Coupling databases to expert systems allows the system to be used to make predictions about plant operations and performance; this capability is in addition to the traditional uses of databases and expert systems for diagnostic support and knowledge remediation. By using either regression analysis or artificial neural networks to model the data structure, expected chemical doses can be ascertained. For simple systems with few inputs, the regression and ANN approaches provide similar results, with the regression analysis more effective at predicting excursions. Based on a risk of underdosing, the regression model can predict the required chemical dose to virtually assure adequate chemical dosing under all circumstances. In another use of the data, the expert system described in this paper can be made more site specific by simple statistical analysis of a particular water plant's records. For the diagnosis component, heuristic/qualitative descriptors can be related to quantitative terms by linking the descriptors to data ranges. The tutorial component can also identify execution options based on similar simple statistics. This paper discusses the development of a coupled system, focusing on regression analysis and artificial neural networks.
| File Size : | 1
file |
| ISBN(s) : | 0898675898 |
| Note : | This product is unavailable in Ukraine, Russia, Belarus |
| Number of Pages : | 16 |
| Published : | 01/01/1991 |
