基于改进神经网络的冶金液压系统故障诊断方法  

Fault Diagnosis Method of Metallurgical Hydraulic System Based on Improved Neural Network

作　　者：胡守英 HU Shouying(Tianjin Rockcheck Steel Group Co.,Ltd.,Tianjin 300352)

机构地区：[1]天津荣程联合钢铁集团有限公司,天津300352

出　　处：《现代制造技术与装备》2024年第12期24-26,共3页Modern Manufacturing Technology and Equipment

摘　　要：提出一种基于改进神经网络的冶金液压系统故障诊断方法。该方法结合小波包变换进行特征提取,利用遗传算法优化网络参数,并引入注意力机制提高诊断精度。在1000 t液压机多故障模式试验中,该方法的诊断准确率高达98.75%,平均诊断时间为0.85 s,均优于传统的反向传播(Back Propagation,BP)神经网络和支持向量机(Support Vector Machine,SVM)方法。A fault diagnosis method of metallurgical hydraulic system based on improved neural network is presented.The method combines wavelet packet transform to extract features,uses genetic algorithm to optimize network parameters,and introduces attention mechanism to improve diagnostic accuracy.In the multi-fault mode test of 1000 t hydraulic press,the diagnostic accuracy of the proposed method is up to 98.75%,and the average diagnostic time is 0.85 s,which is better than the traditional Back Propagation(BP)neural network and Support Vector Machine(SVM)method.

关 键 词：冶金液压系统 故障诊断 改进神经网络 

分 类 号：TF30[冶金工程—冶金机械及自动化] TH137[机械工程—机械制造及自动化] TP183[自动化与计算机技术—控制理论与控制工程]