基于大数据的水电机组轴心轨迹自动识别技术探讨

Big Data Based Automatic Recognition Technology Exploration for Hydropower Units’ Shaft Orbits Shape

针对水电机组轴心轨迹识别问题,提出了从信号采集处理到最终轴心轨迹自动识别的整套流程,并结合大数据技术与算法理论对轴心轨迹信号进行原始波形的分布式存储、特征向量的并行计算,最终用Mahout做轴心轨迹的类型识别。特征提取环节中,在"HU氏不变线矩"算法的基础上,提出增加前三阶矩的权重系数以提供识别准确度,并给出权重系数自适应的实现思路。该流程的关键步骤具有可实践性,已与实际系统、数据结合并验证,可进一步拓展,以提高自动识别与故障诊断水平。

For identifying the shaft orbits shape of hydropower generator units,a complete process from signal sample and process to the automatic identification of shaft orbits shape is proposed.By applying big data technology and algorithm theory,the original waveform files of signals would be stored and transformed in parallel to get typical features.Finally,Mahout algorithm library is called to recognize the type of the orbits shape.In the feature extraction,based on HU’s moment invariants algorithm,the coefficients of the first three moments is proposed to improve recognition accuracy,and the realization idea of adaptive weight coefficients is provided.The key steps of the process are practical and have been verified in the actual system with real wave data,which can be further expanded to improve the level of automatic identification and fault diagnosis.