Estimation and diagnosis using multi-models with application to a wastewater treatment plant

• Multi-model gives the possibility to reduce the complexity of nonlinear systems.
• Advantage of multi-model: extend tools of linear theory to nonlinear systems.
• The estimation involves unmeasurable premise variables depending on the state variables.
• The observer estimates state and unknown input simultaneously.
• Methods investigated using strong nonlinear process: wastewater treatment plant.

Process diagnosis is still considered a challenging engineering problem. Technological and also environmental systems have complex behaviors often involving nonlinear relationships. When confronted to such systems, there is a need to build systems that can operate over a wide range of operating conditions. For that it is very attractive to appeal to a decomposition of the system model into a number of simpler linear models. This paper mainly focuses on the use of multi-models for process diagnosis. It is shown how the traditional tools of the linear automatic can be wide and applied to multi-model structures. A proportional multi-integral observer is used for fault diagnosis using banks of observers to generate structured residuals. The performances of the proposed diagnosis method are highlighted through the application to a wastewater treatment plant model (WWTP), which is an uncertain nonlinear system affected by unknown inputs.