaluation method presented in this paper assumes bench durability test situations and combines schemes for predicting cab input loads and those for predicting fatigue strength. The paper also describes the cases of using the method to investigate the causes of cracks in a light-duty truck during the development stage.

2. Approaches for assessing cab fatigue strength

2.1 Durability test approach The fatigue strength of the cab is generally assessed through both rough-road durability tests (drive tests) and bench durability tests. In a bench durability test, the displacement history of a hydraulic shaker is first determined using, as a target signal, the frame acceleration rates obtained through experimental rough-road driving, and then oscillation inputs equivalent to those experienced by an actual vehicle are applied by the shaker using the displacement history thus determined. In actual bench tests, however, the history is usually amplified in order to shorten the test time. A typical test rigging is shown in Fig. 1, in which the cab is fitted on the frame, the suspension and unsprung components including tires are removed, and the end of each actuator on the shaker is attached to the corresponding spring hanger. The shaker generates vertical vibrations along four axes. Although the shaker cannot generate horizontal inputs, its functions are considered sufficient for evaluating the cabs fatigue strength since the vertical vibration can be regarded as dominant among the effects the inputs to the cab have on the fatigue strength. 2.2 CAE simulation approach Fig. 2 shows the flow of the CAE method used in this research for predicting the cab fatigue strength. The following two items are not supported by the conventional analysis but can be supported by this CAE method that is capable of predicting the fatigue strength taking the history of composite stress inputs through cab mounts into account. (1) Quantitative history of cab input loads (2) Fatigue life calculation based on stress history The CAE method combines the solvers and predic** Function Testing Dept., Research & Dev. Office, MFTBC ** ** Material Engin. Dept., Research & Dev. Office, MFTBC

Engin. Administration Dept., Research & Dev. Office, MFTBC

** Cab Design Dept., Research & Dev. Office, MFTBC *

52

Fatigue Strength Prediction of Truck Cab by CAE

Fig. 3

Static characteristics of cab mounts

Fig. 1

Bench test conditions

Fig. 2

Flowchart of fatigue strength simulation method

tion schemes offered by multiple applications, including static analysis and eigenvalue analysis by NASTRAN, multi-body simulation by ADAMS, and fatigue life analysis by FALANCS. The calculation process is as follows: First, a finite element model of the cab is created using CAD data and then an eigenvalue analysis is performed on that model using NASTRAN. The results of the eigenvalue analysis are then converted into an ADAMSs modal model using ADAMS/FLEX. At the same time, the mounts and related parts are modeled using actual measurement data. The models thus created are combined and ADAMSs multi-body simulation is performed on them to derive the cab input load history. This history is then combined with the results of NASTRANs static analysis to obtain a stress history, from which the fatigue life of each component is calculated. This prediction method was developed as part of the development program for a full model-change of the Mitsubishi CANTER light-duty trucks, which was completed in June 2002.

Fig. 4

Dynamic characteristics of cab mounts

3. Analysis of inputs to cab

3.1 Modeling method 3.1.1 Cab mount modeling The cab is fitted on the frame via cab mounts. The cab mounts of the trucks under development were of a hydraulic damping type filled with a fluid, which features high damping force and low spring constant. Fig. 3 shows the static characteristics and Fig. 4 shows the dynamic characteristics of the hydraulic damping mounts, both for the vertical movements of the mounts. The central portion of the curve in Fig. 3 corresponds to the characteristic of the fluid-filled section of each mount. When the amount of input is above or below the limits, the mount comes in contact with a stopper and the spring constant of the mount suddenly increases. Since the amplitude of inputs is magnified in bench durability tests in order to shorten the test time, the spring constant values after making contact with the stopper are significant. The dynamic characteristics of the cab mount are expressed by a dynamic spring con-

