A novel dynamic parameter method (DPM) based on ANN for safety assessment of corroded pipelines

Corrosion in oil and gas pipelines is inevitable, which seriously affects the safety of pipelines. The safety assessment of corroded pipelines is a matter of urgency. To describe the damage behavior of pipelines, the ChenChu criterion was proposed. However, a key parameter in Chen-Chu criterion was determined empirically. In this paper, a novel dynamic parameter method (DPM) is proposed to determine the key parameter based on artificial neural network (ANN). The empirical parameter in Chen-Chu criterion is optimized by ANN and turns into dynamic one varying with input. The data for training in ANN is obtained by double circular arc (DCA) model. Compared to finite element method and previous prediction equations, the DPM are closer to the experimental data. Besides, by comparing with the experimental data, it is found that DPM is even more accurate than direct ANN method. It provides a new idea for the fusion of solution of traditional methods and data-driven methods.