Fault Detection and Fault-Tolerant Control in Automation Systems

Automation systems underpin critical infrastructure and industrial processes, from manufacturing and chemical plants to autonomous vehicles and energy grids. Fault detection involves identifying anomalies in sensors, actuators or control loops, enabling early intervention before failures propagate. Fault-tolerant control complements detection by reconfiguring controllers or adjusting reference trajectories to maintain stability and performance in the presence of faults. Traditional model-based approaches rely on precise system models and observers to generate residuals that indicate deviations from expected behaviour. In contrast, data-driven and machine-learning methods leverage historical input/output data or online learning to detect and accommodate faults without explicit plant models. Recent advances combine robust control theory, subspace identification, reinforcement learning and adaptive estimation to deliver resilient automation platforms. The global mandate for higher uptime, safety and sustainability has driven research into plug-and-play architectures, event-triggered monitoring and cyber-physical security, ensuring that systems endure both component malfunctions and malicious interventions with minimal performance degradation.

Research from Nature Portfolio

No recent Nature Portfolio content available.

Fault Detection and Fault-Tolerant Control in Automation Systems publication trend

The graph below shows the total number of articles in fault detection and fault-tolerant control in automation systems across all publications each year (not limited to Nature Index journals).

Technical terms

Fault detection: The process of recognising deviations or anomalies in system behaviour indicating potential malfunctions.

Fault-tolerant control: A control strategy that maintains acceptable system performance by reconfiguring or adapting controllers in the presence of faults.

Residual generator: An observer or filter that produces residual signals by comparing measured and estimated outputs to detect inconsistencies.

Data-driven methods: Techniques that derive models or controllers directly from input/output data without requiring explicit system equations.

Model predictive control (MPC): A control approach that optimises future control actions over a prediction horizon subject to constraints.

Reinforcement learning: A machine-learning paradigm where an agent learns optimal control policies through trial-and-error interactions with the environment.

References

1. Data-Driven Fault-Tolerant Tracking Control for Linear Parameter-Varying Systems. IEEE Access (2022).
2. Identification Modelling and Fault-Tolerant Predictive Control for Industrial Input Nonlinear Actuator System. Machines (2023).
3. Fault-Tolerant Control of Skid Steering Vehicles Based on Meta-Reinforcement Learning with Situation Embedding. Actuators (2022).
4. A Dynamic Event-Triggered Secure Monitoring and Control for a Class of Discrete-Time Markovian Jump Systems: A Plug-and-Play Architecture. Information (2024).