Interpolation of Missing Data of Magnetic Flux Leakage in Oil Pipeline Based on Improved Supporting Vector Machine

In the internal detection of submarine pipeline, the data exported from magnetic flux leakage detector may exist some missing data. In order to get accurate data, the magnetic flux leakage data should be preprocessed. The significant part of data preprocessing is to discriminate the missing data, and then to compensate these true values reasonably and effectively. In this paper, SVM algorithm is used in the process of single-channel interpolation to random missing data firstly. Secondly, the SVM traversal algorithm is proposed to achieve the interpolation of whole random missing data block. In order to improve the interpolation accuracy, an improved SVM algorithm is then introduced. The SVM traversal is carried out by using the axial data and the radial data respectively. Then based on the no missing values in the random missing data block, the least squares method is used to obtain the weight for interpolation of statistical missing data. Lastly, the real data exported from magnetic flux leakage detector is used to simulate. The interpolation results are compared with the BP neural network. The results show that this method is more feasible and effective.