Experiments And Analyses On Vibration Of Steam Turbine Rotor

Steam-turbine is one of the three main equipments in thermal power plant. The shutdown will bring about very serious economic losses to electric power producing if the failure occurs. Particularly, in all faults, the fault causing by vibration accounted for the largest percentage. Therefore, vibration can be seen as a key indication of safety assessment of equipments. When a turbine unit is operating normally, the value of its vibration and vibration of change should be low. Once the value of vibration becomes larger or it becomes unstable, there is a certain degree of faults in equipments.In recent years research of Artificial Neural Network (ANN) has made a great progress, and the application of ANN has been conducted in almost every field including fault diagnosis to machine such as steam turbine.The historic-development, current situation, purpose and significance of researching in vibration and ANN are described in the thesis, relative basic theory about faults type and features of turbine unit and the research methods of ANN involved vibration are also expatiated.There are many types of steam turbine vibrating fault that add up to tens. Depending on the reality trial condition just selecting vibration caused by rotor mass-unbalance, shaft miss-alignment, lubrication oil film oscillation, rub-impact to rotor in turbine to take a serial trial study with a rotator simulation test device, the analysis and research are presented and the related conclusion has been drawn applying vibration theory,while some graphic features are acquired。.As the second majority part of this paper, basing the frequency spectra features data acquired from the rig test, a Resilient Back Propagation (RBP) network and a Radial Basis Function (RBF) network models have been established. Simultaneously, using a practical Neural Network Toolbox in MATLAB R2007 environment, the two networks have been trained and the simulation calculations have been performed to verify the accurate and the degree of the fault models and to compare the two models. The simulation experimental results demonstrate that these two diagnosis methods for the four faults are feasible and successful, but the RBF model is better than the RBP model in stability.