An intelligent computing technique in identification problems
Abstract
The paper is devoted to the application of the evolutionary algorithms, gradient methods and artificial neural networks to identification problems in mechanical structures. The special intelligent computing technique (ICT) of global optimization is proposed. The ICT is based on the two-stage strategy. In the first stage the evolutionary algorithm is used as the global optimization method. In the second stage the special local method which combines the gradient method and the artificial neural network is applied. The presented technique has many advantages: (i) it can be applied to problems in which the sensitivity is very hard to compute, (ii) it allows shortening the computing time. The key problem of the presented approach is the application of the artificial neural network to compute the sensitivity analysis. Several numerical tests and examples are presented.
Keywords
References
[1] J. Arabas. Evolutionary Algorithms Lectures (in Polish). WNT, 200l.[2] H. D. Bui. Inverse Problems in the Mechanics of Materials: An Instroduction. CRC Press, Bocca Raton, 1994.
[3] T. Burczyński. Boundary Elements Method in Mechanics (in Polish). WNT, Warszawa 1995.
[4] T. Burczyński, W. Beluch, A. Długosz, P. Orantek, M. Nowakowski. Evolutionary methods in inverse problems of engineering mechanics. In: M. Tanaka and G. S. Dulikravich, eds., Inverse Problems in Engineering Mechanics II, pp. 553-562, Elsevier, 2000.
[5] T. Burczyński, W. Beluch, A. Długosz, W. Kuś, M. Nowakowski, P. Orantek. Evolutionary computation in optimization and identification. Computer Assisted Mechanics and Engineering Sciences, 9: 3-20, 2002.
