Industrial application of fuzzy-neuro process monitoring system

Condition monitoring of industrial processes is an excellent application area where fuzzy logic and neural networks are able to demonstrate its usefulness. In this article, we show that by taking a subjective approach towards statistical process control, a natural link exists between humans´ analytic skill and statistical charts which are tools of formal quality control measures. The application of fuzzy logic and neural networks to online monitoring of the frequency trimming process in the production of crystal quartz resonators is discussed. The basic architecture of the online diagnostic system consists of three main functional blocks: data acquisition and processing, fuzzy chart analysis, and pattern-cause association. The characteristic pattern deduced by the chart analysis module is used to diagnose plausible causes of suboptimal operating conditions. This is done by the pattern-cause associator. Backpropagation neural networks are used to correlate chart patterns and attributes