基于GADF-CNN的盾构盘形滚刀磨损预测模型  

Wear Prediction Model of Shield Disc Cutter Based on GADF-CNN

作　　者：杜康 朱强[1] 秦东晨[1] 陈江义[1] DU Kang;ZHU Qiang;QIN Dongchen;CHEN Jiangyi(School of Mechanical and Power Engineering,Zhengzhou University,He’nan Zhengzhou 450001,China)

机构地区：[1]郑州大学机械与动力工程学院,河南郑州450001

出　　处：《机械设计与制造》2025年第3期115-120,共6页Machinery Design & Manufacture

基　　金：郑洛新国家自主创新示范区产业集群专项课题(181200210100)。

摘　　要：为了提高盾构机掘进效率、有效预测滚刀磨损,采用将格拉姆角差场法(GADF)与改进的卷积神经网络(CNN)结合的预测方法,建立了盾构盘形滚刀磨损预测模型。以某输水工程的盾构数据为研究对象,建立了以滚刀磨损影响因素为主体的磨损数据集,利用格拉姆角差场法转换为二维图像集,以此为输入层训练、优化卷积神经网络模型,确定了最佳网络参数。通过对测试样本的预测结果对比分析,表明GADF-CNN模型的预测误差明显较小,测试样本的平均误差为2.75%,综合预测效果优于简化CSM模型,验证了GADF-CNN模型用于预测盾构盘形滚刀磨损量,具有较高精度和一定可行性。In order to improve the driving efficiency of shield machine and effectively predict the wear of cutter,a prediction model of shield disc cutter wear is established by combining the Gramain Angle difference field method(GADF)and the improved convolutional neural network(CNN).Considering the shield data for a water delivering project,the data set of disc cutter wear is established,which is mainly based on the influencing factors of disc cutter wear.The data set is converted into a two dimensional image set by applying Gramain Angle difference field method,which is used as the input layer,the convolutional neural network model is trained and optimized,and the optimal network parameters are determined.Through the comparison and analysis of the prediction results of the test samples,it is shown that the prediction deviation of GADF-CNN model is relatively smaller,the mean deviation of the test samples is 2.75%,and the comprehensive prediction effect is better than that of simplifiied CSM model.It is verified that the GADF-CNN model has high precision and certain feasibility for predicting the wear quantity of shield disc cutter.

关 键 词：盾构机 盘形滚刀磨损 格拉姆角差场 卷积神经网络 

分 类 号：TH16[机械工程—机械制造及自动化] TP18[自动化与计算机技术—控制理论与控制工程] U455.43[自动化与计算机技术—控制科学与工程]