基于红外成像与卷积神经网络的滚珠丝杠进给系统热误差动态补偿方法  

Dynamic Compensation Method for Thermal Error in Ball Screw Feed System Based on Infrared Imaging and Convolutional Neural Network

作　　者：吴宇航 

机构地区：[1]上海理工大学机械工程学院,上海

出　　处：《建模与仿真》2023年第3期2173-2181,共9页Modeling and Simulation

摘　　要：滚珠丝杠作为机床进给系统的直线进给传动装置,由于受到非均匀热场的影响,其热误差成为影响机床加工精度的重要因素。为了降低非均匀热场产生的影响,本文提出一种基于红外成像与卷积神经网络的滚珠丝杠进给系统热误差动态补偿方法。给出了一套基于多模型集合的滚珠丝杠分段补偿方法,建立了基于红外成像与卷积神经网络的热误差模型。结果表明,相对于无补偿、选取特征点的Back Propagation (BP)神经网络模型补偿和全图层BP (ABP)神经网络模型补偿,定位误差的均方根分别减少了90.3%、53.4%、55.3%。As a linear feed transmission device in the machine tool feed system, the ball screw is affected by the non-uniform thermal field, and its thermal error has become an important factor affecting the machining accuracy of the machine tool. In order to reduce the influence of non-uniform thermal field, this paper proposes a dynamic compensation method for thermal error of ball screw feed sys-tem based on infrared imaging and convolutional neural network. A multi-model ensemble based ball screw segment compensation method is provided, and a thermal error model based on infrared imaging and convolutional neural network is established. The results show that compared with the no compensation method, the Back Propagation (BP) neural network model compensation method with selected feature points, and the full layer BP (ABP) neural network model compensation method, the root mean square error of positioning is reduced by 90.3%, 53.4%, and 55.3%, respec-tively.

关 键 词：热误差 滚珠丝杠 卷积神经网络 机床进给系统 红外成像 神经网络模型 分段补偿 机床加工精度 

分 类 号：TG5[金属学及工艺—金属切削加工及机床]