Research On Identification Method Of Tank Floor Crack Based On Magnetic Flux Leakage Detection

In recent years,the safety accidents in the process of industrial production have not only caused huge losses to the national economy,but also seriously threatened the safety of the people's lives.More and more attention has been paid to industrial safety production.As a common production equipment in the industrial field,the detection method of storage tank is mainly manual,which makes the detection efficiency of storage tank too low,and also threatens the life and health of relevant detection personnel in some dangerous situations.It is of great significance to detect the magnetic flux leakage of the storage tank and improve the defect detection rate in order to ensure the life and health of relevant inspectors and the safe operation of industrial production.In this paper,aiming at the cracks in the bottom plate of 5mm thick storage tank,we use the designed magnetic flux leakage detection module to collect and analyze the signals,improve the traditional defect recognition BP(back propagation)neural network,design the BP neural network recognition program based on genetic algorithm,and compare the recognition effect of the two.The final result shows that compared with the traditional BP neural network,the design is based on the legacy The improved BP neural network can effectively improve the recognition accuracy of the length and depth of the cracks on the tank floor.Firstly,this paper introduces the background significance of the subject and the research status of MFL at home and abroad.Then,the distribution of the magnetic field of the defect under the magnetic dipole model is given.In view of the complexity of the calculation of the magnetic dipole model in the actual project,the finite element analysis is introduced to improve the solution method.Through the special finite element analysis software,the two-dimensional model of MFL detection is established,and the corresponding relationship between defect parameters and each component of MFL magnetic induction intensity is determined,which lays a simulation foundation for the selection of characteristic signals.Complete the hardware structure and software design of MFL detection module.Finally,aiming at the problem that BP neural network fitting is easy to fall into local extremum,genetic algorithm is introduced into neural network defect recognition,and the defect recognition program of BP neural network is improved.Finally,through the comparative analysis of the experimental results,it is determined that the designed BP neural network based on the improved genetic algorithm can further reduce the error on the basis of the traditional defect identification algorithm,and finally control the identification error of the tank floor crack within 8%.