Corrosion Detection System for Oil Pipelines Based on Multi-sensor Data Fusion by Improved Simulated Annealing Neural Network

A system to detect the corrosion of submarine oil pipeline is introduced, it got the original data by 3 groups ultrasonic sensors and flux leakage sensors. We made multiscale wavelet transform and frequency analysis to multichannels original data and extracted multi-attribute parameters from time domain and frequency domain, then we selected the key attribute parameters that have bigger correlativity with the corrosion degrees of oil pipeline among of multi-attribute parameters. The improved simulated Annealing artificial neural network was used to do multisensor data fusion to detect the corrosion degrees of submarin oil transportation pipelines and those key attribute parameters were used to as input vectors of network. The experimental results show that this method is feasible and effective.