Abstract: The damage identification problem in mechanical structures is a process of determining parameters based on numerical analysis from a comparison on measurement data and the output data from a mathematical model. The main idea behind the damage identification problem, using displacement data, is that the structural damage will manifest a changing in the displacement time response of the system. Structural damage detection is displayed as an inverse vibration problem, since the damage evaluation is obtained through the determination of the stiffness coefficient variation. These problems are usually unstable - small variations in the input data, such as random errors inherent to the measurements used in the analysis, can cause large oscillations on the solution. The inverse problem is presented as a well-posed functional form, whose solution is obtained through an optimization procedure. A variety of experimental, numerical and analytical techniques has already been proposed to solve the damage identification problem. These methods are usually classified under different categories, such as frequency and time domain methods, and deterministic and stochastic approaches [1]. Among the stochastic methods, the use of the artificial neural networks, which has already been used successfully in thermal sciences [3], has also been presented as a satisfactory choice to deal with the damage identification problem. In this work, artificial neural network techniques are applied to the inverse vibration problem where the goal is to estimate the unknown time-dependent stiffness coefficients simultaneously in a two degree-of-freedom structure, using a Multilayer Perceptron Neural Network model. The structural paramaters have been assumed as identical to those from Huang's work [2]. Numerical experiments have been carried out with synthetic experimental data considering a noise level of 1\%. Good recoveries have been achieved with this methodology.

References:

[1] L.D. Chiwiacowsky and H.F. Campos Velho (2003): Different Approaches for the Solution of a Backward Heat Conduction Problem, Inverse Problems in Engineering, 11(6), 471-494.

[2] C. H. Huang (2002): An Inverse Vibration Problem for Simultaneously Estimating the Time-dependent Stiffness Coefficients, 4th International Conference on Inverse Problems in Engineering}, Rio de Janeiro, Brazil.

[3] E.H. Shiguemori, H.F. de Campos Velho, J.D.S. da Silva (2004): Estimation of Initial Condition in Heat Conduction by Neural Network, Inverse Problems in Science and Engineering, 12(3), 317-328.