53

Fatigue Strength Prediction of Truck Cab by CAE

Fig. 5

Hydraulic damping mount modeling

Fig. 7

Comparison of frame acceleration between experiment and simulation

Fig. 6

Finite element model and constraint points

Fig. 6 shows the finite element models of the cab and frame and the constraint points on them. The constraint points on the cab correspond to the four cab mount locations; those of the frame are a total of eight locations consisting of the four mount locations and the four actuator attaching locations. The number of modes taken into account is 30 (including 24 for the constrained mode) for the cab and 54 (including 48 for the constrained mode) for the frame. The input analysis model is configured by adding to these cab and frame models the models that represent the shakers actuators and mounts. 3.2 Verification of simulation results The simulation was performed by applying the history determined through experiments to the actuators of the shaker, and the results of the simulation were verified in terms of the cab mount displacement and the acceleration rates of several parts through comparison with the experiment results. Fig. 7 compares the calculated and experimentally measured frame acceleration rates. The simulation results are approximately consistent with the experiment results. The small disagreements may be primarily due to inadequate representation of the elasticity characteristics of the connections between the frame and shaker. To improve the accuracy of the results, therefore, incorporating the elasticity characteristics of the connections is essential. Fig. 8 shows the simulated cab mount displacement compared with the experimental measurements. The figure indicates that the simulation outputs are sufficiently accurate for practical use, as the amplitudes are almost equivalent to the experimental measurements although the history includes some discrepancies.

stant and loss factor for each vibration frequency. Fig. 5 is a model of the hydraulic damping mount represented by a combination of springs and a dashpot(2). The static characteristics can be expressed by directly substituting an actual measurement value for K1 and the dynamic characteristics can be expressed by adjusting the values of K2 and C. As shown in Fig. 4, the calculated dynamic constant roughly agrees with the actual measurements, while the calculated and measured loss factor curves also agree in terms of the resonance frequency and peak values although they are somewhat different in the high vibration frequency range. The model is therefore also capable of expressing the dynamic characteristics. 3.1.2 Input analysis model Rigid structure models have been the most commonly used type of model for motion analyses. However, as the trucks cab and frame are elastic structural bodies, elasticity characteristics should be considered when an accurate analysis is expected. The development team therefore used a modal model based on the constrained mode method as the structural model.

54

Fatigue Strength Prediction of Truck Cab by CAE

Fig. 9

Procedure of fatigue life analysis

Fig. 8

Comparison of cab mount displacement between experiment and simulation

4. Fatigue life prediction

4.1 Calculation method
Fig. 10 Spot-weld modeling

Fatigue life prediction incorporating stress history Stresses that occur in various areas of the cab are combinations of the following two types of stress: 1 Static stress, the generation of which depends on the balance with input loads 2 Dynamic stress, the generation of which depends on the vibration characteristics of the cab The following methods can support the stress history in calculation of the fatigue life and applicable to the above stress types, respectively. 1 Linear superposition method based on static analysis stress and input loads 2 Modal superposition method based on the proper mode analysis and modal displacement Since the experimental data collected so far show that static inputs such as torsional forces and vertical bending forces are dominant among the inputs to the truck cab, only method 1 was adopted in the research. This method is based on the following principle: the stress condition at a given time can be determined by the linear superposition of stresses that result from the input conditions, while the history of the stress condition is determined by multiplying together the stress resulting from a unit load and the input load history. The process of predicting the fatigue life is as follows (see also Fig. 9): (1) The value of the stress resulting from a unit load applied to each of the four cab mounts is calculated. Since shaking of the cab is a kind of constraint-free loading, NASTRANs inertia relief analysis program

4.1.1

is used for the calculation. (2) The stress value derived from (1) above is multiplied by the history of cab input load through each mount (calculated from multi-body simulation), and the stress value and load are superposed for all the input cases to create a time history of the stress. (3) Damage is calculated using rainflow accounting from the stress-time history created in (2) above and the fatigue life is calculated using the corrected Miners rule. 4.1.2 Calculation of stress in spot welds In general, a truck cab is formed of a number of sheets of metal attached by spot welding. Since loads acting on the cab are transmitted through these welds to various parts, stress concentrates on the welds and so they are often the most vulnerable points in terms of strength. FALANCS supports the LBF approach which was proposed by Rupp et al.(3) as a method of analyzing the fatigue life of spot welds. This approach expresses a spot weld as two sheets of shell elements and a beam element (including rigid elements and multi-point constraints (MPCs)) and calculates stresses from element forces (axial force, moments, etc.) to derive the fatigue life. Fig. 10 shows how a spot weld is modeled. The model is configured by two parallel shell elements and a beam element that joins the shell elements. As to the distance between the shell elements and the angle of

