车辆前端吸能部件多目标耐撞性遗传优化

Crashworthiness Optimization for Energy Absorbing Parts in the Front of Vehicle Using Multi-Objective Genetic Algorithm

车辆耐撞性是一个涉及多因素的强非线性问题,而传统响应面模型柔韧性又不足。为克服传统响应面法在拟合车辆耐撞性问题输入与输出之间的映射关系所带来的不精确性和耗时性,这里通过建立一个经试验验证准确性的耐撞性仿真模型,提出采用均匀设计法来提高车辆耐撞性仿真试验的代表性并减少仿真试验次数。通过采用一个具有良好韧性的三层神经网络来拟合车辆耐撞性问题输入与输出之间的映射关系,并基于多目标遗传优化算法NSGA-Ⅱ,得到了一组以峰值加速度、B柱最大位移以及吸能比为耐撞性指标的Pareto解集。优化结果显示,所提方法能够高效、精确的完成车辆耐撞性优化。

For Crashworthiness optimization instinctively involves many non-linear factors while traditional response surface method tends to be less smooth,in order to overcome the inaccuracy and time consuming of traditional response surface method in crashworthiness optimization of vehicle,a crashworthiness simulation model which has been validated by experimental is employed to study the crashworthiness.Uniform design method(UDM)is adopted to enhance the quality of the selected training datasets as well as the efficiency.Then an ANN model with three layers is used to approximate the relationship between inputs and outputs due to its flexibility.Combined with multi-objective genetic algorithm NSGA-Ⅱ,a group of Pareto solution is obtained which use peak acceleration,maximum displacement of B pillar and specific energy as crashworthiness indexes.Optimization results illustrate the successful use of the proposed method in crashworthiness of vehicle,and prove that the proposed method is both of high accuracy and efficiency.