Discrimination and compensation of abnormal values of magnetic flux leakage in oil pipeline based on BP neural network

摘要

In the submarine oil pipeline inspection, the magnetic flux leakage data may exist some abnormal values. In order to obtain the true value, the magnetic flux leakage data should be preprocessed. One of the important parts of data preprocessing is to discriminate the abnormal values, and predict and compensate its true value reasonably and effectively. In this paper, it combines threshold segmentation with 3s-criterion to discriminate the abnormal values of the single-channel MFL data firstly. And then, analyzing the neural network theory and the characteristics of MFL data, it presents a method of compensating the abnormal value which is based on BP neural network. Lastly, it simulates the methods of discrimination and compensation with the abnormal value which is collected from single-channel of magnetic flux leakage detection. The results show that this method is feasible and effective.