Research in the International Journal of Automation and Control describes a novel mathematical framework that can estimate fault risk in complex industrial systems, such as petrochemical reactors. The research might improve reliability of automated processes in the petrochemical sector and in other areas of advanced manufacturing.
The work builds on a predictive defect estimation model. It works for switched non-linear the behaviour of which alternates between different operating modes depending on conditions or control inputs. This is a common modality of chemical reactors, robotic platforms, and automated processes. The accurate prediction of imminent issues with such systems has always been a challenge because system response can vary widely with changing inputs, environmental c…
Research in the International Journal of Automation and Control describes a novel mathematical framework that can estimate fault risk in complex industrial systems, such as petrochemical reactors. The research might improve reliability of automated processes in the petrochemical sector and in other areas of advanced manufacturing.
The work builds on a predictive defect estimation model. It works for switched non-linear the behaviour of which alternates between different operating modes depending on conditions or control inputs. This is a common modality of chemical reactors, robotic platforms, and automated processes. The accurate prediction of imminent issues with such systems has always been a challenge because system response can vary widely with changing inputs, environmental conditions, and internal dynamics.
By using a mathematical representation of the system, a so-called augmented state-space model, variables describing the current condition of the system and failure signals, which indicate the presence of faults, the new model can evaluate how closely a system’s predicted behaviour matches its actual behaviour. Discrepancies between the two are analysed statistically by the model to home in on whether the system is stable or liable to fail.
The researchers point out that traditional fault detection methods have always been limited by restrictive assumptions about system dynamics. The new approach allows for continuous real-time monitoring in industrial environments. Tests with a stirred tank reactor as a standard benchmark for modelling chemical reactions gave fault-free accuracy within 0.05 and successfully detecting both constant and time-varying faults. The current system works to detect faults that develop gradually. Real-world industrial settings introduce additional complexities, such as abrupt faults, high-frequency disturbances, and measurement noise, all of which will require further refinement of the model.
Wang, L. (2025) ‘Design of a model predictive-based fault estimator for faulty nonlinear switched dynamics with guaranteed recursive feasibility’, Int. J. Automation and Control, Vol. 19, No. 7, pp.1–22.