Development of an Expert System Embedding Pattern Recognition Techniques for Pollution Source Identification

摘要

A new methodology for the identification of unknown sources of groundwater pollution is developed. The new methodology is based on the concept of statistical pattern recognition. The statistical pattern recognition algorithm uses Bayes' Optimal Decision Rule. The function of the pattern recognition system is to match statistically an observed set of concentrations in the field with a comparable set obtained by simulating groundwater transport for various disposal conditions. In order to make the application of the methodology easier, an Expert System was developed. The Expert System uses the results obtained by applying the pattern recognition algorithm to select a particular set of pollution source locations and magnitudes. The Expert System also has the capability of adding measures of confidence to alternative selections of sources made by the pattern recognition system, such that these solutions can be ranked according to the subjective confidences supplied by the users. The performance of the pattern recognition system and the Expert System was evaluated for a selected study area.