Oil And Gas Pipeline Corrosion Diagnosis And Prediction Based On Big Data

There are many factors affecting the pipeline corrosion,meanwhile,some of these factors have the characteristics of ambiguity and randomness,and there are complex correlations between them.Therefore,the traditional methods of pipeline corrosion diagnosis and prediction often have inapplicability.With strong adaptive learning capabilities,strong nonlinear mapping capabilities and parallel processing capabilities,big data and artificial neural network are very suitable for solving multi-factor correlation problems that traditional methods can't solve.Therefore,the application of big data analysis and mining methods to pipeline corrosion diagnosis and prediction is an important development direction of pipeline corrosion control and life prediction.In addition,as an important part of corrosion data,the accuracy of pipeline inspection data is often constrained by internal detectors and environmental factors.Therefore,it is necessary to study the comparison of pipeline inspection data and the detection probability of the internal detector to reduce the data error and determine the performance of the internal detector,thereby improving the accuracy of pipeline corrosion diagnosis and prediction.Based on the analysis of pipeline big data,a pipeline corrosion database for data mining analysis has been built.The database uses the geographical location information as a benchmark to build a data logic model.It provides a data platform for pipeline corrosion diagnosis,prediction,data comparison and probabilistic analysis.Based on the analysis of the causes of errors in the internal detection data,seven steps of alignment of key points have been proposed according to the characteristics of the internal detection data.In the defect matching,a method of matching with three positions information and the rule of corrosion points classification are proposed,providing reference for updating the corrosion database and reducing error.According to the internal detection of excavation data and the missing detection of defect points,a POD model based on excavation data and a POD model based on defect size have been proposed.A probability model has been constructed according to aforementioned two models from two aspects of defect rate and defect size.Based on the actual data,the performance of the internal detector has been determined,and the accuracy of the internal detection data has been verified.Based on the analysis of the corrosive factors of soil,a T-S neural network model has been constructed,and the network model has been trained according to the classification criteria of corrosion factors.The test results show that the error can be controlled within a small range and the corrosiveness of soil can be diagnosed well.On the basis of grey correlation analysis,a corrosion prediction model based on BP neural network has been established.In this model,the parameters of corrosion influencing factors are taken as input,and the corrosion rate is used as output.The advantages of neural networks are fully utilized.The results of tests and predictions in the case analysis are reliable.