IIC Journal of Innovation 11th Edition | Page 14

Early AI Diagnostics at Westinghouse Figure 3: Westinghouse Steam Purity Analyzer System, ca. 1981. The box was about 5 feet high. When PDS was unveiled in 1982, the Steam Chemistry Engineering section, then managed by the author, had a program of monitoring and was using data loggers in the monitor packages. A typical steam purity analyzer system is shown in Figure 3. It was placed on a sample line drawing steam from the steam generator. The sample was condensed and cooled before it reached the analyzer panel. The analyzer system included chemical monitors, a strip chart recorder and a data logger that was capable of recording a week of data on a cassette tape. 7 Thus, the chemists had a significant amount of online data in digital form and the testing of PDS started with chemistry. Initially, the sensors were treated as correct and a small system was created. The next iteration started to diagnose the condition and accuracy of the sensors, which were generally less reliable than the equipment they monitored because they were operating near the detection limits. Chemical sensors also had the possibility of failing with a reasonable reading. A sodium sensor, for example, might stop responding while displaying 1 ppb (µg/kg), which was a reading within the normal range. The 7 In the 1980s, most power plants did not have steam samples sent to the central sampling system, so a separate Steam Purity Analyzer System was required. The usual samples to the central panel included makeup water, condensate, polisher effluent (if a condensate polisher were present), final feedwater (economizer inlet) and boiler drum water (if the unit had a steam drum). IIC Journal of Innovation - 10 -