基于神经网络和遗传算法的火炮结构动力学优化

为研究炮口扰动优化问题,提出采用非线性有限元、试验设计、神经网络和遗传算法相结合的方法进行火炮结构动力学优化。建立了某大口径火炮上装部分非线性有限元动力学模型,结合试验设计进行了火炮结构动力学分析。以试验数据为训练样本,建立了基于贝叶斯正则化算法的反向传播(BP)神经网络来模拟火炮总体结构参数与炮口扰动之间的非线性映射关系。构造了炮口扰动优化目标函数,利用遗传算法对目标函数进行求解,实现了火炮总体结构参数的动力学优化。研究表明所建立的火炮总体结构参数与炮口扰动之间的非线性映射关系具有很高的可信度,运用该方法进行火炮结构动力学优化行之有效。

Structural Dynamics Optimization of Gun Based on Neural Networks and Genetic Algorithms

In order to study the optimization of muzzle disturbance, a new method of gun structural dynamics optimization based on nonlinear finite element method, experimental design, neural networks and genetic algorithms is proposed. A dynamic model of a large caliber gun is established based on the nonlinear finite element method, and the structural dynamics analysis of the gun is made based on experimental design. With experimental data as training samples, a back-propagation (BP) neural network is established to simulate the nonlinear mapping between the structural parameters and muzzle disturbance index based on Bayesian regularization algorithm. The optimal objective function of muzzle disturbance is constructed, the genetic algorithms is applied to solve the objective function, and the optimal design for structural parameters of the gun is realized. The results show that nonlinear relationship between the structural parameters and muzzle disturbance index established by the method is proved to be highly reliable, and the method is accurate and feasible to optimize the muzzle disturbance.