The Research On Residual Strength Of Pipes Containing Corrosion Development Defects Based On Machine Learning

The pressure pipeline is an important infrastructure for national economic construction.And the corrosion defect is an important factor leading to the failure of pressure pipelines.It is particularly important to accurately predict the residual strength of pipelines containing corrosion defects to ensure the safe and stable operation of pressure pipelines.However,the traditional evaluation methods to predict pipeline defects are complex.And the prediction accuracy is low.Although a large number of studies have been carried out on pipeline defects using finite element methods at home and abroad,the development trend of pipeline defects and the residual strength of pipeline after the development of defects have not been thoroughly studied.In this paper,two prediction methods for pipelines external corrosion rate and pipelines residual strength are established based on BP neural network（BPNN）and optimization algorithms.The prediction of intensity change trends provides theoretical and technical support for pipeline operation scheduling,inspection,and maintenance.The specific research contents and main results are as follows:（1）An improved firefly optimization algorithm（IFA）is proposed to solve the problem of BPNN falling into local optimality due to improper setting of initial weights and thresholds.The initial weights and thresholds of BPNN are optimized by this method.The FA-BPNN,PSO-BPNN,and GA-BPNN model is established to compared with IFA-BPNN.Those three models are optimized based on the Firefly Algorithm（FA）,Particle Swarm Optimization Algorithm（PSO）,and Genetic Algorithm（GA）.（2）Based on the blasting test data of defective pipelines,a failure pressure assessment model for defective pipelines is established.It is based on the IFA-BPNN,FA-BPNN,PSO-BPNN,GA-BPNN,and the BPNN.When using the IFA algorithm to optimize BPNN,it reaches the optimal fitness of 0.3913.At the same time,the BPNN is iterated to 35 generations which is the least.And the fitness value is the best.Compared with the other four models established in this paper,the prediction accuracy of the IFA-BPNN is the highest.The maximum relative error of the IFA-BPNN is 10.12%,the minimum is only 0.01%,the average relative error is 1.91%,the coefficient of determination R²is 0.9939.（3）To evaluate the development trend of external pipeline corrosion defects,an external pipeline corrosion rate evaluation model based on IFA-BPNN,FA-BPNN,PSO-BPNN,GA-BPNN,BPNN is established according to the external pipeline corrosion rate detection data set.The IFA only needs 33 iterations to iterate to the optimal fitness value of 1.59×10^-4,which is improved compared to the other three optimization algorithms.The prediction results of the IFA-BPNN are better than the other four models.The maximum relative error of the IFA-BPNN is9.93%,the minimum relative error is 0.05%,the average relative error is 4.08%,and R²is 0.9928.The accuracy and robustness of the tool are verified to predict pipeline corrosion rate.（4）The pressure pipeline residual strength prediction software is written on the Windows platform according to the IFA-BPNN pipeline residual strength model with defects,the IFA-BPNN pipeline external corrosion rate evaluation model with the best performance in the pre-training test,and the"Recommended Practices for Corrosion Pipeline Evaluation"（SY/T10048-2016）.The software is divided into three modules:the pipeline external corrosion rate evaluation module,the pipeline failure pressure evaluation module,and the pipeline residual strength analysis module.（5）The residual strength prediction software is used to evaluate a certain defect in a natural gas pipeline in Northwest China.It is found that the average corrosion rate of a defect area of the pipeline is 0.1925mm/a,the current residual strength of the pipeline at the defect is 20.95MPa,and the actual operating pressure of the pipeline is 4-8MPa.So the pipeline can continue to operate safely.After the pipeline has been in operation for 14 years,the residual strength of the pipeline at the defect has dropped to 10.4MPa.The pipeline state monitoring needs to be strengthened for continued operation,and the pipeline operating pressure needs to be selected according to the actual situation.