Abstract:In this paper a Multi-Frequency Excitation and Spectrogram Eddy Current System and an inverse neural model were used to detect and identify natural flaws in steam generator tubes. It is shown that the applied dynamic neural model of the ECT sensor offers very high speed of operation and guarantees reliability of the recognition results.
“…Thus, these methods cannot be used for the solution of the 3D eddy current inverse problem that is considered here. The last group of techniques are based on artificial neural networks or fuzzy logic techniques [8,9], and are therefore very fast. Nevertheless, their application is rather limited to the area in parameter space, for which the model has been trained.…”
Section: Eddy Current Type Testing -Type Ndtmentioning
confidence: 99%
“…Among these techniques one can find the deterministic and stochastic algorithms, pre-calculated data approach, methods based on the evolution strategy or statistics, linear or quadratic models, artificial neural network or fuzzy-logic, e.g. [3][4][5][6][7][8][9].However, engineering optimization requires highly accurate numerical models, which imply an excessive computational cost, e.g. 3D simulations for complicated geometries.…”
Two-level approach for solving the inverse problem of defects identification in Eddy Current Testing - type NDTThis work deals with the inverse problem associated to 3D crack identification inside a conductive material using eddy current measurements. In order to accelerate the time-consuming direct optimization, the reconstruction is provided by the minimization of a last-square functional of the data-model misfit using space mapping (SM) methodology. This technique enables to shift the optimization burden from a time consuming and accurate model to the less precise but faster coarse surrogate model. In this work, the finite element method (FEM) is used as a fine model while the model based on the volume integral method (VIM) serves as a coarse model. The application of the proposed method to the shape reconstruction allows to shorten the evaluation time that is required to provide the proper parameter estimation of surface defects.
“…Thus, these methods cannot be used for the solution of the 3D eddy current inverse problem that is considered here. The last group of techniques are based on artificial neural networks or fuzzy logic techniques [8,9], and are therefore very fast. Nevertheless, their application is rather limited to the area in parameter space, for which the model has been trained.…”
Section: Eddy Current Type Testing -Type Ndtmentioning
confidence: 99%
“…Among these techniques one can find the deterministic and stochastic algorithms, pre-calculated data approach, methods based on the evolution strategy or statistics, linear or quadratic models, artificial neural network or fuzzy-logic, e.g. [3][4][5][6][7][8][9].However, engineering optimization requires highly accurate numerical models, which imply an excessive computational cost, e.g. 3D simulations for complicated geometries.…”
Two-level approach for solving the inverse problem of defects identification in Eddy Current Testing - type NDTThis work deals with the inverse problem associated to 3D crack identification inside a conductive material using eddy current measurements. In order to accelerate the time-consuming direct optimization, the reconstruction is provided by the minimization of a last-square functional of the data-model misfit using space mapping (SM) methodology. This technique enables to shift the optimization burden from a time consuming and accurate model to the less precise but faster coarse surrogate model. In this work, the finite element method (FEM) is used as a fine model while the model based on the volume integral method (VIM) serves as a coarse model. The application of the proposed method to the shape reconstruction allows to shorten the evaluation time that is required to provide the proper parameter estimation of surface defects.
“…The postprocessing of surface proper- [10], the estimation of defect shape [11], the detection and characterization of plural defects [12], the determination of crack size and shape [13], the determination of electrical conductivity profiles from the inversion of multifrequency data [14], eddy-current multifrequency system simulation and flaws identification [15], and to detect and identify natural flaws in steam generator tubes [16] are reported in the technical literatures. In this paper, the idea of using neural network is to develop an alternative method to replace the grid systems as discussed earlier.…”
Section: Neural Network Model Aided Post-processingmentioning
The use of planar-type sensors for the estimation of system properties has gained considerable importance in recent times because of its noncontact and nondestructive nature. The impedance of a coil in proximity of any conducting/nonconducting, magnetic/nonmagnetic surface is a complex function of many parameters, such as conductivity, permeability, and permittivity of near-surface materials, liftoff and coil pitch of the coil, etc. The transfer impedance (i.e., the ratio between the sensing voltage and the exciting current) of the planar-type microelectromagnetic sensors consisting of exciting and sensing coils is used for the estimation of the near-surface system properties. Two methods have been discussed for the postprocessing of output parameters from the measured impedance data. Based on the estimation of near-surface properties, it is possible to detect the existence of defects, to predict the degradation of material, fatigue, etc.
“…An alternative of the grid system is to adopt a neural network aided estimation [ 28 – 35 ]. Figure 29 shows a neural network aided model developed for the estimation of near-surface material properties.…”
Section: Applications Of Planar Electromagnetic Sensorsmentioning
High performance planar electromagnetic sensors, their modeling and a few applications have been reported in this paper. The researches employing planar type electromagnetic sensors have started quite a few years back with the initial emphasis on the inspection of defects on printed circuit board. The use of the planar type sensing system has been extended for the evaluation of near-surface material properties such as conductivity, permittivity, permeability etc and can also be used for the inspection of defects in the nearsurface of materials. Recently the sensor has been used for the inspection of quality of saxophone reeds and dairy products. The electromagnetic responses of planar interdigital sensors with pork meats have been investigated.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.