Corrosion and breakage detection and prediction of fracture in piping systems using artificial neural networks and wavelet transform

In this paper, a new method of diagnostics in nuclear power plant piping systems using neural network and wavelet transform will be presented. Occurring corrosion in nuclear power cycle increase the potential of breakage and leakage of radioactive materials to the environment in the future. Power cycle dynamic nature leading to subsequent corrosion after the first corrosion takes place. Effects such as irregular high frequency disturbances and noise make it difficult for the learning process of neural networks. Whatever the relationship between the input and output of neural networks is higher; the higher is the efficiency of neural network. On this basis, using time-frequency analysis and filtering properties of wavelet transform in broom the input variables, redundant signals from transient disturbances are removed from the input. Using the new variables, troubleshooting based on neural network is trained. Simulation results on experimental data related to a doctoral thesis at a university in Japan, indicating the success of the proposed method.

Keywords:

troubleshooting , Wavelet , neural network , the frequency spectrum