Research On Mechanical Property Estimation Methods Of Ferromagnetic Materials Based On Electromagnetic NDE

Mechanical properties are significant indicators for evaluating performance of ferromagnetic materials,so it is important to study mechanical properties.Traditionally,mechanical properties such as yield strength and tensile strength are obtained by destructive way of off-line stretching.In this thesis,the estimation methods of yield strength,tensile strength and elongation of ferromagnetic materials are designed based on electromagnetic non-destructive evaluation technology.The main content is summarized as follows:(1)The basic theory of four electromagnetic non-destructive evaluation technology,including magnetic Barkhausen noise technology,multi-frequency eddy current technology,incremental permeability technology,harmonic analysis of tangential magnetic field technology is introduced,and conventional feature extraction method of electromagnetic signals is analyzed.(2)Considering the relationship among features and between features and target parameters,the ReliefF-Corr algorithm is proposed by combining ReliefF algorithm and correlation analysis algorithm.Redundant and unimportant electromagnetic parameters are removed by ReliefF-Corr algorithm,which is based on the importance of electromagnetic parameters to mechanical properties.And the effectiveness of the algorithm is verified by experiments.(3)A mathematical model between electromagnetic parameters and mechanical properties is established based on BP neural network and ReliefF-Corr feature selection algorithm,the input vectors are respectively electromagnetic parameters of single electromagnetic method and feature fusion of multi-electromagnetic methods.The differences in modeling effects of different electromagnetic methods are analyzed and compared.(4)The theory of alternating least squares(ALS)algorithm based on matrix decomposition is studied,and ALS model for estimating mechanical properties of ferromagnetic materials based on electromagnetic NDE technology is designed,the influence of the implicit factor spatial dimension k on mechanical properties estimation is discussed.Combining the ReliefF-Corr feature selection algorithm,the ALS modeling effect between single electromagnetic method and feature fusion of multielectromagnetic methods,is compared and analyzed.The experimental results show that fusion of multi-electromagnetic methods has good generality,and it has high estimation accuracy for three mechanical properties of each steel grade.For elongation,the estimation accuracy of fusion multi-electromagnetic method is significantly higher than that of each single method;For yield strength,the estimation accuracy of fusion multi-electromagnetic method is sightly higher than that of each single method;For tensile strength,the estimation accuracy of fusion multi-electromagnetic method is not much different from each single method.