Research And Application On Diagnosis Methods Of Rotor Thermal Bending Fault For Turbo Unit

In recent years, the catastrophic accidents caused by the fault of turbo generator units lead to significant property damage and casualties, so the fault diagnosis to the equipment, the identification of potential failure, the reasonable maintenance plan and the implement equipment state maintenance, are of great significance to ensure the safe and reliable operation of the unit.Many scholars at home and abroad made a significant contribution in the research of fault diagnosis after decades of efforts, but there are still many bottleneck problem need to break through. According to the above problem, in this paper, the steam turbine generator unit fault diagnosis technology research has been done as the following work:(1) Fault model was set up. Establishing the rotor dynamics model, making a thorough analysis of failure mechanism and studying the influence of turbine generator common fault vibration characteristics and changes of the thermal parameters on the fault. Based on the FTA and FMEA analysis, the correlation between failure mode and failure cause, location, and the correlation between the abnormal monitoring parameters and the different cause to the fault are established, providing theoretical basis for fault diagnosis.(2) Fault diagnosis knowledge base was established. On the basis of fault model analysis, extracting the fault symptoms, and proposing the fault symptom quantization method, converting the traditional text semantic quantitative qualitative description fault symptoms to the digital information which can be identified by the computer automatically. Then basing the fault diagnosis knowledge base to store fault diagnosis information according to certain rules, providing fault diagnosis knowledge base.(3) The fault diagnosis model was established. Fault diagnosis model is divided into the fault warning process and fault diagnosis process. Creep failure is the diagnosis object of the fault early warning, using the k-means clustering method to analyze fault vibration signs based on time series analysis to achieve early fault diagnosis; the objects of fault diagnosis are the result of the fault early warning and the sudden failure, using hyper sphere multi-class vector machines (HSMC-SVM) method to achieve precision diagnosis of fault modes. At the same time matching the abnormal parameters and the knowledge of the knowledge base, to solve the cause, the position and the effect of the fault, and give a reasonable maintenance plan.(4)The fault diagnosis expert system was built. On the basis of the completion the fault model, the diagnosis knowledge base and the diagnosis model, relying on the Delphi system development software, a fault diagnosis expert knowledge base which can be applied at the actual scene was set, installed and debugged on site, obtaining good results.