CAJ | 학술논문

水电机组振动故障成因与故障征兆之间呈复杂的非线性关系，传统方法难以描述。当前研究常采用模式识别方法，如支持向量机、神经网络等实现振动故障诊断。该文在现有研究基础上，引进相关向量机(relevance vector machine， RVM)对诊断过程进行改进。相比传统方法，该文所提方法在学习过程中参数设置简单，在输出结果时给出了分类的可靠性，适合实际工程应用。同时，该方法在决策过程中，能够根据训练数据分布情况，自动选取决策结构，进一步提高诊断的速度与准确性。将该文所提诊断方法用于水电机组振动故障诊断实例，取得良好效果，验证了算法的有效性。
수전궤조진동고장성인여고장정조지간정복잡적비선성관계，전통방법난이묘술。당전연구상채용모식식별방법，여지지향량궤、신경망락등실현진동고장진단。해문재현유연구기출상，인진상관향량궤(relevance vector machine， RVM)대진단과정진행개진。상비전통방법，해문소제방법재학습과정중삼수설치간단，재수출결과시급출료분류적가고성，괄합실제공정응용。동시，해방법재결책과정중，능구근거훈련수거분포정황，자동선취결책결구，진일보제고진단적속도여준학성。장해문소제진단방법용우수전궤조진동고장진단실례，취득량호효과，험증료산법적유효성。
The functions between vibrating fault symptoms and their causes for hydroelectric generating units are nonlinear, and are hard to be described by conventional approaches. One usual method for the vibrating fault diagnosis is to use the pattern recognition approaches like the support vector machine and neural networks. Following the current work, we proposed the Relevance Vector Machine (RVM) based approach to optimize the diagnostic performance. Compared with conventional approaches, the proposed approach avoids the problem of parameter setting while learning, and offers probabilistic outputs. These make RVM more suitable for real applications; Moreover, the proposed approach could automatically select the optimal decision structure according to the training sample distribution, and increase the diagnostic speed and accuracy. Finally, we applied the proposed approach to a real diagnosis of the Hydroelectric Generating Unit vibrating faults, and satisfactory results have been obtained in the experiments which have validated the effectiveness of the proposed approach.