Post cyber-attack state reconstruction for nonlinear processes using machine learning

Wu, Zhe; Chen, Scarlett; Rincón, David; Christofides, Panagiotis D.

doi:10.1016/j.cherd.2020.04.018

Cited by 23 publications

(20 citation statements)

References 18 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Also inspired by [38], Proposition 3 demonstrates the fact that the nonlinear system of Eq. 1 can be rendered negative for all times, so that the actual state x can be driven towards the origin under implementation of the stabilizing controller h(x), which utilizes the estimated state, derived from the updated parameters θ(t k ) computed by the RL agent of Algorithm 1.…”

Section: Closed-loop Stability Analysis Of Rl-based Empcmentioning

confidence: 95%

See 1 more Smart Citation

A reinforcement learning-based economic model predictive control framework for autonomous operation of chemical reactors

Alhazmi

Albalawi

Sarathy

2022

Chemical Engineering Journal

View full text Add to dashboard Cite

Economic model predictive control (EMPC) is a promising methodology for optimal operation of dynamical processes that has been shown to improve process economics considerably. However, EMPC performance relies heavily on the accuracy of the process model used. As an alternative to model-based control strategies, reinforcement learning (RL) has been investigated as a model-free control methodology, but issues regarding its safety and stability remain an open research challenge. This work presents a novel framework for integrating EMPC and RL for online model parameter estimation of a class of nonlinear systems. In this framework, EMPC optimally operates the closed loop system while maintaining closed loop stability and recursive feasibility. At the same time, to optimize the process, the RL agent continuously compares the measured state of the process with the model's predictions (nominal states), and modifies model parameters accordingly. The major advantage of this framework is its simplicity; state-of-the-art RL algorithms and EMPC schemes can be employed with minimal modifications. The performance of the proposed framework is illustrated on a network of reactions with challenging dynamics and practical significance. This framework allows control, optimization, and model correction to be performed online and continuously, making autonomous reactor operation more attainable.

show abstract

Section: Closed-loop Stability Analysis Of Rl-based Empcmentioning

confidence: 95%

“…Inspired by the work in [38], Proposition 2 provides an upper bound on the deviation between the actual state and the estimated state obtained from the RL agent in Algorithm 1.…”

Section: Closed-loop Stability Analysis Of Rl-based Empcmentioning

confidence: 99%

A reinforcement learning-based economic model predictive control framework for autonomous operation of chemical reactors

Alhazmi

Albalawi

Sarathy

2022

Chemical Engineering Journal

View full text Add to dashboard Cite

show abstract

“…Proof: The proof of Proposition 1 follows the same lines of the proof of Proposition 2 in [32] and it is omitted for the sake for brevity.…”

Section: Remarkmentioning

confidence: 99%

Tracking Model Predictive Control Paradigm for Underwater Optical Communication

Alalwan

Albalawi

Laleg‐Kirati

2021

IEEE Open J. Commun. Soc.

View full text Add to dashboard Cite

High-precision positioning of two underwater mobile robots is investigated in this work. To achieve good performance in underwater communication, control algorithms are implemented to maintain the position of the receiver robot aligned with that of the transmitter in the presence of measurement noise and process uncertainty. Although recent research works have successfully integrated control algorithms with Extended Kalman Filter (EKF) estimator to track the desired position of the transmitter, other aspects besides the convergence to the equilibrium point such as operational constraints and input constraints were not taken into account within these controllers. Such inability of these control algorithms may degrade the performance of the controlled process. Motivated by the above considerations, a tracking Model Predictive Control (MPC) with an EKF-based estimator is developed to both estimate the process states online and drive the actual system to the desired equilibrium point while meeting input and state constraints. The closed-loop stability and the recursive feasibility of the proposed tracking MPC scheme are rigorously proved. To demonstrate the applicability of the proposed control design, the performance of the tracking MPC with that of the conventional Proportional (P), Proportional Integral Derivative (PID), and Linear Quadratic Regulator (LQR) controllers are compared.

show abstract

“…[13] Integrated cyberattack detection and advanced control methods using model predictive control (MPC [14] ) also have been developed as an attempt to identify attacks while guaranteeing closed-loop stability even in the presence of cyberattacks. [15,16] In ref. [15], a neural network-based detection method combined with a model predictive controller for nonlinear systems was designed to potentially detect sensor tampering.…”

Section: Introductionmentioning

confidence: 99%

“…After the detection of cyberattacks, to retain control of the system and ensure safe operation ref. [16] proposes a state reconstruction algorithm using machine learning methods based on the falsified state measurements. Moreover, cyberattack detection and MPC techniques with economics‐based objective functions, named economic model predictive controllers (EMPC [ 17 ] ), have been investigated in light of Lyapunov‐based constraint properties (Lyapunov‐based EMPC or LEMPC [ 18 ] ) to maintain closed‐loop stability of the system, under sufficient conditions, both in the absence of and in the presence of cyberattacks.…”

Section: Introductionmentioning

confidence: 99%

Handling of stealthy sensor and actuator cyberattacks on evolving nonlinear process systems

Oyama

Rangan

Durand

2021

J Adv Manuf & Process

View full text Add to dashboard Cite

Cyberattacks on control systems in the chemical process industries cause concern regarding how they can impact finances, safety, and production levels of companies. A key practical challenge for cyberattack detection and handling using process information is that process behavior evolves over time. Conceivably, changes in process dynamics might cause some detection strategies to flag a change in the dynamics as an attack due to the new data appearing abnormal compared to data from before the dynamics changed. In this work, we utilize several case studies to probe the question of what might be the impacts, benefits, and limitations of cyberattack detection and handling policies when the process dynamics change over time. The goal of this work is to characterize, through simulation studies, characteristics, which might be desirable and undesirable in cyberattack detection and handling procedures when

show abstract

Post cyber-attack state reconstruction for nonlinear processes using machine learning

Cited by 23 publications

References 18 publications

A reinforcement learning-based economic model predictive control framework for autonomous operation of chemical reactors

A reinforcement learning-based economic model predictive control framework for autonomous operation of chemical reactors

Tracking Model Predictive Control Paradigm for Underwater Optical Communication

Handling of stealthy sensor and actuator cyberattacks on evolving nonlinear process systems

Contact Info

Product

Resources

About