Abstract
Background: Most NN research only conducted qualitative analysis, analyzing its accuracy, with certain limitations, without studying its NN model, error convergence process, and pressure ratio; There is relatively limited research on the application of NN optimized by GA to pipelines containing flaws; Moreover, the residual strength evaluation of GA-BP NN has the advantages of high global search ability, efficiency not limited by constant differences, and the use of probability search instead of path search, which has a wide application prospect. Objective: Using MATLAB software to estimate the residual strength of oil and gas pipelines with the GA, artificial NN BP, and GA-BP NN. Methods: This study uses MATLAB software to anticipate the residual strength of oil and gas pipelines using a genetic algorithm (GA), an artificial neural network (NN) BP, and a GA-optimized BP NN (GA-BP NN). These models' predictions were compared to widely acknowledged residual strength evaluation standards, such as ASME B31G Modified, BS7910, PCORRC, DNV RP F101 and SHELL92. Results: The GA-BP NN’s dependability and applicability were thoroughly tested, resulting in faster convergence and improved system performance following optimization. Using five different residual strength criteria, the GA-BP NN model outperformed the normal BP NN model in terms of prediction accuracy. Conclusion: These findings have crucial implications for forecasting the residual strength of corrosive oil and gas pipelines.
