A multi-objective optimization approach for simultaneously lightweighting and maximizing functional performance of vehicle body structure

Abstract

This study presents a hybrid approach to integrate the comprehensive sensitivity analysis method, support vector machine technology, modified non-dominated sorting genetic algorithm-II method and the technique for order preference by similarity to ideal solution, which have been applied to multi-objective lightweight optimization of the B-pillar structure of an automobile. First, numerical models of the static–dynamic stiffness and the crashworthiness performance of automobile are established and validated by experimental testing. Then, the comprehensive sensitivity analysis method is used to define the final optimization variables. Experimental design and support vector machine based surrogate model techniques are introduced to establish the approximate model; subsequently, the modified non-dominated sorting genetic algorithm-II algorithm is applied to the multi-objective lightweight optimization design of the B-pillar structure, and the non-dominated solution set is determined. The principal component analysis method is applied to determine the weight of each objective. Finally, the technique for order preference by similarity to ideal solution method is used to rank Pareto front from best to worst to obtain the optimal solution; furthermore, a comparison between the original model and optimized design denotes that the mass of the B-pillar being reduced by 22.55% under the other impacting indicators is well guaranteed. Therefore, the proposed hybrid approach provided promising prospects in the lightweight and crashworthiness optimization application of the B-pillar.

A data-driven self-adaptive parameter tuning framework for composite automobile part optimization design

“…The results from the SSI-PSO with the proposed DSPT framework and compared algorithms are sorted from lowest to highest. The order is set as algorithms' score respectively as equation (19). The Best results, Worst results, Median results, Mean results, Std (standard deviation), and IoS could be found in Tables 1-4 for details.…”

“…In Figure 1, the surrogate models, which represent relationships between design variables and responses by numerical method, are adopted in the optimization process to relieve problem complexity and time consumption. There are different suitable surrogate models for different problems, such as the radial basis function for composite bumper beam optimization, 17 the hybridized radial basis function based neural network and response surface method for the front end structure optimization, 18 the support vector machine for the B-pillar structure optimization, 19 and the neural network for the automobile body optimization. 20 Different modeling methods may also result different characteristics for the same problems.…”

Electric vehicles body frame structure design method: An approach to design electric vehicle body structure based on battery arrangement

“…A total of 100 experimental samples were designed on the basis of the optimal Latin hypercube design to maximize the accuracy of the fitting of the approximate model according to references 50,51 and engineering experiments. The estimates of the precision of RSMs are listed in Table 9, which shows that G b and G t meet the requirement of the evaluation accuracy ( R 2 > 0.9 and RMSE<0.2).…”