55

Fatigue Strength Prediction of Truck Cab by CAE

Fig. 13

Crack initiation of spot-weld

Fig. 11

Comparison of fatigue life at body panel between experiment and simulation

4.2.1 Non-welded panel sections Fig. 11 compares the calculated and experimentally measured fatigue lives. In the calculation, the data length corresponding to one history iteration is regarded as one vibration cycle. The fatigue-life calculation based on the S N analysis predicted all the cracks that also occurred in the physical experiment, but the calculated fatigue life was much shorter than the experimental result. This discrepancy may be primarily because the calculation did not support the plastic stress (strain). In fact, the fatigue life calculation that was performed based on N analysis using an N curve obtained by a simplified method in which the materials tensile strength and Youngs modulus were used(4), showed closer agreement between the calculation and physical experiment. This implies that data on basic material properties must be accumulated in the future for accurate prediction of fatigue life by CAE simulation. 4.2.2 Spot welds The fatigue life calculation results were compared with the results of the experiment conducted for spot welds on the floor panel. The damage distribution calculated by the fatigue life simulation is shown in Fig. 12. The circled points in the figure correspond to the locations where cracks occurred in the experiment. The calculation result indicates major damage in many of these locations. The points indicated by red circles correspond to the locations in which cracks were detected visually, while those indicated by yellow circles correspond to the locations where cracks were found by cutting spot welds after the experiment. Fig. 13 is a photograph showing the cross section of a spot weld. The crack in the weld has not yet progressed to the surface, so it cannot be detected visually. There were several locations where cracks actually occurred in the experiment but calculated damage was rather slight. One possible cause for this discrepancy is that the calculation did not support changes in the stress distribution caused by the occurrence of cracks in certain welds, and this might have prevented the simulation from predicting the high stress areas that would have consequentially occurred. Another cause may be the use of an fixed value as the coefficient when calculating the stress from the separation load and bending moment although this method requires the coefficient be varied according to the shape of the part around the weld. Other factors contributing of the discrepancy may include the effects of variation in diameter among spot welds and the initial stresses resulting from welding

Fig. 12

Damage distribution of spot-welds

the beam element (both are important factors that significantly influence the element forces), the model assumes that the shell element is at the mid-plane of each shell and the beam element is perpendicular to the shell elements. 4.2 Verification of simulation results The fatigue life simulation results were verified by comparing them with the experimental measurements on the prototype that was fabricated in the initial stage of development of the vehicle. Since vibration inputs within the evaluation standard range did not cause cracks in a sufficient number of locations for a significant comparison, a marginal durability test was conducted by applying inputs of magnitudes beyond the standard range and a number of cracks sufficient for the comparison were produced. The evaluation was performed separately on non-welded panel sections and spot welds.

56

Fatigue Strength Prediction of Truck Cab by CAE

that were not supported by this method that calculates the stress from the beam element forces. Part A in Fig. 12 is a spot weld for which the simulation predicted major damage but the physical test produced no crack. This weld spot is liable to be affected by contact. The linear-analysis-based simulation, which cannot support contact, may have created an impractically large damage output. Many issues remain to be solved before the method can be made sufficiently accurate for quantitatively predicting fatigue life, nevertheless the present stage of the simulation can predict crack locations with reasonable accuracy.

Fig. 14

Roof deck installation

5. Application of the simulation method

The companys practice of bench durability tests for truck cabs includes, as one of the standard evaluation items, the fatigue strength under the load of a roof deck installed on the cab (including the mass of cargo on it). A cab with roof deck installed is shown in Fig. 14. Physical durability tests revealed presence of locations where cracks did not occur without a roof deck but did occur when the deck was installed. A simulation was therefore attempted using the method presented here to analyze factors contributing to the decrease in fatigue strength due to this change in condition.

