Tarik Hacib

Microwave Characterization Using Least-Square Support Vector Machines

This paper presents the use of the least-square support vector machines (LS-SVM) technique, combined with the finite element method (FEM), to evaluate the microwave properties of dielectric materials. The LS-SVM is a statistical learning method that has good generalization capability and learning performance. The FEM is used to create the data set required to train the LS-SVM. The performance of LS-SVM model depends on a careful setting of its associated hyper-parameters. Different tuning techniques for optimizing the LS-SVM hyper-parameters are studied: cross validation (CV), genetic algorithms (GA), heuristic approach, and Bayesian regularization (BR). Results show that BR provides a good compromise between accuracy and computational cost.

Microwave characterization using ridge polynomial neural networks and least-square support vector machines

Motivated by the slow learning properties of multilayer perceptrons which utilize computationally intensive training algorithms and can get trapped in local minima, this work deals with ridge polynomial neural networks (RPNN) and least-square support vector machines (LSSVM) technique. RPNN and LSSVM are combined with the finite element method (FEM), to evaluate the dielectric materials properties. RPNN maintain fast learning properties and powerful mapping capabilities of single layer high order neural networks. LSSVM is a statistical learning method that has good generalization capability and learning performance. Experimental results show that LSSVM can achieve good accuracy and faster speed than those using conventional methods.

Eddy current characterization of small cracks using least square support vector machine

Eddy current (EC) sensors are used for non-destructive testing since they are able to probe conductive materials. Despite being a conventional technique for defect detection and localization, the main weakness of this technique is that defect characterization, of the exact determination of the shape and dimension, is still a question to be answered. In this work, we demonstrate the capability of small crack sizing using signals acquired from an EC sensor. We report our effort to develop a systematic approach to estimate the size of rectangular and thin defects (length and depth) in a conductive plate. The achieved approach by the novel combination of a finite element method (FEM) with a statistical learning method is called least square support vector machines (LS-SVM). First, we use the FEM to design the forward problem. Next, an algorithm is used to find an adaptive database. Finally, the LS-SVM is used to solve the inverse problems, creating polynomial functions able to approximate the correlation between the crack dimension and the signal picked up from the EC sensor. Several methods are used to find the parameters of the LS-SVM. In this study, the particle swarm optimization (PSO) and genetic algorithm (GA) are proposed for tuning the LS-SVM. The results of the design and the inversions were compared to both simulated and experimental data, with accuracy experimentally verified. These suggested results prove the applicability of the presented approach.

The combination of adaptive database SDM and multi-output SVM for eddy current testing

Purpose – Eddy current testing (ECT) is a nondestructive testing method for the detection of flaws that uses electromagnetic induction to find defects in conductive materials. In this method, eddy currents are generated in a conductive material by a changing magnetic field. A defect is detected when there is a disruption in the flow of the eddy current. The purpose of this paper is to develop a new noniterative inversion methodology for detecting degradation (defect characterization) such as cracking, corrosion and erosion from the measurement of the impedance variations. Design/methodology/approach – The methodology is based on multi-output support vector machines (SVM) combined with the adaptive database schema design method (SDM). The forward problem was solved numerically using finite element method (FEM), with its accuracy experimentally verified. The multi-output SVM is a statistical learning method that has good generalization capability and learning performance. FEM is used to create the adaptive ...

Ridge Polynomial Neural Network for Non-destructive Eddy Current Evaluation

Motivated by the slow learning properties of Multi-Layer Perceptrons (MLP) which utilize computationally intensive training algorithms, such as the backpropagation learning algorithm, and can get trapped in local minima, this work deals with ridge Polynomial Neural Networks (RPNN), which maintain fast learning properties and powerful mapping capabilities of single layer High Order Neural Networks (HONN). The RPNN is constructed from a number of increasing orders of Pi-Sigma units, which are used to solving inverse problems in electromagnetic Non-Destructive Evaluation (NDE). The mentioned inverse problems were solved using Artificial Neural Network (ANN) for building polynomial functions to approximate the correlation between searched parameters and field distribution over the surface. The inversion methodology combines the RPNN network and the Finite Element Method (FEM). The RPNN are used as inverse models. FEM allows the generation of the data sets required by the RPNN parameter adjustment. A data set is constituted of input (normalized impedance, frequency) and output (lift-off and conductivity) pairs. In particular, this paper investigates a method for measurement the lift-off and the electrical conductivity of conductive workpiece. The results show the applicability of RPNN to solve non-destructive eddy current problems instead of using traditional iterative inversion methods which can be very time-consuming. RPNN results clearly demonstrate that the network generate higher profit returns with fast convergence on various noisy NDE signals.

