Pipeline Working State Recognition Based On Data Feature Extraction And Analysis

Since the 1970 s,the global oil and gas chemical industry has ushered in rapid development,and the quality of products and industrial output are increasing year by year.Countries around the world have laid a large number of transportation pipelines to ensure the efficient transportation of oil and gas.Therefore,how to monitor the working status of pipelines in the process of production practice and ensure the safe transportation of pipelines is the focus of enterprises and governments around the world.At present,the commonly used pipeline working condition monitoring methods such as hardware monitoring method and software monitoring method have the problems of low monitoring efficiency,difficult to get rid of manual intervention,high cost or high recognition error rate.In order to solve these problems,this thesis designs a pipeline working state recognition method based on data feature extraction and analysis,and uses time series data prediction for auxiliary research.Based on this,the corresponding software system is implemented.The main work is as follows :(1)A pipeline working state recognition method based on data feature extraction and analysis is designed and implemented.The process of this method mainly includes :data screening and cleaning,data preprocessing and feature extraction,time series anomaly detection,sequence similarity judgment and state recognition.In the process of implementing the process,local weighted regression algorithm,unsupervised anomaly detection toolkit,dynamic time warping algorithm and other technologies are used.The method proposed in this thesis has a good effect in the experiment,and can maintain a high accuracy when it has considerable efficiency.(2)The time series prediction method is applied to pipeline working state recognition.On the basis of studying the more mature time series prediction model in the past,combined with the uniqueness of this study,this thesis designs and implements a pipeline working state recognition method based on time series data prediction for auxiliary research,and applies deep learning to this study.Good results have been achieved and the feasibility and accuracy of pipeline working state recognition have been improved.(3)A pipeline working state recognition system based on the above method is realized.Based on the actual production requirements and the practice of software engineering projects,this thesis implements a fully functional pipeline working state recognition system.The system realizes the above-mentioned pipeline working state identification method,and constructs a standard data management module and a complete user-role-permission management mechanism,which enables users to conveniently monitor and accurately identify the pipeline working state.The implementation of the system improves the accuracy and efficiency of pipeline working state recognition,and provides feasibility and reference for the safe operation and maintenance of pipeline transportation industry.In summary,under the background of pipeline transportation engineering practice,this thesis studies and improves the existing technical methods in this field,designs and implements a new pipeline working state recognition method,and proves the feasibility and accuracy of the method under a large number of experimental verifications.Finally,based on the software engineering project,a pipeline working state recognition system that meets the engineering requirements is realized.