基于神经网络近似模型的液压支架掩护梁轻量化设计  

Lightweight Design for Shield Beam of Hydraulic Support Based on Neural Network Approximation Model

作　　者：王邦祥[1] 陆金桂[1] WANG Bangxiang;LU Jingui(School of Mechanical Engineering,Nanjing Tech University,Nanjing,Jiangsu 211816,China)

机构地区：[1]南京工业大学机械与动力工程学院,江苏南京211816

出　　处：《矿业研究与开发》2018年第8期121-124,共4页Mining Research and Development

基　　金：国家"十二五"科技支撑计划资助项目(2013BAF02B11)

摘　　要：针对传统液压支架掩护梁轻量化设计效率低问题,将实验设计方法、近似模型和遗传算法相结合,提出了基于神经网络近似模型的掩护梁轻量化设计方法。以扭转加载工况条件下的掩护梁的最小质量为目标函数,选取4个对质量和强度影响较大的结构参数作为设计变量,建立了掩护梁轻量化设计模型,通过数值计算软件ANSYS获取每个样本的质量和最大应力,利用神经网络对样本集进行多维非线性拟合,构建神经网络掩护梁质量和应力近似模型,用近似模型代替耗时的数值计算,最终得到优化后的掩护梁质量减轻了8.87%,验证了轻量化设计方法的合理性和可行性。Aiming at the low efficiency of lightweight design of traditional hydraulic support beam, the experimental design method, approximate model and genetic algorithm were combined. And a lightweight design method of shield beam based on neural network approximation model was proposed. Taking the minimum mass of shield beam under twisting loading condition as the objective function, four structural parameters which had great influence on mass and strength were selected as design variables, and a lightweight design model of shield beam was established. The mass and the maximum stress of each sample were obtained by the numerical software ANSYS, and multidimensional nonlinear fitting of the sample sets was carried out by neural network. Then, neural network approximation model of shield beam mass and stress was constructed to replace the time-consuming numerical calculation. Finally, the mass of optimized shield beam was reduced by 8.87%, which verified the rationality and feasibility of the lightweight design method.

关 键 词：液压支架 掩护梁 神经网络 遗传算法 轻量化 

分 类 号：TD355.41[矿业工程—矿井建设]