基于径向基函数神经网络的白车身减重优化研究

Lightweight Design of BIW Based on Radial Basis Function Neural Networks

针对白车身变量多、性能响应复杂问题,在车身结构的板厚优化设计中引入近似模型方法,提高设计效率。以某轿车车身结构为轻量化设计对象,通过灵敏度分析确定优化设计变量,基于径向基函数神经网络近似模型进行全局优化,在不降低刚度和模态性能的情况下实现白车身减重目标。通过最优拉丁超立方试验设计构造样本点,用径向基函数神经网络法构造刚度和模态的近似模型,并用自适应模拟退火法进行优化求解。结果表明,径向基函数神经网络模型能较好地模拟车身结构刚度和模态响应问题,提高了整体的设计效率。通过有限元模型的验证,基于近似模型的优化结果精度较高,实例白车身减重达5.73%。

For improving optimal design efficiency,the approximation model method is introduced into the size optimization of car body structure because of many parameters and complex performance response. Taking a certain car body structure as lightweight design object,in purpose of BIW lightweight without reducing the stiffness and modal performance,the optimal design variables are determined through the sensitivity analysis and global optimization is executed based on radial basis function neural network approximate model. The sample points are produced by Optimal Latin hypercube design,the approximation model is built by using Radial Basis Functions（RBF）neural network,and the optimal design is executed by Adaptive Simulated Annealing（ASA）. The results show that RBF neural network model can simulate the stiffness and modal response problems of car body structure,and improve the efficiency of BIW lightweight design significantly. The accuracy of the optimization result that based on the approximation model is verified by the original finite element model. 5.73% of the BIW mass is reduced.