FAILURE STRESS PREDICTION OF DEFECTIVE PIPELINE BASED ON PSO–GPR

The traditional prediction method of failure stress of pipeline with defects has the problem of large error. Aiming at this problem, the failure stress prediction model of pipeline with defects based on PSO–GPR (particle swarm optimization–Gaussian process regression) is established by using MATLAB software. By testing the experimental data of 60 groups of pipelines with defects, it is found that both the predicted results and the measured results are within 95% confidence interval, which indicates that the mean value can be used as the predicted results. The analysis of the prediction results shows that the minimum relative error between the prediction results of Gaussian process regression and the measured results is 0.003%, the maximum relative error is 1.205%, and the average relative error is 0.319%. The scattered points based on the prediction results and the measured results all fall in the ±1.3% error zone of curve y = x, which verifies the accuracy of the Gaussian process regression prediction model and provides more accurate auxiliary judgment help for practical engineering application and pipeline maintenance.