Study On Collaborative Optimization Of Car Body Structure

In recent years, energy consumption and environmental pollution have become increasingly serious. In order to save energy and reduce emission, lightweight automobile is undoubtedly one of the effective means. Under the pressure of rising oil prices and heavy smog, it is urgently necessary to develop automotive lightweight techniques. The Lightweight design is a process that not only reduces the weight of automobiles but also keeps or even enhances the performance of all aspects of automobiles. Therefore, the lightweight design of automobiles is a comprehensive optimization problem considering multi conditions. In this paper, the collaborative optimization(CO) method is used in the lightweight design of car body structure. By taking the thickness of some components of the car body as design variables, and CO method as the optimization strategy, a comprehensive optimization of the weight, side impact safety, static stiffness and modal performance of the car is achieved.Firstly, the basic principle of the CO method is studied in detail. In order to make the optimization results more accurate and reliable, the key technologies related to the optimization process are studied, including the design of experiments method, approximation models and optimization strategies.A side impact finite element model under the rules of ECE Regulation No.95and finite element models for Body-in-White(BIW) under static bending mode, torsion mode and natural mode of vibration are established and analyzed. The results of these analyses provide data for the optimization.The integration technology and secondary development language of Nastran, LS-DYNA, HyperMesh and LS-PrePost are studied. By using the data exchanging function and the software integrating function in Isight, the automatic simulation platforms under different working conditions are built.Sensitivity analyses of the thickness of15key components under various loading conditions are carried out. By comparing the main effect of each variable in the regression model, the sensitive parts under each loading conditions are determined. Through using optimal Latin hypercube design method to sample variables’ thickness and the automatic simulation platforms in Isight, RBF approximation models of the static bending stiffness, the static torsional stiffness, the first order natural frequency and the car’s whole weight are established. As a result of the strong nonlinear characteristics in the side-impact system, this paper uses the cross validation method to minimize the mean square error so as to establish ensemble of metamodels for the intrusion and the velocity of B pillar. The results show that, the precision of the ensemble of metamodel is higher than single metamodel.The CO mathematical model for car body structure is set up and according to the mathematical model, the optimization platform is built in Isight. In order to accelerate the convergence of the optimization system, this paper presents3strategies, which is, accurately setting the relaxation factor, using the adaptive simulated annealing genetic algorithm(ASA) for system level optimization due to its capability of global search, and the downhill simplex algorithm for the subspace level optimization due to its fast searching ability, using optimal Latin hypercube design method to generate the initial value of the system level design variables to avoid an unreasonable set of the number of iterations. The verification result shows that the optimization result is accurate and reliable. Study in this paper proves that applying the collaborative optimization method to the comprehensive optimization of car body is feasible, effective and accurate.