Research On Fault Diagnosis Of Turbine Shafting Based On Neural Network

With the speedy development of product modernization, steam turbines become larger and achieve higher capacity, so inspecting the running status and fault diagnosis of steam turbine equipment also becomes more and more important in order to improve production efficiency and maintain safe operation. The traditional steam turbine fault diagnosis requires experienced staff for the complicated fault diagnosis process caused by the diversity and the non-determined factors of the shafting fault diagnosis of steam turbines, so the intelligent fault diagnosis technologies develops gradually. Artificial neural network is an mathematical mode for distributed information parallel co- processing algorithm capable of improving the accuracy of fault diagnosis with the ability of self-learning, association and memory, finding the optimal solution with high speed and dealing with no linear problem, and the application to steam turbine fault diagnosis is of great importance to economic and social.This paper starting from the fault diagnosis concept introduced, the paper has the exiting the shafting vibration fault diagnosis technologies of steam turbines analyzed and the deficiency revealed. A study of neural network shows its conception and treatment ways meet the demands for the shafting vibration fault diagnosis of steam turbines, so the study and design of the fault diagnosis system based on neural network have an important implication both in theory and practice. On basis of describing the BP neural network algorithm as well as its improvement and related knowledge, the paper has neural networks techniques introduced to the shafting vibration fault diagnosis to propose a fault diagnosis model based on neural networks, and the design idea and the principle scheme are elaborated.Some improvement methods are proposed in the paper to solving the primary problem of BP neural network as the slow convergence, difficulty in determining neuron hidden layer nodes and the existence of local minimum. A fault diagnosis model which is verified by test samples is built with Matlab, and the result shows the improved BP neural network model has great improvements in numerical precision and accuracy. The achievement of fault diagnosis using the improved BP neural network model shows the feasibility of applying artificial neural network technologies to the field of the shafting fault diagnosis of steam turbines.