Structural optimization using evolution strategies and neural networks (Q1971478): Difference between revisions

The authors attack structural optimization problems by means of a combination of evolution strategies (ES) and artificial neural networks (ANN). Since the computation of objective and constraint functions of real world problems may be time-consuming, these functions are simulated by means of an ANN. The latter is trained using specified data and a ``feedforward error back propagation'' training procedure. The versions of ES proposed by \textit{H.-P. Schwefel} [Numerische Optimierung von Computer-Modellen mittels der Evolutionsstrategie. Mit einer vergleichenden Einführung in die Hill-Climbing- und Zufallsstrategie. Interdisciplinar Systems Research. 26. Basel-Stuttgart: Birkhäuser Verlag (1977; Zbl 0347.65035)] are applied. The authors give an extensive introduction into ES and AN, and present several examples of application of the described method.