Fault Diagnosis And Monitoring System For Turbine Vibration Based On Manifold Learning

Power Industry plays a pivotal role in the development of the national economy, which is closely related with people’s livelihood, and it is also the key industries in the world economy. With the rapid development of modern production and the expansion of the demand for electricity, the entire power industry will surely have an ongoing and stable development. As the main device in power industry, turbine has complex structure, and works in poor conditions. So the faults are inevitably happened. In order to improve the production efficiency, and ensure the safe operation of the turbine equipment, condition monitoring and fault diagnosis of steam turbine is becoming increasingly important.Firstly, the paper studies the theory of Turbine Vibration Monitoring and Fault Diagnosis. The development and application of manifold learning are introduced. And common fault diagnosis methods are summarized and compared. Three typical manifold learning algorithms of Laplacian Eigenmaps, Isometric Feature Mapping and Locally Linear Embedding are described in detail. Finally, Locally Linear Embedding is simpler and higher accuracy than the other two algorithms by comparing. The paper also analyzes the principles of classification of linear separable support vector machine and non-linear support vector machine. Diagnostic results are compared when the sample is extracted respectively by manifold learning method and wavelet packet analysis, and then diagnosed by support vector machine classifier. And the former has a higher diagnostic accuracy.It introduces the concept of virtual instrumentation. The paper develops a vibration monitoring and analysis system based on LabVIEW. The signal acquisition and signal-line and off-line vibration analysis are achieved.