Pipeline Defects Recognition Technology Based On Neural Networks

Magnetic flux leakage technology is one of the most effective detection of pipeline defects, due to the high reliability and speed, this technology is used more and more in the pipeline defect detection. Identification of defects is an important component of the magnetic flux leakage testing system, only the correct identification can be able to provide accurate information for the owners to determine whether the pipe can be used. Along with the modern computer technology, the pipeline detection is not just limited on find the defects, even more important is to be able to carry out quantitative analysis of the defects, that is, inversion of magnetic flux leakage signals to the defects in the actual parameters.In this thesis, the finite element analysis method was used, a three-dimensional inspection model has been established, signals of Magnetic Flux Leakage has been got. Then, we use neural network algorithm, ultimately obtained the geometric parameters of the defect by inversion of magnetic flux leakage signals. The main results are as follows:1. Based on the principle, we studied the systems and processes of magnetic flux leakage testing, and analyzed the relationship between the parameters and magnetic flux leakage.2. Based on the theory of finite element, the software of finite element analysis is used to setting up of the MFL model and simulation of the MFL signals.Then the simulation signal that has the same character as the MFL signal which is really detected by the MFL detector.3. The recognition work is done for the columnar-shape defect as an example using the neural network algorithm. In this paper, 60 groups data of MFL samples acquired from the defects with different dimensions by the finite element analysis software named Ansoft .40 groups data are used as the study samples .The other 20 groups data are used for the examination of the neural network model. The results showed that the characteristics sample obtained by finite element analysis is valid, and neural network algorithm is an effective method of identification on piping defect.Finally, we make a summary of the transcript and outlook the further research of pipeline defects recognition and parameters analysis.