Fig. 15

Comparison of front mount force between roof deck installed cab and uninstalled cab

5.1 Effects of roof deck on Fig.16 Comparison of component force frequency characteristic between input loads roof deck installed cab and uninstalled cab The team first studied the difference in input loads on the cab between the two conditions, with and shown in Fig. 16. The graph indicates that only rolling without the roof deck. Fig. 15 shows the results of anainputs have a peak frequency of 3 Hz when the cab is lyzing the input load history and the frequency response installed with the roof deck although there are no peak at one of the front cab mounts under the two conditions. frequencies without the roof deck. There are no other The load history graph shows that when the cab is component forces that have peak frequencies around 3 installed with the roof deck, the vibration amplitude Hz. This means that installing the roof deck significantly increases rolling inputs. increases by approximately 50 % as compared with the case without the roof deck and the frequency response 5.2 Effects of roof deck on fatigue life graph shows that resonance occurs at around 3 Hz frequency. Next, the team studied the effects the change in input load discussed above might have on the fatigue The cab is supported by mounts at four locations, life. The distribution of damage in a location on the cab and so it should be appropriate to break down the vertiwas compared between when the location was under cal input loads on the cab into four component forces the roof deck load and when it was free from the load. bouncing, pitching, rolling, and twisting(5) in order to study the causes of the above-mentioned findings. The The results showed that when the deck was installed, frequency characteristics of each component force are the magnitude of the largest damage was 20 times that

57

Fatigue Strength Prediction of Truck Cab by CAE

Fig. 17

Comparison of damage distribution between roof deck installed cab and uninstalled cab

(strain) history allowed the locations of cracks to be predicted both in nonwelded panel sections and spot welds, although the prediction was not sufficiently accurate in quantitative terms for spot welds. (3) The study using the CAE simulation method on the effects caused by installing on the cab, a roof deck on fatigue life revealed the mechanism of changes in the input mode and changes in the crack life that resulted from the difference in the test condition. In future studies, the team will seek to improve the accuracy of the fatigue life simulation, with the ultimate objective of building a CAE simulation system that can predict the fatigue strength under rough-road durability test conditions.
References (1) K. Koibuchi: Fatigue Life Prediction and its Software, speech collection for JSAE symposium New Proposals for Future Fatigue Endurance Evaluation of Vehicle Bodies, JSAE, 2000 T. Sakamoto: Development of Engine Rubber Mount with Hydrauic Damping, Journal of JSAE, Vol. 36, No. 12, 1982 A. Rupp, K. Storzel and V. Grubisic: Computer Aided Dimensioning of Spot-Welded Automotive Structures, SAE Technical Paper 95071, 1995 LMS FALANCS Theory Manual Version 2.9 Y. Kondo, H. Ono and K. Hamano: Study of Load Applied to Vehicle Body during Rough-Road Driving, Mitsubishi Motors Technical Review, NO. 1, 1988

Fig. 18

Damage distribution by each component force (2)

of the cab with no deck installed (Fig. 17). In order to investigate the contribution of the above-mentioned component forces to the damage at that location, a fatigue life simulation was conducted for each of the input component forces. The results are shown in Fig. 18. The calculation result indicates that bouncing input has little contribution but rolling input has a significant contribution to the damage. In other words, the fatigue strength of the location is largely influenced by rolling inputs. The results of studies on the causes of cracks in a cab on which a roof deck is installed are summarized as follows: (1) Installing the roof deck significantly increases rolling inputs. (2) Rolling inputs contribute significantly to the fatigue strength of the investigated location. These two facts may explain the causes of cracks.

(3)

(4) (5)

6. Summary
This paper discussed a fatigue strength prediction method that was developed to simulate the bench durability test and examined the cases of its application. The results yielded through the research are summarized below. (1) Multi-body simulation using an elastic body model enabled cab input loads to be calculated with sufficient accuracy for practical application. (2) The fatigue life simulation that supported stress

Shinichi CHIBA

Kimihiko AOYAMA

Kenji YANABU

Hideo TACHIBANA

Katsushi MATSUDA

Masashi UCHIKURA

58