数值模拟与卷积神经网络混合的堆垛机立柱性能预测方法

Prediction method of performance of stacker column mixed with numerical simulation and convolutional neural network

当立柱高度、载荷、结构尺寸等变量因素耦合影响堆垛机立柱结构性能时,数值模拟计算会变得非常复杂且工作量剧增,难以应对企业堆垛机大规模个性化敏捷设计的实际需求。文中通过数值模拟手段并结合正交试验法,对25~40 m高层堆垛机变截面立柱的力学性能进行分析;以此为基础,通过卷积神经网络算法建立了立柱性能智能化预测模型,与有限元计算结果比较平均相对误差低于10%而计算效率大幅提高,可以为高层变截面立柱个性化设计的快速评估提供新的使能工具。

When the coupling of column height,load,structural size and other variables affects the structural performance of stacker columns,the numerical simulation calculation will become very complicated and the workload will increase sharply,which makes it difficult to meet the actual needs of large-scale personalized agile design of enterprise stackers.Through numerical simulation and orthogonal test,the mechanical properties of variable cross-section columns of 25-40 m high-rise stackers were analyzed.Based on this,an intelligent prediction model of column performance was established by convolution neural network algorithm.Compared with the finite element calculation results,the average relative error is less than 10%and the calculation efficiency is greatly improved,which provides a new enabling tool for rapid evaluation of personalized design of high-rise variable cross-section columns.