Regularização e hibridização aplicadas à identificação de danos em estruturas esbeltas através de um algoritmo de propagação de ondas acústicas

• Autor(a) principal: Baptista, Josué Costa
• Data de Publicação: 2017
• Tipo de documento: Dissertação
• Idioma: por
• Título da fonte: Biblioteca Digital de Teses e Dissertações da UERJ
• Texto Completo: http://www.bdtd.uerj.br/handle/1/13833
• topic: Damage identification; Inverse problems; Regularization; Hybridization; Identificação de danos; Problemas inversos; Regularização; Hibridização; Estruturas Modelos matemáticos; Impedância acústica; Otimização matemática; CNPQ::CIENCIAS EXATAS E DA TERRA::FISICA

The modified sequential algebraic algorithm (MSAA) is a routine developed to solve the plane acoustic wave propagation problem in inhomogeneous media. This method accurately predicts the echo generated by the inhomogeneities in a propagating medium. Used for solving the inverse problem of damage identification, by a stochastic optimization approach, MSAA shows to be eficient to identify rough profiles of impedance with low noise level in the input data. As noise level increases, the `ill-posed' feature of the inverse problem is accentuated resulting in an error propagation which raises absolute error to 30%. The purpose of this work is to apply the regularization techniques of Tikhonov (RT) and Total Variation (RTV) and to implement a hybrid optimization strategy, composed by the stochastic method of Luus-Jaakola and the deterministic procedure of Levenberg- Marquardt, to reduce the error propagation in the damage identification. In the numerical examples, 40 parameters of 5 different damages are successfully identified for signals of a synthetic experiment with SNR between 18 and 27 dB. For all tested cases, the regularization and hybridization procedures decreased the maximum absolute error to levels lower than 4%, and for the roughest impedance profile, the mean squared error was reduced by more than one order of magnitude and the error propagation was avoided.