基于孤立森林算法的风电齿轮箱轴承故障检测方法

Fault Detection Method for Wind Turbine Gearbox Bearings Based on Isolation Forest Algorithm

为精准检测齿轮箱轴承故障,实时进行有效监测以保证风电齿轮箱健康运行,提出一种基于孤立森林算法的风电齿轮箱轴承故障检测方法。首先,以齿轮箱轴承温度为故障检测模型的输出变量,采用多尺度图相关算法选择输入变量;然后,提取输入变量的均方根和包络线进行自组织映射神经网络特征融合;最后,以融合值为模型输入量,使用孤立森林算法进行异常点检测。实例验证表明,经过特征融合后模型的平均检测精度提高21.56%,平均运行时间缩短0.123 s,整体性能优于常用的BPNN,RF,SVM模型。

In order to accurately detect the faults of gearbox bearings and effectively monitor in real time to ensure the healthy operation of wind turbine gearbox,a fault detection method for wind turbine gearbox bearings is proposed based on isolation forest algorithm.Firstly,the temperature of gearbox bearings is taken as output variable of fault detection model,and the input variable is selected using a multi-scale graph correlation algorithm;then,the root mean square and envelope of input variable are extracted for feature-level fusion of self-organising mapping neural network;finally,the fused value is used as model inputs,and the isolation forest algorithm is used for anomaly detection.The example validation shows that the average detection accuracy of the model after feature fusion is improved by 21.56%,and the average running time is shortened by 0.123 s.The overall performance is better than that of commonly used BPNN,RF and SVM models.