基于GAF-CNN的柴油机振动信号故障诊断  

Fault Diagnosis of Diesel Engine Vibration Signal Based on GAF-CNN

作　　者：李少康 陈龙 陈辉[1,2] 管聪 LI ShaoKang;CHEN Long;CHEN Hui;GUAN Cong(Key Laboratory of High Performance Ship Technology of Ministry of Education,Wuhan University of Technology,Wuhan 430063,China;School of Naval Architecture,Ocean and Energy Power Engineering,Wuhan University of Technology,Wuhan 430063,China;Systems Engineering Research Institute,CSSC,Beijing 100094,China)

机构地区：[1]武汉理工大学高性能船舶技术教育部重点实验室,武汉430063 [2]武汉理工大学船海与能源动力工程学院,武汉430063 [3]中国船舶集团有限公司系统工程研究院,北京100094

出　　处：《武汉理工大学学报（交通科学与工程版）》2023年第4期648-653,共6页Journal of Wuhan University of Technology（Transportation Science & Engineering）

基　　金：国家重点研发计划项目(2019YFE0104600);国家自然科学基金(51909200)。

摘　　要：文中提出一种基于格拉姆角场-卷积神经网络(GAF-CNN)的故障诊断方法.利用格拉姆角场将一维柴油机振动信号转化为二维图像,通过超参数寻优的方法确定CNN模型网络结构,通过Dropout技术和Adam优化器让模型更好更快地实现拟和,最终将二维图像导入训练好的CNN模型进行实验验证.结果表明:GAF-CNN对训练集样本和测试集样本的故障诊断率分别为100%和98%,与传统的CNN方法相比具有更高的准确率及稳定性.A fault diagnosis method based on Gram Angle Field Convolutional Neural Network(GAF-CNN)was proposed.One-dimensional diesel engine vibration signal was transformed into two-dimensional image by Gram angle field,and CNN model network structure was determined by super-parameter optimization method.Through Dropout technology and Adam optimizer,the model can be fitted better and faster,and finally the two-dimensional image was imported into the trained CNN model for experimental verification.The results show that the fault diagnosis rates of GAF-CNN for training set samples and test set samples are 100%and 98%respectively,which is more accurate and stable than the traditional CNN method.

关 键 词：柴油机 故障诊断 卷积神经网络 格拉姆角场 

分 类 号：U676.4[交通运输工程—船舶及航道工程]