Research On The Method Of Pipeline Defect Classification Based On SVM And ELM

Pipeline transportation which is irreplaceable has been playing an increasingly important role in the development of the national economy. However, with pipelines increasing and time passing by, pipe network leakage accident due to corrosion, natural disasters and other factors occurs frequently. Therefore, the research on the method of pipeline defect classification is of great significance. In recent years, with the wavelet analysis, fuzzy control and other advanced control theory develop, in the pipeline leakage detection and evaluation of research on magnetic technology, the integration of intelligent algorithms and control theory is an inevitable trend.In this paper, the complexity of the pipeline defect classification is proposed based on SVM and ELM. The main technical problems are solved in two aspects:defect classification and definition of magnetic flux leakage signal characteristics, methods for pipeline defect classification improvements.Firstly, the pipeline magnetic flux leakage defect classification and defect signal characteristics defined. The definition of the main pipe defect classification is divided into a large area of corrosion defects, circumferential scratches, circumferential slit, corrosion spots, needle, axial scratches, axial slit seven categories. In order to facilitate SVM and ELM modeling and simulation, seven categories defects conferred 1-7 category labels. For MFL signal characteristics to define the characteristics, the main requirement is that the definition features can represent and distinguish defects. So that we can get meaningful model using support vector machine model and extreme learning machine.Secondly, the use of SVM and ELM modeling and simulation. Using support vector machines and extreme learning machine simulation, they all need 6 attribute data is divided into training set samples and testing samples. The training set used to train the mathematical model, the test set of samples used to test the accuracy of predictions. Through the simulation conclusion:support vector machines has higher prediction accuracy, but it takes a long time, slow speed; extreme learning machine speed, but the prediction accuracy rate lower.Finally, the support vector machine with both advantages and disadvantages of extreme learning machine to improve defect classification process. Improved defect classification method according to certain rules defects divide defects into serious defects and non-serious defects, so you can use two classification methods which are more focused. As for the serious defects using extreme learning machine has the advantages of fast speed, fast classification. And then use the support vector machine to get a more accurate classification results. For non-serious defects due to less demanding of time, the direct use of SVM training simulation. The simulation analysis concluded:After the defect into serious and non-serious defects, defects classification results are more accurate and more in line with actual needs.