Bayesian regularization neural networks for optimizing fluid flow processes

This paper details the application of neural networks and evolutionary strategies for shape optimization problems. These, commonly grouped under the name soft computing methods, are quite recent methods and in demand for application to engineering optimization tasks. For complex problems, like in non-convex optimization, such heuristic techniques are able to outperform conventional optimization methods. We show that the use of progressive network models can yield satisfactory results when optimizing engineering relevant applications. Numerical examples are given.