IIC Journal of Innovation 11th Edition | Page 18

Early AI Diagnostics at Westinghouse turbine diagnostic system, GTAID, is used to follow over 1200 gas turbines. Including the system shell programmers in the team meant that a request for a new feature could be accommodated in a few days. Because roadblocks due to limitations of PDS could be removed quickly, the knowledge engineers could continue working, almost without interruption. The development group, at its peak about 1989, included two programmers who maintained the expert system shell, two chemists working on ChemAID, four mechanical engineers working as knowledge engineers on GenAID, four engineers working on TurbinAID and a physicist who acted as an instrument inventor so that when a measurement was needed, there was a way to get it. In addition, specialists from the entire engineering staff of Westinghouse made contributions to individual rules. A third important lesson for the construction of such systems was the step by step approach. This approach promotes transparency, and transparency is useful in explaining a diagnosis. It is also valuable when a section of the rulebase must be modified. Modification of complex rules invites unintended consequences. Modification of simple, one-step rules is much less likely to have unintended consequences. L ESSONS L EARNED One important lesson to be found in this narrative is that the audience for the expert system is important to its potential commercial success. GenAID ® was written for the operator, who might know that something was wrong, but was usually not able to diagnose the problem. ChemAID ® was written for the chemist, who could generally diagnose the chemistry problems. ChemAID ® did not survive the lean years of the 1990s. The author would like to acknowledge helpful conversations with Eric Harper, one of the PDS programmers, who first suggested writing this paper. The author would also like to thank all the people who contributed to the development of the Westinghouse AI Diagnostics systems for a very rewarding and stimulating decade during the 1980s. A second important lesson was the integration of the development team. IIC Journal of Innovation - 14 -