(373d) Data-Driven Prescriptive Maintenance Scheduling and Process Optimization

Maintenance refers to acts undertaken to improve the availability and integrity of ageing productive systems, and is at the nexus of the broader concepts of system resilience, system effectiveness, and complexity. Compromised system resilience can be catastrophic, with consequences such as cost to human life due to process safety incidents, lost revenue due to downtime, as well as damage to the system and the environment. Maintenance scheduling can improve system effectiveness, however due to system complexity, the optimal allocation of resources in selecting when and where to maintain process equipment is non-trivial. Existing approaches have focused on the use of equipment reliability models, degradation signal models, and fault detection models to help address this challenge. Limited attention has been given to future failure prediction in the literature, and efforts have typically focused on probability of failure prediction without consideration of process information, system resilience, or risk via inclusion of the consequences of equipment failure.

This research leverages the availability of data, and complex data-driven models to help guide the optimal allocation of resources in complex systems via the inclusion of information about process operations and equipment condition to obtain optimal maintenance schedules. Equipment data is fed to a non-linear machine learning regression model to determine the remaining useful life (RUL) distribution of equipment for future failure prediction. Knowledge of future failure is then used by a maintenance scheduling model to determine the optimal maintenance schedules via multi-objective optimization of system effectiveness and system resilience as quantified by safety metrics. The results of this research are a set of Pareto-optimal data-driven maintenance schedules from which the decision-maker can select. This research involves automated and dynamic assessment of the risks associated with process hazards, and can be used to help ensure system resilience.

References

1. Onel, Melis, Chris A. Kieslich, and Efstratios N. Pistikopoulos. "A nonlinear support vector machine‐based feature selection approach for fault detection and diagnosis: Application to the Tennessee Eastman process." AIChE Journal 65.3 (2019): 992-1005.
2. Jain, Prerna, Efstratios N. Pistikopoulos, and M. Sam Mannan. "Process resilience analysis based data-driven maintenance optimization: Application to cooling tower operations." Computers & Chemical Engineering 121 (2019): 27-45.
3. Baptista, Marcia, et al. "Forecasting fault events for predictive maintenance using data-driven techniques and ARMA modeling." Computers & Industrial Engineering 115 (2018): 41-53.
4. Yildirim, Murat, Xu Andy Sun, and Nagi Z. Gebraeel. "Sensor-driven condition-based generator maintenance scheduling—part II: Incorporating operations." IEEE Transactions on Power Systems 31.6 (2016): 4263-4271.
5. Bousdekis, Alexandros, et al. "Review, analysis and synthesis of prognostic-based decision support methods for condition based maintenance." Journal of Intelligent Manufacturing (2016): 1-14.
6. Van Horenbeek, Adriaan, and Liliane Pintelon. "A Dynamic Prognostic Maintenance Policy for Multi-Component Systems." IFAC Proceedings Volumes 45.31 (2012): 115-120.
7. Vassiliadis, Constantinos G., et al. "Simultaneous maintenance considerations and production planning in multi-purpose plants." Annual Reliability and Maintainability Symposium. 2000 Proceedings. International Symposium on Product Quality and Integrity (Cat. No. 00CH37055). IEEE, 2000.