基于KELM和多传感器信息融合的风电齿轮箱故障诊断

Fault Diagnosis of Wind Turbine Gearbox Based on KELM and Multi-sensor Information Fusion

为提高风电齿轮箱的运行效率,降低风电场的运行维护成本,结合时域统计特征分析和多传感器信息融合技术,提出了一种基于灰狼优化核极限学习机(GWO-KELM)的风电齿轮箱状态监测新方法。首先,计算原始振动信号不同的时域统计特征参数,并采用并行叠加的方式对特征级和数据级进行信息融合以得到融合数据集。其次,利用融合数据集,建立了基于GWO-KELM的故障分类识别模型。最后,运用所提方法对QPZZ-Ⅱ旋转机械振动试验台齿轮箱实测数据进行状态监测,实例结果表明了该方法的有效性和可行性,与其他同类方法相比,所提方法具有最佳分类性能。

To improve the operation efficiency of wind turbine gearbox(WTB) and reduce the operation and maintenance costs of wind farm, a novel condition monitoring method of grey wolf optimization-based kernel extreme learning machine(GWO-KELM) is proposed, which combines time-domain statistical feature analysis and multi-sensor information fusion technology. Firstly, different time-domain indicator feature parameters of the original vibration signal are calculated, and a fusion data set from the feature level and the data level can be obtained by means of parallel superposition. Secondly, a WTB fault classification recognition model based on GWO-KELM is established using the fusion data set. Finally, combining with the measured gearbox data of QPZZ-Ⅱ rotating mechanical vibration test bench, the proposed method is adopted to realize the gearbox fault monitoring. The example results show the effectiveness and feasibility of the proposed method, and it has the best classification performance compared with other similar methods.