Zonotopic recursive least-squares parameter estimation: Application to fault detection

The uncertain parameters estimation problem for linear discrete-time systems modeled in regression form and assuming an unknown but bounded description of the noise, is the main subject of this work. Particularly, zonotopic sets for bounding the parametric uncertainty are considered. Then, a zonotop...

Descripción completa

Detalles Bibliográficos
Autores: Samada Rigo, Sergio Emil, Puig Cayuela, Vicenç|||0000-0002-6364-6429, Nejjari Akhi-Elarab, Fatiha|||0000-0001-9118-632X
Tipo de recurso: artículo
Fecha de publicación:2023
País:España
Institución:Universitat Politècnica de Catalunya (UPC)
Repositorio:UPCommons. Portal del coneixement obert de la UPC
Idioma:inglés
OAI Identifier:oai:upcommons.upc.edu:2117/383370
Acceso en línea:https://hdl.handle.net/2117/383370
https://dx.doi.org/10.1002/acs.3557
Access Level:acceso abierto
Palabra clave:System failures (Engineering)
Parameter estimation
Bayesian inference
Deterministic systems: Parameter estimation
Parametric fault detection
Stochastic systems
Zonotopic least squares
Enginyeria de sistemes
Estimació de paràmetres
Àrees temàtiques de la UPC::Informàtica::Automàtica i control
Descripción
Sumario:The uncertain parameters estimation problem for linear discrete-time systems modeled in regression form and assuming an unknown but bounded description of the noise, is the main subject of this work. Particularly, zonotopic sets for bounding the parametric uncertainty are considered. Then, a zonotopic recursive least-squares (ZRLS) estimator is proposed and compared with its stochastic counterpart (RLS), as well as with the set-membership (SM) approach. Likewise, the parameter estimation problem is addressed from the Bayesian general framework to define the relationship between stochastic and deterministic approaches. Moreover, both ZRLS and SM applications applied to fault detection are assessed taking as a reference the minimum detectable fault in the worst case. Finally, a well-known quadruple-tank process is used to illustrate the estimation and fault detection results.