Enhancement of nondestructive evaluation techniques for magnetic and nonmagnetic structural components (Final report for doctoral fellowship)

In this report, research works performed in the Structural Safety Engineering Group of OEC/JNC are summarized as the final report of the doctoral fellowship. The main objective of this study is for the enhancement of the nondestructive evaluation techniques for structural components of both magnetic and nonmagnetic material. Studies in three topics have been carried out aiming at the quantitative evaluation of crack with the eddy current testing and the validation of a natural magnetic field based NDE method for detecting mechanical damages in a paramagnetic material. In the first part of the study, an approach to the reconstruction of the natural crack was proposed and implemented with an idealized crack model for its validation. In the second part, the correlation of the natural magnetization and the mechanical damages in the SUS304 stainless steel was investigated by using an experimental approach. In part 3, an inverse method of the measured magnetic fields is proposed for the reconstruction of magnetic charges in the inspected material by using an optimization method and wavelet. As the first work, an approach to the reconstruction of an idealized natural crack of non-vanishing conductivity is proposed with use of signals of eddy current testing. Two numerical models are introduced at first for modeling the natural crack in order to represented it with a set of crack parameters. A method for the rapid prediction of the eddy current testing signals coming from these idealized cracks is given then by extending a knowledge based fast forward solver to the case of a non-vanishing conductivity. Based on this fast forward solver, the inverse algorithm of conjugate gradient method is updated to identify the crack parameters. Several examples are presented finally as a validation of the proposed strategy. The results show that both the two numerical models can give reasonable reconstruction results for signal of low noise. The model concerning the touch of crack surfaces with a conducting band region surrounded by the crack edge, however, is proved more efficient than the model using a conductivity distribution from the point of view of both reconstruction speed and accuracy. In the second part, an experimental study on the behaviors of the magnetization induced by mechanical damages is carried out. By introducing mechanical damages to a test-piece with a tension or/and a zero-tension fatigue testing and measuring the corresponding leakage flux signal, natural magnetization change is proved and found increasing with the mechanical damages (viz. plastic deformation or fatigue damages) though a saturation occurs when damage gets too large. From the experimental results of fatigue testing utilizing test-pieces with a central slit, it was verified that observing the natural leakage flux density (leakage flux without applying external magnetic field) is a reasonable way to identify fatigue cracks. A feature parameter (area of the ε ∼ B hysteresis curve) of the in-situ magnetic field signal measured during the fatigue testing is proposed for predicting the fatigue damages, which is found depending on the cyclic number of the applied loading. At last, residual magnetic fields of a magnetized test-piece are also measured and found depending on the applied plastic deformation in case that the plastic strain is not too small. From these experimental results, it is found that the approach detecting natural magnetization is applicable for monitoring the damage status though it may be not efficient for a scanning inspection concerning its small signal magnitude. On the other hand, the method employing permanent magnet is robust against the environment noise but possibly not valid for the ISI of a structural component with a relative low damage level. For practical application, efforts to evaluate the feasibility of the proposed method are necessary for more testing conditions especially its suitability in a practical environment. For reconstructing the magnetic charges induced by mechanical damages in a test piece of SUS304 stainless steel, inversion methods base on the optimization method and wavelet are proposed and validated by using the measuring results of the second part. The approach for solving this typical ill-posed inverse problem is selected as a way in the least square method category. Concerning the ill-poseness of the system of equations, an iteration algorithm is adopted to its solving in which the designations of initial profile and the total number of iterations are taken as the means of regularization. From examples using simulated input data, it is verified that the approach gives good reconstruction results in case of signals with a relative high S/N ratio. For improving the robustness of the proposed method, a Galerkin procedure with base functions chosen as the Daubechies' wavelet is also developed for discretizing the governing equation. By comparing the reconstruction results of the least square method and those using wavelet discretization, it is found that the wavelet used approach is more feasible in the inversion of noise polluted signals. Moreover, the selection of the maximum iteration number and the weight coefficients used in the objective function are also discussed. Reconstruction of 1-D and 2-D magnetic charges with the least square strategy and reconstruction of an 1-D problem with the wavelet used method are carried out from both simulated and measured magnetic field signals, which in turn, are used as the verification of the suitability of the proposed inversion strategy. (author)