基于卷积神经网络和传感器数据的风力发电机转子断裂故障诊断  

Fault Diagnosis of Wind Turbine Rotor Fracture Based on Convolutional Neural Network and Sensor Data

作　　者：俞勤新 杨晓峰 YU Qinxin;YANG Xiaofeng(LongYuan Power Group(Shanghai)New Energy Co.,Ltd.,Shanghai 200122,China)

机构地区：[1]龙源电力集团(上海)新能源有限公司,上海200122

出　　处：《微特电机》2024年第5期35-39,共5页Small & Special Electrical Machines

摘　　要：提出基于卷积神经网络和传感器数据的新能源风力发电机转子断裂故障诊断算法。根据转子运行状态与裂纹轴向刚度之间的关联关系,构建了发电机转子裂纹模型。对多个传感器采集的发电机转子振动数据进行融合处理后,将原始的一维数据转换为m×n二维矩阵的形式,利用多通道卷积神经网络对传感器数据信息进行分析,并将输出结果代入到发电机转子裂纹模型中,实现对转子断裂故障的诊断。在测试结果中,算法对不同数据源下的电机转子断裂故障诊断结果,误差稳定在0.03μm以内。The fault diagnosis algorithm of new energy wind turbine rotor fault based on convolutional neural network and sensor data was proposed.Based on the correlation between the rotor running state and the axial stiffness of the crack,the generator rotor crack model was established.After integrating the generator rotor vibration data collected by multiple sensors,the original one-dimensional data was converted into the form of mn two-dimensional matrix,analyzing the sensor data information by using the multi-channel convolutional neural network,and putting the output results into the generator rotor crack model to realize the diagnosis of rotor fracture fault.In the test results,the error of the design algorithm is stable to within 0.03μm for the generator rotor crack fault diagnosis results under different data sources.

关 键 词：卷积神经网络 传感器数据 发电机转子 断裂故障 发电机转子裂纹模型 二维矩阵 

分 类 号：TM307.1[电气工程—电机] TM315