基于卷积神经网络的风电齿轮箱故障诊断  

Fault Diagnosis of Wind Power Gearbox Based on Convolutional Neural Network

作　　者：成钧澜 张会广 邵丽丽 李兵兵 蔡高原 魏庆海 Cheng Junlan;Zhang Huiguang;Shao Lili;Li Bingbing;Cai Gaoyuan;Wei Qinghai(Datang Zhongnan Electric Power Experimental Research Institute,Zhengzhou Henan 450000,China)

机构地区：[1]大唐中南电力试验研究院,河南郑州450000

出　　处：《现代工业经济和信息化》2025年第1期231-234,共4页Modern Industrial Economy and Informationization

摘　　要：齿轮箱是双馈风机的重要组成部分,对整台风力机设备的长期安全运行极为重要。提出了一种以行星齿轮箱振动数据为输入参数的多层卷积神经网络方法,预测准确率最高可达0.9980,优于以振动数据的时频特征值为输入参数的BP神经网络。证明了卷积神经网络用于齿轮箱故障诊断的可行性。With the proposal of national 3060 target,the proportion of wind power generation is increasing year by year.Gearbox is an important part of doubly-fed wind turbine,which is extremely important for the long-term safe operation of the whole wind turbine equipment.A multilayer convolutional neural network method with planetary gearbox vibration data as input parameters is proposed,and the prediction accuracy is up to 0.9980,which is better than the BP neural network with the time-frequency eigenvalue of vibration data as input parameters.The feasibility of convolutional neural network for gearbox fault diagnosis is proved.

关 键 词：风电 齿轮箱 卷积神经网络 故障诊断 

分 类 号：TM315[电气工程—电机] TP277[自动化与计算机技术—检测技术与自动化装置]