A process flowsheet expert system for a fluidized bed calciner which solidifies high-level radioactive liquid waste was developed from pilot-plant data using a commercial inductive classification program. After initial classification of the data, the resulting rules were inspected and adjusted to match existing knowledge of process chemistry. The final expert system predicts performance of process flowsheets based upon the chemical composition of the calciner feed and has been successfully used to identify potential operational problems prior to calciner pilot-plant testing of new flowsheets and to provide starting parameters for pilot-plant tests. By using inductive classification techniques to develop the initial rules from the calciner pilot-plant data and using existing process knowledge to verify the accuracy of these rules, an effective expert system was developed with a minimum amount of effort. This method may be applied for developing expert systems for other processes where numerous operating data are available and only general process chemistry effects are known. (ERA citation 17:027900)
展开▼