Support vector machines for measuring dielectric properties of materials

Purpose – The dielectric properties of materials (complex permittivity) can be deduced from the admittance measured at the discontinuity plane of a coaxial open‐ended probe. This implies the implementation of an inversion procedure. The purpose of this paper is to develop a new non‐iterative inversion methodology in the field of microwave characterization allowing reducing the computation cost comparatively to iterative procedures.Design/methodology/approach – The inversion methodology combines the support vector machine (SVM) technique and the finite element method (FEM). The SVM are used as inverse models. They show good approximation and generalization capabilities. FEM allows the generation of the data sets required by the SVM parameter adjustment. A data set is constituted of input (complex admittance and frequency) and output (complex permittivity) pairs.Findings – The results show the applicability of SVM to solve microwave inverse problems instead of using traditional iterative inversion methods w...

An approach based on ANFIS and input selection procedure for microwave characterization of dielectric materials

The principle of microwave characterization of dielectric materials using open-ended coaxial line probe is to link the dielectric properties of the sample under test to the measurements of the probe admittance (Y(f) = G(f)+ jB(f )). The purpose of this paper is to develop an alternative inversion tool able to predict the evolution of the complex permittivity (e = e′ – je″) on a broad band frequency (f from 1 MHz to 1.8 GHz).,The inverse problem is solved using adaptive network based fuzzy inference system (ANFIS) which needs the creation of a database for its learning. Unfortunately, train ANFIS using f, G and B as inputs has given unsatisfying results. Therefore, an inputs selection procedure is used to select the three optimal inputs from new inputs, created mathematically from original ones, using the Jang method.,Inversion results of measurements give, after training, in real time the complex permittivity of solid and liquid samples with a very good accuracy which prove the applicability of ANFIS to solve inverse problems in microwave characterization.,The originality of this paper consists on the use of ANFIS with input selection procedure based on the Jang method to solve the inverse problem where the three optimal inputs are selected from 26 new inputs created mathematically from original ones (f, G and B).

Electromagnetism-like mechanism algorithm and least square support vector machine for estimation the defect in nondestructive evaluation

Eddy Current Testing (ECT) is a fast and effective method for detecting and sizing most of the default in conducting materials. The size estimation of an unknown defect from the measurement of the impedance variations is an important technique in industrial area. This paper considers to solve this problem by the novel combination of the Least Square Support Vector Machines (LS-SVM) and Finite Element Method (FEM). The FEM is used to modelling the eddy current sensor. In this context, Comsol Multiphysics resolution using a 3D electromagnetic formulation have been considered to create a database required to train the LS-SVM. Several method exist to find the LS-SVM parameters. Electromagnetism-like Mechanism (EM) algorithm is proposed. A good agreement is obtained between the numerical results and the experimental measure.

Electromagnetic Acoustic Transducer for cracks detection in conductive material

This paper is concerned with the characterization methodologies of defects in conducting materials by an Electromagnetic Acoustic Transducer (EMAT) testing system. It has been developed to create a virtual environment for Non-Destructive Testing (NDT) before implementing it in real, to study the change effect on the defect geometry at the signal received. EMAT is a new technology, which provides a noncontact process of testing materials compared to ultrasonic testing technique. This work is based on the simulation of two-dimensional numerical model, using Finite Element Method, (FEM) like a simulator model forward analysis, which includes the calculation of induced eddy current, the Lorentz force, and mechanical displacement inside conducting material. Results obtained shows that the model is capable of detecting the depth, the width and the location of the surface defect in an Aluminum material, using the mechanical displacement amplitude.

Generalized RBF Neural Network and FEM for Material Characterization Through Inverse Analysis

This paper describes a new methodology for using artificial neural networks (ANN) and finite element method (FEM) in an electromagnetic inverse problem (IP) of parameters identification. The approach is used to identify unknown parameters of ferromagnetic materials. The methodology used in this study consists in the simulation of a large number of parameters in a material under test, using the FEM. Both variations in relative magnetic permeability and electric conductivity of the material under test are considered. Then, the obtained results are used to generate a set of vectors for the training of generalized radial basis function neural networks (RBFNN). Finally, the obtained neural network (NN) is used to evaluate a group of new materials, simulated by the FEM, but not belonging to the original dataset. The reached results demonstrate the efficiency of the proposed approach.