Single and multiple faults in system actuators and sensors for ethanol production C.A. Aceves-Lara 1, 2, 3 , D. Fragkoulis 4 , G. Roux 5, 6 and B. Dahhou 5, 6 1 Université de Toulouse ; UPS, INSA, INP, LISBP ; F-31077 Toulouse, France 2 INRA, UMR792, Ingénierie des Systèmes Biologiques et des Procédés, Toulouse, France 3 CNRS, UMR5504, Toulouse, France 135 avenue de rangueil, Toulouse Cedex, F-31077 (Tel : 0561559444 ; e-mail : aceves@insa-toulouse.fr) 4 Halkis Institute of Technology, 34400,Psahna,Evia, Greece (dfragoulis@gmail.com). 5 CNRS ; LAAS ; 7 avenue du Colonel Roche, F-31077 Toulouse, France 6 Université de Toulouse ; UPS, INSA, INP, ISAE ; UT1, UTM, LAAS ; F-31077 Toulouse, France (Tel : 0561559494 ; e-mail : [roux, dahhou]@laas.fr) Abstract: This paper focuses on the study of a fault detection and isolation system strategy applied to an ethanol production bioprocess that involves a two-stage bioreactor with a cell recycling loop to reach high biomass concentrations. The various actuators and sensors of this type of bioprocess, impose to develop a fault detection and isolation strategy. In this work, we study the case where multiple faults (with a small delay between them) or simultaneous faults can occur. We assume that only one type of faults can occur at a time and we will focus on the actuators and sensors faults. The sensor fault problem will be reformulated to an actuator fault one by introducing a state variable transformation, so that an augmented system is constructed. Thus we will design a nonlinear model based on an adaptive observer method, for detection, isolation and identification of actuator and sensor single and multiple faults. These approaches use the system model and the outputs of the adaptive observers to generate residuals. Residuals are defined in such way to isolate the faulty instrument after detecting the fault. The validity of the method will be tested in simulation in a nonlinear model of a two-stage ethanol bioprocess with a cell recycling loop. Keywords: Bioprocess, Fault Detection, Fault Isolation, Fault Identification, Observers, Ethanol Production. 1 INTRODUCTION The new century impose environmental challenges such as water supply, global warming and new energy sources for substitution of fossil fuels. These two last are closely dependent. In our days ethanol is the main biofuel used in Europe whose production is now based on old technology with performance that requires innovative culture strategies to optimize productivity. In order to overcome this challenge, an innovative bioprocess has been studied by Aldiguier (2006) and Ben Chaabane et al. (2006). A two-stage continuous bioreactor with a cell recycling loop allowed a productivity of 41 kg.m -3 .h -1 to be reached with an ethanol titer of 8.3°GL in the second bioreactor (Ben Chaabane et al. 2006). The two previous works propose a static and a dynamic optimizations of this two-stage continuous bioreactor. In the first time, a steady-state optimization for ethanol production was carried out using a mathematical model based on the mass balance for a two-stage bioreactor. The volume ratio (V 1 /V 2 ) and substrate feed concentrations (S f1 and S f2 ) were varied during optimization in order to optimize process design (Aceves-Lara et al. 2010a). In addition, an approach of dynamic optimization of ethanol production by using an optimal closed loop control was studied (Aceves-Lara et al. 2010b). Two algorithms were proposed for applied a model predictive control (MPC); a Pattern Search algorithm (PS) and an Interactive Ant Colony Algorithm (IACA). The final objective of this work is to validate online the method proposed in an experimental pilot. The various actuators and sensors of this kind of bioprocess impose to develop a fault detection and isolation system strategy. Fault detection and isolation (FDI) techniques could prevent from all the undesirable consequences of the faults and so it is becoming an attractive topic. Stimulated by this growing demand for improving the reliability many methods of FDI have been developed during the last decades with the aim to reduce the isolation time of the procedure. The various research groups propose approaches of FDI based on the expertise of their own field and/or experiment on a specific class of systems. The diversity of the solutions was also enriched by the growing interest of the industry. Model based fault detection and diagnosis systems have found extensive use because of the fast response to abrupt failure and the implementation of the model based FDI in real-time algorithms. The most common methods for model based FDI are either based on state or parameter estimation. A comprehensive review of the different FDI methods and their applicability to a given physical system has been presented in Isermann (1994). The methods based on observers are rather well developed, especially for the linear systems. Various types of observers were created according to the nature of the considered problem. More recent work treats theoretical development for fault detection and isolation methods of nonlinear systems (García and Frank, 1997; Frank et al., 1999; Hammouri et al., 1999; Nijmeijer and Fossen, 1999; 8th IFAC Symposium on Fault Detection, Supervision and Safety of Technical Processes (SAFEPROCESS) August 29-31, 2012. Mexico City, Mexico 978-3-902823-09-0/12/$20.00 © 2012 IFAC 228 10.3182/20120829-3-MX-2028.00099