Polynomial Residuals for Fault Detection and Isolation

Polynomial Residual Generators for Fault Detection and Isolation in Dynamic Processes. Seminar of two hours for the University of Science & Technology Beijing - USTB.
 

Abstract

• This talk presents a model-based strategy for detecting and isolating sensor faults in complex dynamic processes by exploiting input–output polynomial descriptions. Residual signals are generated by dynamic filters obtained from the left null space of disturbance channels, which ensures decoupling from unknown inputs while preserving sensitivity to faults. The resulting residual generators are assembled into banks that enable isolation at both input and output sensors. The approach admits a principled tuning of the filter polynomial to shape transient behaviour and to maximise steady-state fault sensitivity, whilst mitigating the influence of measurement noise and modelling error. The method is algorithmically simple, relying only on measured inputs and outputs, yet it aligns with well-established canonical forms and can be implemented in discrete or continuous time. Its effectiveness is illustrated on two representative applications: a thermal-power process and a general aviation aircraft, where minimal detectable faults and detection delays are assessed under realistic noise and uncertainty. Compared with observer-based schemes, the proposed synthesis reduces design burden without sacrificing diagnostic capability. The talk summarises the mathematics, the design workflow, and practical guidelines for threshold selection, and highlights conditions that guarantee systematic disturbance decoupling and reliable fault isolation across a wide range of operating regimes.

Keywords

• Fault detection and isolation, polynomial residual generators, disturbance decoupling, input-output canonical forms, industrial and aeronautical case studies.

Main Points

• Polynomial input-output modelling and canonical forms
• Residual generator synthesis
• Banked filters for Fault Detection and Isolation
• Industrial validation on a thermal-power process
• Aeronautical validation on a general aviation aircraft
• Robustness and reliability analysis
• Simulation-based optimisation.

Downloads: Talk Slide Files

• Seminar Slides (PDF file, 1 MB).

Selected References

• Bonfè, M., Castaldi, P., Geri, W., & Simani, S. (2004). Residual Generator Computation for Fault Detection of a General Aviation Aircraft. IFAC. PDF file (0.15 MB).
• Simani, S., Fantuzzi, C., & Beghelli, S. (2000). Diagnosis Techniques for Sensor Faults of Industrial Processes. IEEE Transactions on Control Systems Technology, 8(5), 848-855. PDF file (0.16 MB).
• Bonfè, M., Castaldi, P., Geri, W., & Simani, S. (2005). Residual Generator Computation via Polynomial Approach for Fault Detection and Isolation in Dynamic Processes. Technical Note. PDF file (0.4 MB).

Previous Page

Simani Home Page

Dipartimento di Ingegneria