基于GWO-CNN的船用齿轮类机械故障诊断智能算法的实现  

Implementation of Intelligent Algorithm for Fault Diagnosis of Marine Gear Machinery Based on GWO-CNN

作　　者：吕旭峰 韩佳颖[1] Lv Xu-feng;Han Jia-ying(School of Maritime Transportation,Tianjin University of Technology,Tianjin 300384)

机构地区：[1]天津理工大学海运学院,天津300384

出　　处：《内燃机与配件》2024年第24期83-85,共3页Internal Combustion Engine & Parts

摘　　要：故障数据样本少、泛化能力和鲁棒性较差对船用齿轮诊断准确率有着重要影响,采用Solidworks软件和ANSYS软件进行齿轮建模及仿真模拟,获得特征数据建立船用齿轮诊断数据集。在灰狼算法优化卷积神经网络(GWO-CNN)模型中加入仿真数据融合后的实验诊断准确率为99.29%,解决了卷积神经网络模型鲁棒性较差的弊端。The small number of fault data samples,poor generalization ability,and robustness have a significant impact on the accuracy of marine gear diagnosis.Solidworks software and ANSYS software are used for gear modeling and simulation to obtain feature data and establish a marine gear diagnosis dataset.The experimental diagnostic accuracy of incorporating simulation data fusion into the grey wolf algorithm optimized convolutional neural network(GWO-CNN)model is 99.29%,which solves the problem of poor robustness of convolutional neural network models.

关 键 词：齿轮故障诊断 卷积神经网络 灰狼算法 仿真模拟 

分 类 号：TG457.25[金属学及工艺—焊接]