Оцінка перспективності артезіанської свердловини з використанням експертних знань

Abstract

A decision support system to evaluate the potential of drinking water mining from an artesian well is introduced. The model is based on an interval type-2 fuzzy set. It does not work with the entire input vector at once, but only with the subset of parameters available at the moment. Unlike the formal approach, based on evaluating each parameter against the bounds of its acceptable values, the present procedure does not raise the requirement of having the complete data on all parameters for the system to function correctly. Of all parameters available at the moment, not all are given as inputs to the fuzzy logic system. Prior to this, an informative feature combination detection procedure takes place, in order to reduce the dimensionality of the problem and exclude redundant features. The initial shape of the membership functions is chosen arbitrarily, given the nature of experimental data. Based on input and output variables and the rules knowledge base, generated from the experimental data, a type-1 fuzzy logic system is built. In order to improve the adequacy of the model’s reflection of the training data, membership functions parameters optimization is performed using a genetic algorithm. The type-1 membership functions obtained as a result of the optimization process are transformed into type-2 membership functions with uncertain means. The system’s output is an interval of values of the output variable. The interval’s width describes the uncertainty related to the decision taken; the values of its left and right bounds allow to provide a preliminary estimation of this artesian well’s potential, without having to wait for the geological exploration to complete.