Research On Quantitative Estimation Method Of Mechanical Properties Of Ferromagnetic Materials Based On Pulse Eddy Current

Ferromagnetic materials are widely used in railway,aerospace and construction,whose mechanical properties directly affect the safety of a variety of major equipment.At present,the detection of the mechanical properties of ferromagnetic materials relies on destructive testing,which cannot be used for online inspection of materials.As a non-destructive testing technology,pulsed eddy current testing technology,which owns the advantages of high detection speed and low detection cost,is especially suitable for the detection of mechanical properties of ferromagnetic materials.Therefore,this paper proposes a quantitative estimation method for mechanical properties of ferromagnetic materials based on pulse eddy current,which possesses theoretical research sense and engineering value.This paper firstly studies the related theory of pulsed eddy current technology,describes the basic principle,and analyzes the feasibility of quantitative estimation of ferromagnetic materials.Then the overall framework of pulsed eddy current testing system is designed,while the design process of each module of the system is discussed in detail,including the design of signal excitation module,detection probe,etc.Thirdly,this paper studies the feature extraction of pulsed eddy current signals thoroughly.The traditional characteristics of pulsed eddy current signals,such as differential signal peaks and spectral amplitudes,are applied to the quantitative estimation of mechanical properties,and a feature extraction method based on ensemble empirical mode decomposition and Hilbert-Huang transform is proposed to extract the marginal spectral peak and marginal spectral energy as new features.The experimental results show that the new features and traditional features have good stability and they are suitable for the quantitative estimation of mechanical properties.Moreover,the paper introduces BP neural network algorithm,and quantitatively estimates the mechanical performance parameters in both the complete training set and the incomplete training set.The results show that the algorithm has a high accuracy for the quantitative estimation of mechanical performance parameters.Finally,aiming to solve the problem that the BP neural network reaches local minimum easily,it is the genetic algorithm that is used to optimize BP neural network.According to results,BP neural network with genetic algorithms can effectively enhance the accuracy of quantitative estimation of mechanical performance parameters.