The development process of an expert system for the automated interpretation of large EPMA data sets

Abstract The applicability of the artificial intelligence language OPS5 to represent chemical knowledge in the field of X-ray analysis is evaluated. The problem studied here involves the automated interpretation of large numbers of X-ray spectra obtained by electron probe microanalysis of single particles. The algorithm used during the data reduction phase is outlined and the expert system's data and knowledge base are discussed. Special attention has been paid to the incremental growth process of the knowledge base of the system. Starting from a limited prototype system, which was based on the increase/decrease of probability values of chemical elements, a more powerful expert system was constructed by adding chemical knowledge to its rule base, enabling the system to deal with more complex X-ray spectra. Evaluation of the system's performance shows that it is able to interpret 80–90% of the spectra of a complex data set correctly.