基于GA优化神经网络的变刚度复合材料板簧刚强度预测研究

Research on Stiffness and Strength Prediction of Variable Stiffness Composite Leaf Spring Based on GA-Optimized Neural Network

将形状记忆合金混编三维织物驱动器(SMA驱动器)植入复合材料板簧中,可以使大型板式结构具有变刚度潜力。为提高驱动器空间布置参数的寻优效率,文章基于BP神经网络预测不同驱动器布置参数下的板簧刚强度,并采用遗传算法提升预测精度。结果表明,经过遗传算法(Genetic Algorithm,GA)优化后的BP神经网络(Back Propagation Neural Networks)模型对板簧刚强度性能的平均预测精度为96.9%,优于传统BP神经网络的平均预测精度(92.7%),并且寻优效率比传统智能算法提升了172倍。该研究为大型板式结构的空间布置参数寻优算法提供了有益参考。

Embedding Shape Memory Alloy composite 3D fabric actuators(SMA actuators)into composite leaf spring can endow large plate-like structures with variable stiffness potential.To enhance the optimization efficiency of space layout parameters of actuators,this study uses a BP(Back Propagation)neural network to predict the leaf spring strength under different actuator layout parameters and employs Genetic Algorithm(GA)to improve prediction accuracy.Results show that the BPNN(Back Propagation Neural Network)model optimized by GA achieves an average prediction accuracy of 96.9%for leaf spring strength performance,outperforming the average prediction accuracy of traditional BPNN(92.7%),and enhancing optimization efficiency by 172 times compared to traditional intelligent algorithms.This research provides valuable insights for optimizing the space layout parameters of large plate-like structures.