Research On Failure Type And Refinement Algorithm Of The Typical Fault Detection In Pipeline Being Used

As one of the five major transportations, the industrial pipeline is the main research object in this paper. In the paper we emphatically discussed several frequently occurred failure types in pipeline being used, then chose the methods of Nondestructive Examination for several typical fault, and finally achieved the pattern recognition of defects.In the paper, we put the five kinds of common NDE as a breakthrough point. For the detection must be done in the process of running, so after the theory discussion and experimental demonstration and combined with the complex working conditions, we concluded that the suitable NDE methods for the pipeline are X-ray testing and Ultrasonic testing.May occur in the pipeline failure can be divided into four categories:corrosion defect, crack defect, weld defect and damage defect by the third party. By analysing the expert experience, the recurrent failure types are weld defect and corrosion defect. And then in order to facilitate for comparison, we experimented on two kinds of different diameter pipeline with different defect types.When detecting by the X-ray testing, we use MATLAB process image, including denoising, enhancement, edge detection and feature extraction. In image processing, after repeated attempts, the Two-order Butterworth Low-pass filter was been used in denoising, the Image Fuzzy Set Enhancement operator was been used in enhancement, and Laplace of Gaussian function was been used in edge detection. When detecting by the Ultrasonic, by using circular arc curve fitting based on Newton interpolation method, we curve fitted the Ultrasonic thickness data. The trend diagram of the wall thickness would come into our view. By analyzing the variation it could be very intuitive to get the defect information.By comparing the results of the two methods, we can concluded that in the detection of weld defect X-ray detection is very outstanding, but for the average corrosion defect detection it is not well behaved, and the Ultrasonic detection is more suitable for the detection of various corrosion defects. The Ultrasonic detection is taken as an evidence of X-ray detection. It can not only achieve a high accuracy of detection, but also save the cost.In the pattern recognition part, we chose seven features to effectively describe the defect, including Convexity-concavity, Circular-degree, Flatness-degree, Aspect-ratio, Symmetry and Sharp degree of the end. And finally according to these seven features we set different threshold. At the last, by using the Binary Tree Structure Classifier, the Pattern Recognition of the defect can be finished very successfully.