Sensor fault and unknown input estimation based on proportional integral observer applied to LPV descriptor systems

Adriana Aguilera González (1) , Didier Theilliol (2) , Manuel Adam Medina (1) , Carlos Manuel Astorga Zaragoza (1) , Mickael Rodrigues (3)

1 CENIDET - Centro Nacional de Investigación y Desarrollo Tecnológico [Mexico]
2 CRAN - Centre de Recherche en Automatique de Nancy
3 LAGEP - Laboratoire d'automatique et de génie des procédés

In this paper a detection, isolation and fault estimation scheme using a Proportional- Integral (PI) Observer that provides a robust estimation against measurements noise and unknown inputs is presented. The main contribution consists in the synthesis of a PI-Observer for descriptor LPV systems, can be useful not only for state estimation, but also to estimate sensor faults and unknown inputs. The proposed PI-Observer can provide an alternative fault diagnosis scheme. Once a fault is detected, a bank of observers is activated for the purpose of fault isolation. This scheme reconstructs the sensor faults based on augmented state equations where an auxiliary state is assigned to represent the dynamic behavior of the fault. An illustrative example is presented in simulation.

Domaines: Automatique / Robotique

Adriana Aguilera González, Didier Theilliol, Manuel Adam Medina, Carlos Manuel Astorga Zaragoza, Mickael Rodrigues. Sensor fault and unknown input estimation based on proportional integral observer applied to LPV descriptor systems. 8th IFAC Symposium on Fault Detection, Supervision and Safety of Technical Processes, SAFEPROCESS 2012, Aug 2012, Mexico City, Mexico. ⟨10.3182/20120829-3-MX-2028.00177⟩. ⟨hal-00727304⟩