10.2190/JNEU-C7TL-8PXL-7A25 JNEUC7TL8PXL7A25 157 Application of Artificial Neural Networks to Update and Modify a DOS-Based Environmental Expert System 157 174 20020509 20020509 20020509 20020509 JNEUC7TL8PXL7A25.pdf http://baywood.metapress.com/link.asp?target=contribution&id=JNEUC7TL8PXL7A25 2 Matthew R. Thompson and William F. McTernan An Artificial Neural Network (ANN) model was developed to mimic the exact output from a DOS-based Environmental Expert System. Computer codes developed originally for mainframe computers and transported into the DOS environment routinely do not receive modifications necessary to perform under more modern operating systems unless there is sufficient financial incentive. Software written for the environmental market, particularly the classroom market, rarely has this level of incentive, resulting in much previously usable software being rendered obsolete. Much of this software still can play a critical role in the education of future environmental scientists and engineers. The subject research investigated one potential solution to this problem: the development of ANN models capable of producing the exact results of the earlier DOS code while having the capability of ready modification given new information or circumstances. This research illustrated the overall utility of ANN s in this capacity, as a 100 percent compatibility between the underlying Expert System and the ANN was achieved. In addition, the ANN was readily modified to include new information. The ANN developed extends the useful life of the Expert System with minimal developmental costs, without extensive re-programming or retrofitting of the original code.