(64b) A Stylized Trend Analysis Approach for Process Monitoring and Fault Diagnosis

A stylized trend analysis was developed to identify recurring or standard operating procedures (SOP) such as equipment switching or maintenance from plant process history data to detect faults and near misses due to operator errors or equipment failures. Trend based fault detection and diagnosis systems have not been implemented widely in the chemical process industries because known fault scenarios are typically required. Known fault scenarios were not required for the stylized trend analysis because faults and near misses were identified by comparing trends with normal SOP transition responses. Large volumes of historical data were processed automatically allowing corrective action to be taken prior to an incident. The stylized trend analysis was demonstrated to detect unsteady-state transition faults on historical distributed control system (DCS) data simulated for a continuous process ethylene plant dryer switching operation.