Genetic based learning of a civil engineering problem

The Classifier System (CS) (Goldberg (1989), Wilson (1994)) is a machine learning process: the machine (a computer program) 'learns' about a particular environment and is then capable of making beneficial decisions or predictions concerning that environment (Forsyth, 1989). The learning method within a CS is approached from an Evolutionary Computation angle: a rule base attempting to define an environment, is optimised with a Genetic Algorithm (GA) (Goldberg, 1989) until that environment is deemed to be 'learnt'. CSs are of interest to Engineers, particularly those interested in the fields of intelligent real-time monitoring and control, planning, scheduling, signal processing, operations research, failure analysis and forecasting as well as general dynamic system modelling and prediction (Barclay, 1993). This paper is concerned with the latter of these fields. More specifically, it explains the research being undertaken at UWC regarding the Derivation of Reservoir Control Strategies using a form of CS.