Suspension And Steering System Design And Performance Matching For In-wheel Motor Electric Vehicle

The in-wheel motor electric vehicle (IMEV) is concerned as future vehicle with theadvantage of low energy consumption, low emission and with flexible layout design. However,the dynamic characteristics of IMEV are influenced by the changes of mass distribution andchassis structure. In this paper, a suspension and steering system oriented to IMEV is presentedaiming at improving the travel performance. This system is composed of a double pivot frontsuspension system, a twist beam rear suspension system, a rack and pinon steering system.Multi-body dynamic and finite element (FE) theories are applied in mathematical modelingof the IMEV and benchmarking vehicle. The vuiral models are validated by real road tests. Thesuspension and steering system is designed and optimized based on the vural models usingmultiple theories.The handling and stability performance of IMEV is influenced by the expended lateraldimension caused by installation of in-wheel motors. In this case, a double pivot suspension isdesigned. The suspension parameters are defined by a virtual hinge in this design. Thus abetter suspension parameter combination is obtained in a narrow design space. Motioninterference and steering ratios variation occurred in preliminary design. Analysis shows theproblem is caused by the variation of kingpin parameters during steering. Optimization foreliminating the instability of kingpin parameter is processed. Seven design variables areselected based on sensitivity analysis results. The kingpin parameters are more stable duringsteering using optimal design. Variation of kingpin inclination is reduced from4°to0.32°, thevariation of caster angle is reduced from5°to1.18°. The motion interference and variation ofsteering ratios are eliminated.Torsional stiffness of twist beam in rear suspension is a significant factor influencing thevehicle travel performance. The relationship between tortional stiffness and length of twistbeam is obtained by FE simulation and stiffness tests. Ride performance simulation is processedusing full vehicle models equipping different twist beams. The vehicle ride performance isnegatively correlated with torsional stiffness of twist beam in low and middle speed, while therelationship is positive when in high speed. The vehicle handling performance is also influencedespecially the corning characteristics. A robust design based on6method is processed for steering system. The analyticalmathematical model for steering linkage is inducted. The deterministic and robust optimizationsare processed respectively based on the mathematical model. Both of the optimal results aresatisfied with ackerman steer angle, while the result of robust design is away from failureboundaries with6.547sigma level, which shows better robustness comparing with resultobtained by deterministic design.In preliminary phase of chassis design, the ride performance and stability indices are non-ideal. Thus a performance matching design of suspension and steering system is processed.Considering the number of evaluating indices of handling performance is too many, thePrincipal Component Analysis (PCA) method is applied for reducing dimension processing.The evaluation indices are reduced from21to5after calculation. Then, a chassis parameterdesign is processed aiming at matching and optimizing the ride and handling performance usingthe indices obtained by PCA dimension reduction. The result is satisfied with requirement ofvehicle travel performance.Multiple surrogate models are applied in suspension and steering matching design.Theapplications of second and third order Response Surface Model (RSM), Kriging Model (KRG)and Radial Basis Function Neuro Network Model (RBF) in chassis performance matchingproblem are analysised and compare respectively. In case of suspension and steering matchingdesign problem, results show that each surrogate model is able to gain an optimal design.However, a surrogate model with better prediction accuracy does not certainly obtain a betterdesign result. Thus application of multiple surrogate models should be considered inengineering problems.