Development and experimental validation of a predictive resilient framework for cyber‐physical systems

Abstract

Cyber‐physical process systems integrate process dynamics, control, sensing, actuation, and communication, making them vulnerable to attacks that falsify data, alter control actions, or disrupt networks and drive operations toward unsafe states. This work presents a safety‐focused resilience framework that adds an independent protection layer to existing control systems. Using trusted state and controller‐output measurements, a controller digital twin (DT) checks whether control actions are consistent, while a process DT predicts future behavior after detecting a compromise. These predictions are assessed using the minimum safety margin and the time to minimum safety margin, which indicate risk severity and urgency and classify conditions as Safe, Possible Critical, or Safety Critical. Mitigation actions are chosen when they can restore an acceptable safety margin; otherwise, the system proceeds to safe shutdown. Simulated continuous stirred‐tank reactor and experimental continuous stirred‐tank heater studies show the layer detects attacks, supports recovery, and triggers automatic shutdown when needed.