Ultrasonic characterization of porous silicon using a genetic algorithm to solve the inverse problem

Abstract

International audienceThis paper presents a method for ultrasonic characterization of porous silicon in which a genetic algorithm based optimization is used to solve the inverse problem. A one dimensional model describing wave propagation through a water immersed sample is used in order to compute transmission spectra. Then, a water immersion wide bandwidth measurement is performed using insertion/substitution method and the spectrum of signals transmitted through the sample is calculated using Fast Fourier Transform. In order to obtain parameters such as thickness, longitudinal wave velocity or density, a genetic algorithm based optimization is used.A validation of the method is performed using aluminum plates with two different thicknesses as references: a good agreement on acoustical parameters can be observed, even in the case where ultrasonic signals overlap.Finally, two samples, i.e. a bulk silicon wafer and a porous silicon layer etched on silicon wafer, are evaluated. A good agreement between retrieved values and theoretical ones is observed. Hypothesis to explain slight discrepancies are proposed

Laser-excited elastic guided waves reveal the complex mechanics of nanoporous silicon

“…These however require measurements of bulk waves in different crystallographic directions 27 , which evidently is problematic for thin porous membranes. Therefore, these methods generally only reported values for the longitudinal bulk wave velocity in the thickness direction 28,29 . Furthermore, most of these methods typically use a coupling medium between the transducers and the sample under investigation.…”

Inverse-Problem-Based Accuracy Control for Arbitrary-Resolution Fairing of Quasiuniform Cubic B-Spline Curves

“…The numerical results confirm that this method produced convergent and stable numerical solutions [15]. Bustillo et al presented a method for ultrasonic characterization of porous silicon in which a genetic algorithm-based optimization is used to solve the inverse problem [16]. Capasso et al solved inverse problems of differential equations by a "generalized collage" method and application to a mean field stochastic model [17].…”

Determination of microscopic parameters of piezoceramic materials under electrical loading using genetic algorithm

“…Then, Goldberg developed a floating-point based genetic algorithms. This optimization method is currently used in several research fields [3], [10], [11]. In this study, we have chosen a fitness function based on the least-square distance between theoretical and experimental curves.…”

“…For simplicity, the inverse of distance is used. In order to limit premature convergence risk, a parameter is added to the distance to limit the maximal value of fitness function (5) [10]. This parameter is set at 1, enabling a good convergence of this optimization problem [10].…”

“…GA population size and mutation rate choice is very important to guarantee a good convergence behavior during optimization. In this study, population size is set at 20 chromosomes and the mutation rate at 20%, as proposed in literature for optimization procedures [10], [15] to ensure a good convergence of real-based genetic algorithms. End condition is fixed at 500 generations in order to achieve a good precision on retrieved parameters.…”