1,721,144 research outputs found
Application of Probabilistic Neural Networks to Eddy Current Non Destructive Test Problems
No full textThe aim of this paper is to propose the use of elliptical basis function probabilistic neural networks in automatic defects classification. The aim of classification is to assign a new pattern to a class, on the basis of already classified patterns. In eddy current based inspection, the patterns are sequences of complex voltages whose shape depends on the kind of defect affecting the conductive object under test. 1
About the role of hysteresis in magnetic penetration at extremely low frequency
No full textIn this paper a discussion about the role of magnetic hysteresis on the magnetic field distortion and attenuation is presented. A one-dimensional FDITD algorithm has been used to evaluate the weight of hysteresis in determining the magnetic field attenuation for a magnetic iron hollow indefinite thin cylinder. On the basis of such an analysis a second 2D-FEM numerical algorithm has been implemented taking into account only the virgin magnetization curve, therefore, not describing the hysteretic cycles and simulations have been done for a finite thin hollow cylinder geometry. To investigate the approximation due to such a strategy an experimental setup has been implemented and results have been compared. Data have shown that for the investigated geometry and materials the error remains small validating the simplified methodology at least from an engineering point of view
Image Reconstruction of Defects in Metallic Plates Using a Multi-Frequency Detector System and a Discrete Geometric Approach
No full textWe present an inversion procedure for the image reconstruction of defects in metallic plates, using a multifrequency eddy-current system. The solution of the eddy-current forward problem is achieved by means of a discrete geometric approach, while the inverse problem is resolved with an iterative linearization algorithm based on sensitivity data. In particular, we propose a suitable measurement point on the region under test using a probe coil exited by means a multifrequency signal, in order to improve the amount of usable data and the accuracy of the inverse procedure
Neural blind separation for electromagnetic source localization and assessment
No full textIn this paper we present a possible approach to electromagnetic source localization using a hybrid blind separation - minimum search inversion algorithm. The total electrical field versus time emitted by the working antennas, located at different and unknown geographical positions, is used to reconstruct each separate contribute via a suitable neural network technique. When the emitted electric field and the related base frequencies have been separated for each emitting antenna, the unknown location of each emitter, is determined with a minimum-search numerical technique. The theory here presented has been applied with success to a practical problem dealing with amplitude-modulated ratio-transmissions
Numerical Modeling for the Localization and the Assessment of Electromagnetic Field Sources
No full text- …
