神经网络近似模型在液压支架顶梁轻量化设计中的应用

摘要

For the top beam of hydraulic support to meet the requirements of working conditions,the goal of the project is to minimize the quality.A lightweight design method for the top beam was proposed combined with neural network approximation model and genetic algorithm.Firstly,using ANSYS to build a parametric model to the top beam,taking the quality of the top beam as the objective function,5 design variables that affect quality and intensity were selected,the optimization model of the top beam was established.Then,Optimal Lat-in Hypercube sampling method and ANSYS were used to get training samples.Neural network was applied to nonlinear fitting of sample set,and the approximate model of neural network was established.The approximate model was used to approximate the quality and maximal stress of the top beam,genetic algorithm was applied to solve the optimization model of the top beam,and the optimal solution was obtained finally.The optimization results showed that the quality of the top beam was 8 038.2 kg,which reduced by 9.66%.The maximum stress value was less than the yield strength of the top beam material and met the fatigue life requirement.%针对液压支架顶梁在满足工况要求的前提下质量需要达到最小的工程目标,提出了神经网络近似模型和遗传算法相结合的顶梁轻量化设计方法:首先运用ANSYS建立顶梁参数化模型,以顶梁质量为目标函数,选取5个对质量和强度影响较大的设计变量,建立了顶梁优化模型;然后用优化拉丁方采样方法和ANSYS获取训练样本,利用神经网络对样本集进行非线性拟合,建立神经网络近似模型,对顶梁质量和最大应力进行近似计算,用遗传算法求解顶梁优化模型,最终得到最优解.优化结果表明,顶梁优化后的质量为8038.2 kg,减轻了9.66%,最大应力值小于顶梁材料的屈服强度且满足疲劳寿命要求.