Research And Implementation Of Pipeline Defect Detection And Recognition Method

Pipeline transportation plays an important role in the defense industry and national economy. The rapid and accurate detection of pipeline is an important part in reducing the risk of pipeline accidents, ensuring the safety of pipeline operation. Ultrasonic guided wave technology has unique advantages in the field of pipeline nondestructive testing for its high detection speed and efficiency. What’s more, it can detect the pipeline without stripping the coating layer.The identification of defects type is an important research field in ultrasonic guided wave detection technology of pipeline. The research on an effective feature optimization and pattern recognition method for guided wave signal, realizing the accurate identification of pipeline defects, is of great significance to finding the defects and preventing the pipeline accidents. In this paper, L(0,2) mode ultrasonic guided wave were excited to detect artificial prefabricated holes, cracks and pits on the basis of theoretical research on pipeline detection technology. The feature optimization, pattern recognition algorithm and the realization of the defect recognition of guided wave signals were studied. The main researches are as follows:(1) The feature extraction method combined by time and frequency domain, and the feature optimization algorithm based on PCA were put forward. On the basis of existing feature acquisition and optimization methods of guided wave signal, the application effect of digital filter, wavelet analysis and wavelet packet analysis on the noise reduction of guided wave signal were analyzed. The kurtosis coefficient, skewness coefficient, dispersion coefficient, shape coefficient and wavelet packet energy spectrum of different defect signals were also collected to constitute original feature parameters matrix. The algorithm and implementation of the Principal Component Analysis(PCA) were introduced, and the effect of chosen number of principal component factors was discussed. With the use of PCA to realize feature optimization of original feature in guided wave signal, the redundant features are eliminated and the feature dimension is reduced, which provides basis for the effective recognition of defects.(2) The defect recognition method based on Support Vector Machine(SVM) was proposed. The principle and classification process of SVM was studied, and the classifier was constructed with this algorithm. The influence of kernel function and the corresponding parameters on the performance of the classifier were analyzed. Compared with BP neural network, this method has a better recognition efficiency and generalization ability. At the same time, it avoid the dependence on the designer’s experience while using BP neural networks it is needed. Combined with the feature optimization of PCA method, the effective recognition of three types of pipeline defects, such as prefabricated holes, cracks and pits were achieved.(3) The hardware system of pipeline guided wave diagnosis instrument and software system of defect recognition were designed as well as a pipeline defect sample system, on the basis of ultrasonic guided wave theory and defect signal recognition algorithm. In addition, a pipeline with artificial defects was tested, which proves that this hardware and software system designed can achieve the defect detection of pipelines effectively.