基于油液在线监测的风机齿轮箱磨损状态识别

Wear state identification of wind turbine gearbox based on oil online monitoring

针对风机齿轮箱磨损状态识别中传统算法识别准确率低的问题,该文提出了一种基于BP神经网络的风机齿轮箱磨损状态的识别方法。该方法基于油液在线监测的多维数据样本,构建BP神经网络风机齿轮箱磨损状态识别模型。仿真实验中,在经过对测试数据学习训练后,BP神经网络和传统聚类算法分别对一组测试数据进行磨损状态识别,BP神经网络模型的风机齿轮箱磨损状态识别准确率可达到98%,而相同样本下,传统聚类算法的识别准确率仅为90.4%。对比实验结果表明,所提方法具有更高的识别准确率。

This research provides a method for recognizing wear status of wind turbine gearboxes based on BP neural network to address the problem of previous algorithms’low recognition accuracy in identifying wear condition of wind turbine gearboxes.This method creates the BP neural network wind turbine gearbox wear status identification model using multi⁃dimensional data samples from oil online monitoring.The BP neural network and the standard clustering algorithm recognize the wear state of a set of test data in the simulation experiment after learning and training on the test data.The BP neural network model’s wear state recognition accuracy for the wind turbine gearbox can reach 98%.The standard clustering algorithm’s recognition accuracy under the same sample is just 90.4%.The proposed method has a greater recognition accuracy,according to the findings of comparative experiments.