Research On The Recognition Method Based On SVM To The Pipeline Magnetic Flux Leakage Signals

NDT pipe on the protection of industrial production has a very important sense.Widely used pipeline inspection technology at home and abroad mainly include ultrasonic testing,eddy current testing and magnetic flux leakage detection and so on.Nowadays,scholars both at home and abroad are all focused on defects of pipeline in the technology research magnetic flux leakage testing,but only a few studies on the impact of Pipeline components,such as pipeline repair fill plates,flanges,branch pipe,and weld.Because of the structural characteristics of these channels,magnetized pipeline will produce magnetic flux leakage,if handled carelessly,we get the leakage magnetic field generated by these components confuse with defect leakage field,it would result in miscarriage of justice or missing,reducing the detection accuracy,and pipeline inspection and maintenance will be Inconvenient.In this paper,in order to study the pipeline defects and the distinction approach between the various components.we mainly study the magnetic flux leakage testing,to analyze the changes of leakage magnetic field in the pipeline under different circumstances.Based on a comprehensive analysis of magnetic flux leakage technology research status and trends at home and abroad,as well as features、advantages and technical limitations of multi-sensor information fusion technology,support vector machine technology and other new technologies,according to the characteristics of magnetic flux leakage testing signal,analyzing technology applicability,we study the way、the algorithm and the Intelligent system of magnetic flux leakage testing signal recognition.Firstly,according to the basic principles of magnetic flux leakage testing,we use finite element analysis and experiment to analyze the correlation between characteristics of pipeline components magnetic flux leakage signal as well as shape defects and changes in magnetic flux leakage signals.They are the basis for establishing signal segmentation and recognition methods.The actual detection of the MFL signal is very complex,so it should be analyzed.We can use signal amplification,noise reduction techniques to filter out the baseline,noise,etc.In this paper,image segmentation technique is used to segment by a number of sensors for magnetic flux leakage signals,it can be more effective to integrate multi-sensor information,to obtain three-dimensional object for investigation,so to help obtain a more comprehensive evaluation of the defects and pipeline components leakage magnetic field.Generally,the obtained magnetic flux leakage testing signals sample size is little,however,because of the particular specifications of the pipeline and its components are fixed,certain values can be obtained by the test in advance,such as the peak range of pipeline components magnetic flux leakage signal,the magnetic flux leakage signal reference value of different size compensation boards and defects and so on,so as to maximize the advantages of support vectors.In this paper,we use C-SVM based on kernel to distinguish piping components and defects,to estimate the size of defects and fill plates,to infer the state of pipeline current position through the interpretation of the signal.Finally,we establish a magnetic flux leakage testing signal adaptive recognition system,then use sample database and knowledge base combined approach,giving full play to the advantages of human-computer interaction,to do MFL Signal object segmentation,feature extraction and classification.Through practice proved,this method has high analysis efficiency and precision.All this provide a basis for the problem of automated and intelligent magnetic flux leakage testing signal recognition.