Linear parameter-varying modeling and decentralized fault diagnosis for large sets of interacting subsystem

Fault Detection and Isolation (FDI) systems aim at the early detection and localization of faults. They are the key for the establishment of condition-based maintenance on industrial processes, which allows for cost reduction, and a safer and more reliable operation. The research on FDI has been mostly done on a centralized approach, where the information of the supervised system is gathered in a central node. However, this approach becomes unfeasible on large-scale systems due to limitations on computational power or communication infrastructure. Here a decentralized approach is necessary (see Figure 17), since the fault diagnosis is decomposed into different sub-problems aiming at the supervision of smaller parts of the system (subsystems). This lowers the technical requirements, and improves the reliability and security [1].

Concepts of centralized and decentralized fault diagnosis

This project aims at developing a model-based decentralized FDI system, for a process made of large sets of interacting subsystems. A relevant example of such a process is a wind farm, where the turbines are interacting with each other through the wake effect [2]. This project addresses the next challenges:

  • How to identify Linear Parameter-Varying models [3, 4] for the subsystems from available data (accounting for the dynamic dependence on operating conditions). Continuous-time frequency domain identification [5, 6] will be used since (i) a continuous-time model is closer to the physics than a discrete-time model, and (ii) the modeling can be done in a user defined frequency band.
  • How to extend to a decentralized framework the FDI methods currently developed in a centralized way [7, 8], while dealing with the interactions of the subsystems and optimizing the computational load and data exchange. Both additive and multiplicative faults will be considered.

The case study will be wind farms, with focus on the supervision of energy production, and the temperature of certain components of the turbines (See Figure below)

FDI system

References:

[1]   M. Blanke, M. Kinnaert, J. Lunze, M. Staroswiecki & J. Schröder, “Diagnosis and fault-tolerant control,” New York: Springer-Verlag, 2006.

[2]   L. J. Vermeer, J. N. Sorensen, and A. Crespo., “Wind turbine wake aerodynamics,” Progress in Aerospace Sciences, vol. 39, no. 6-7, pp. 467–510, 2003.

[3]   R. Tóth, Modeling and Identification of Linear Parameter-Varying Systems. Berlin: Springer, 2010.

[4]   J. De Caigny, R. Pintelon, J. F. Camino, and J. Swevers, “Interpolated Modeling of LPV Systems,” IEEE Trans. on Control Systems Technology, vol. 22, no. 6, pp. 2232-2246, 2014.

[5]   R. Pintelon, J. Schoukens, and P. Guillaume, “Box-Jenkins identification revisited—Part III Multivariable systems,” Automatica, vol. 43, no. 5, pp. 868-875, 2007.

[6]   R. Pintelon, B. Peeters, and P. Guillaume, “Continuous-time operational modal analysis in the presence of harmonic disturbances–The multivariate case,” Mechanical Systems and Signal Processing, vol. 24, no. 1, pp. 90–105, 2010.

[7]   M. Basseville and I.V. Nikiforov, “Detection of abrupt changes: theory and application,” Prentice-Hall, 1993

[8]   A. Zolghadri, B. Bergeon, and M. Monsion, “A two-ellipsoid overlap test for on-line failure detection,” Automatica, Vol. 29, No. 6, pp. 1517-1522, 1993.

Back to top