A Digital Twin for Event Prediction and Mitigation
Abnormal events including equipment faults, process disturbances, unplanned changes, and unexpected combinations can cause huge process costs, including degraded product, lost production time, and engagement of automatic safety systems that effect the entire plant. One of the important roles of the operator is to keep the process running, even if in a degraded state, while these issues are addressed or pass. Alarms are the first step, but being able to see qualitative changes to the process earlier allows maintenance to be done, or adjustments to be made in a timely and less costly manner.
Most plants do not have real-time mechanistic models detailed enough to provide their operators with real-time advice for developing situations. PPCL’s geometry-based methods can use process history and engineering knowledge to detect developing events and directly advise your operators.
By modelling fault-free operation far from process disturbances, a Geometric Process Control (GPC) model is very sensitive not just to single variables, but to the relationship between variables, allowing early warning of developing faults and abnormal events. The operator is informed that not only the process has gone outside the envelope, but also the variables involved, allowing them to pinpoint the issue much faster than with traditional fault-detection methods like PCA and PLS.
Who builds these GPC models? Your existing process and control engineers after training and coaching from PPCL. Experienced users say they can develop and commission a new model in a day because thanks to the inherently visual nature of geometry there are no equations to formulate and solve and no heavy maths. The knowledge you will need is good understanding of the process, its operating procedures and actual operating practices that Unit process engineers and control engineers possess.
First presented on: 27th and 28th April 2022.
Presenter: Alan Mahoney PhD