(250f) Uncertainty-Informed Dynamic Risk Control of Safety-Critical Reaction Processes with Bayesian State Estimation

The paradigm shift into Industry 4.0 has created unique opportunities and challenges for the safe and optimal operations of chemical and energy systems. With this shift, there exists a critical need for real-time process safety management, which features a departure from the conventional protocol assessing a static picture of the process facility (e.g., hazard and operability analysis) [1-2]. Toward this direction, recent works have exploited model-based control to characterize a set of process state variables with theoretical guarantee of safe and stable dynamic operation under uncertainties [3-4]. Model predictive control (MPC) has also been extended to incorporate dynamic risk metrics [5] for fault prognosis, particularly leveraging the moving horizon estimation (MHE) to identify potential faults at the early developing stage [6-7]. However, to attain the desired process safety performance, a key open research question lies in how to immune the advanced control strategies against the dynamic errors of online measurement and state estimation under random noises and/or unavailable measurements.

To address this gap, this work aims to improve the robustness of risk-based multi-parametric MPC strategy developed in our previous work [7] by incorporating system uncertainties into the control problem. A Bayesian Point Mass Filter (PMF) is first utilized to confront non-Gaussian noise by interfacing with the nonlinear dynamic risk model during online implementation. The PMF efficacy will be benchmarked with a classical Kalman filter based on linearized risk models. A constrained moving horizon estimator is then employed to incorporate process safety considerations for fault prognosis while identifying the bounding set of estimation errors [8]. The MHE optimization problem is formulated and solved offline via multi-parametric programming to yield explicit solutions as piecewise affine functions of process inputs, outputs, disturbances, etc. In this way, both the MHE estimator and the controller have the potential to provide offline solutions to reduce the online computational load to a look-up problem. The risk-based MPC is then extended to incorporate tube-based robust control formulations [9] which can smoothly optimize the control actions with respect to process safety and uncertainties. On this basis, the trade-offs between operational optimality, system safety under uncertainties, and online computational load are analyzed. In its entirety, this framework provides a methodical approach for real-time process safety management by employing proactive fault mitigation, integrating process risk metrics and advanced control techniques, while comprehensively combating uncertainties in state measurement and in the process itself. We show that its application is practical and generalizable for both continuous and batch reaction systems.

References

1. [1] Lee, J., Cameron, I. & Hassall, M. Improving process safety: What roles for digitalization and industry 4.0? Process Safety and Environmental Protection 132, 325–339 (2019).
2. [2] Amin, M. T., Arunthavanathan, R., Alauddin, M., & Khan, F. (2022). State-of-the-art in process safety and digital system. In Methods in Chemical Process Safety (Vol. 6, pp. 25-59). Elsevier.
3. [3] Ahooyi, T. M., Soroush, M., Arbogast, J. E., Seider, W. D., & Oktem, U. G. (2016). Model‐predictive safety system for proactive detection of operation hazards. AIChE Journal, 62(6), 2024-2042.
4. [4] Albalawi, F., Durand, H., & Christofides, P. D. (2018). Process operational safety via model predictive control: Recent results and future research directions. Computers & Chemical Engineering, 114, 171-190.
5. [5] Khan, F., Hashemi, S. J., Paltrinieri, N., Amyotte, P., Cozzani, V., & Reniers, G. (2016). Dynamic risk management: A contemporary approach to process safety management. Current Opinion in Chemical Engineering, 14, 9–17.
6. [6] Bhadriraju, B., Kwon, J. S. I., & Khan, F. (2021). Risk-based fault prediction of chemical processes using operable adaptive sparse identification of systems (OASIS). Computers & Chemical Engineering, 152, 107378.
7. [7] Ali, M., Cai, X., Khan, F. I., Pistikopoulos, E. N., & Tian, Y. (2023). Dynamic risk-based process design and operational optimization via multi-parametric programming. Digital Chemical Engineering, 7, 100096.
8. [8] Voelker, A., Kouramas, K., & Pistikopoulos, E. N. (2013). Moving horizon estimation: Error dynamics and bounding error sets for robust control. Automatica, 49(4), 943-948.
9. [9] Limón, D., Alvarado, I., Alamo, T. E. F. C., & Camacho, E. F. (2010). Robust tube-based MPC for tracking of constrained linear systems with additive disturbances. Journal of Process Control, 20(3), 248-260.