Narrow Results

In this paper, a different use of soft computing is shown. Commonly, genetic algorithms are used to solve problems that haven't an analytic formulation; neural networks are used like a black box able to model a system; fuzzy systems try to model human knowledge, transferring the action rules from the expert to a universal support based on fuzzy logic. In this work, instead, genetic algorithms are a way to solve a problem whose formulation doesn't allow to use another global optimization method; the result of the optimization process is used to train a neural network whose structure allows to describe the acquired knowledge in terms of linguistic variables. The efforts are to design an intelligent controller able to solve the problem and give knowledge about what the plant operators have to do, because actually they are unable to manage a good strategy in a risk situation.