融合集合经验模态分解与宽度学习的齿轮箱故障预警方法

Early warning method of gearbox fault based on EEMD and broad learning algorithm

为实现风力发电机齿轮箱的预测性维护,针对齿轮箱油温超温故障,提出一种基于集合经验模态分解(EEMD)和宽度学习算法的融合预警模型。以健康状态下的齿轮箱数据为判别基准,首先对齿轮箱油温时序信号进行EEMD分解得到时频特性,再采用宽度学习算法,利用数据采集与监视控制系统数据对齿轮箱进行建模,分别以马氏距离和重构油温曲线与真实油温曲线的关联度为指标,从时间维度和相关变量维度评价齿轮箱的健康程度。通过计算两种算法的交叉熵将二者的预警结果融合,从而兼顾预警方法的准确性和快速性。对实际风场中齿轮箱油温超温故障发生前后记录的数据进行仿真分析,验证了EEMD变点与宽度学习算法的融合方法在齿轮箱油温超温早期故障预警上的可行性。

To achieve predictive maintenance of the gearbox of wind turbines,a fusion early warning model based on Ensemble Empirical Mode Decision(EEMD)and broad learning algorithm was proposed aiming at the over temperature fault of gearbox oil temperature.Using the gearbox data in a healthy state as a criterion,the gearbox oil temperature timing signal were decomposed by EEMD to obtain the time-frequency characteristics,and then the Supervisory Control and Data Acquisition(SCADA)data was used to model the gearbox oil temperature with broad learning algorithm.Mahalanobis distance and the correlation between the fitting result and the true value were used as indicators to evaluate the health of gearbox from the time dimension and the related variable dimension.By calculating the cross entropy of two algorithms,the early warning results of the two were fused,so as to take into account the accuracy and rapidity of the early warning method.The data recorded before and after the gearbox oil temperature overtemperature fault in the actual wind field were analyzed,and the results verified that the fusion method of EEMD change point and width learning algorithm was feasible in the early fault warning of gearbox oil temperature over temperature.