Development of Adaptive Fault and Anomaly Detection for Industrial Processes and the Application of Reinforcement Learning (RL) for Automatic Fault Recovery
The Operational Excellence (OpEx) team at Spartan Controls is actively involved in several initiatives for developing advanced process control (APC) solutions to the oil sands industry. The OpEx team collaborates with Professor Biao Huang’s research group through the NSERC Industrial Research Chair (IRC) in the Control of Oil Sands Processes program for solutions that require extensive research and development. This proposed project will complement the on-going joint research efforts with the development of new data analysis techniques to address the APC problems. Due to increasing number of sensors in industries, a large amount of process and alarm data are collected that has a useful information about the process. To detect, isolate, and identify any kinds of potential abnormalities and possible faults, the real time process monitoring and fault diagnosis algorithm should be applied. Also, to detect possible faults and anomalies which were not in the historical data, the adaptive framework is applied to capture any possible future faults and anomalies. Moreover, to make the task of taking actions to deal with the faults more automated, a RL-based techniques are going to be used.