Knowledge-Based Process Control Using Rough Sets

摘要: A process model for heavy oil upgrading was developed using a machine learning system based on rough sets. The has incorporated temporal patterns control-loop responses and key relationships between the variables at low (feedback) high (supervisory) control levels. predicted reactor temperature distribution with 90 to 95 percent accuracy. Accuracy depended number of training cycles resolution used. advantages sets approach were: 1) it allowed use qualitative quantitative information in model; 2) provided unified description events patterns; 3) permitted “raw” sensor data without preprocessing.