基于BP神经网络的盘形滚刀磨损预测研究

Research on the Wear Prediction of Disc Cutters Based on BP Neural Network

为防止全断面掘进机滚刀磨损超限、减少刀具检修维护的费用,需要准确地对滚刀磨损进行预测。文章依据磨粒磨损机理和滚刀受力模型,推导了滚刀径向磨损量与贯入度、安装半径、刀盘转速间的数学模型;建立了基于BP神经网络的滚刀径向磨损量预测模型,并采用遗传算法(GA)和粒子群算法(PSO)分别对预测模型进行优化;以中国西北地区某TBM掘进工程为实例对预测模型进行验证。结果表明:以掘进机推力、滚刀安装半径和刀盘转速作为神经网络输入节点能较为准确地预测滚刀径向磨损,且该神经网络预测模型有较高的预测精度。最后基于GA-BP神经网络预测模型,采用MATLAB和C#混合编程的方法设计了该预测模型的人机交互界面。

The accurate prediction of disc cutter wear is essential to prevent excessive wear of disc cutters on the full-face tunnel boring machine and to reduce costs on cutter tool repair and maintenance.Based on the abrasive wear mechanism and the disc cutter force model,this paper derives a mathematical model between the radial wear of the disc cutter and the penetration,installation radius,rotation speed of the cutter head;and it establishes a prediction model for the radial wear of the disc cutters based on the BP neural network,which is optimized by using the genetic algorithm(GA)and particle swarm optimization(PSO),respectively.The prediction model is verified by a TBM tunnelling project case in Nortwest China,and the results show that the radial wear of the disc cutters can be predicted more accurately by using the thrust of the TBM,the installation radius of the disc cutters and the rotation speed of the cutter head as the input nodes of the neural network,with higher prediction accuracy.Finally,it designs the human-computer interface for the GA-BP neural network based prediction model by using a hybrid programming method of MATLAB and C#.