Radial Basis Function Network-Based Pipeline Monitoring Signal Optimization

Considering pipeline integrity monitoring based on piezoelectric sensors and Lamb wave, a radial basis function network (RBFN) based pipeline monitoring signal optimization method is proposed. This approach compensates for the time domain sampling length using the spatial domain sampling of the pipeline's Lamb wave sensing signals. In this method, several compensation piezoelectric sensors are arranged near the original piezoelectric sensor, and the sensing signal of each sensor is synthesized in turn after a time delay. Then the RBFN is trained using the synthetic sampling time and signal. This allows the synthetic error between the time delayed signal and the trained RBFN fitting value to be obtained. After calculation of the synthetic error of each delay time, the synthetic signal corresponding to the minimum point of the synthetic error is selected as the best synthetic signal. Finally, the optimized pipeline monitoring Lamb wave signal is obtained until all the piezoelectric sensors are synthesized. The proposed method was investigated on a DN100 L360MB pipeline. Experimental results indicate that the proposed method can effectively optimize the sampling length of the pipeline monitoring signal, thus improving the accuracy of the subsequent processing and analysis of the Lamb wave sensing signal. In addition, the optimized Lamb wave sensing signal is more simi-lar to the Lamb wave excitation signal as the spatial spacing decreases when the length of the spatial domain sampling is constant.