Korean J. Chem. Eng., 15(3), 310-316 (1998) FUZZY CONTROL OF ETHANOL CONCENTRATION FOR EMULSAN PRODUCTION IN A FED-BATCH CULTIVATION OF A CINETOBA CTER CALCOA CETICUS RAG-1 Jeong-Woo Choi t, Seung-Mok Oh, Hyun-Goo Choi, Sang-Baek Lee*, Kwang-Soon Lee and Won-Hong Lee Department of Chemical Engineering, Sogang University, C.P.O. Box 1142, Seoul 100-611, Korea *Department of Chemical Engineering, Cheju National University, Cheju, Korea (Received 3 December 1997. accepted 21 March 1998) Abstract-A fuzzy control system was organized and applied to the control of ethanol tamcentration in a fed- batch cultivation process for emuisan production by Acinetobacter calcoaceticus RAG-1. The membership func- tions and fuzzy rules were determined by sets of data and experiences obtained from the preliminary culture ex- periments. The input variables, error (the difference between the set point value and the process variable) and the change of the error, were fuzzified by using the membership functions and the output variable, change of the ethanol feed rate, was inferred based on the membership functions and the given fuzzy rules. To obtain the nu- merical value for the output variable, the center-of-gravity method was used in the defuzzification procedure. The results showed that the ethanol concentration was well regulated around optimal level and the emulsan yield was in- creased compared with that of the cultivation controlled by the conventional feedback control loop. Key words : Fuzzy Control, Fed-batch Cultivation, Emulsan, Acinetobacter calcoaceticus RAG-1 INTRODUCTION Modem control theory and intelligent system instrumenta- tion have been applied to the great number of bioprocesses as one of the key features of development in biotechnoiogy. To achieve the optimal control of the objective process, pro- cess modeling is considered as an essential tool to be accom- plished first. In process modeling, the main goal is to obtain the accurate relationship between the manipulated variables and controlled variables. However, the accurate process model de- scribing the dynamic characteristics of the process can not be easily obtained since most of chemical or biological processes include several nonlinear elements and unexpected disturbances. Therefore, the processes are usually operated based on the in- formation obtained from the experiences and knowledge of the experts [Stephanopoulos, 1987; Konstantinov and Yoshi- da, 1992]. Knowledge based control system has been developed and widely applied to complex processes, such as biological pro- cesses, with control problems due to the nonlinear and uncer- tain dynamic characteris'tics of the processes [Dohnal, 1985; Stephanopoulos, 1986; Czogala and Rawlik, 1989; Thibault et al., 1990]. Since knowledge based control system has a flex- ibility to handle the process information mathematically and/ or linguistically, it is often chosen as an alternative control scheme to the conventional ones. The fuzzy control algorithm, one of the simplest architecture of knowledge based control tTo whom all correspondence should be addressed. E-mail : jwchoi@ccs.sogang.ac.kr 310 systems, was developed for the effective use of uncertain and qualitative information, and has been applied to the simula- tion or control in many fields of the practical processes [Filev et al., 1985; Kishimoto et al., 1991; Yamakawa, 1992; Zhang and Edmunds, 1992; Alfafara et al., 1993; Katayama et al., 1993]. Considering the complexity and uncertainty of biologi- cal processes, fuzzy algorithm can be chosen as one of the most appropriate control scheme in the operation of biopro- cesses. A polyanionic heteropolysaccharide produced by Acinetobac- ter calcoaceticus RAG-l, emulsan, is an excellent bioemul- sifter which can be widely applied in many fields of oil-as- sociated and consumer product industries [Reisfeld et al., 1972; Rosenberg et al., 1979a,b, 1980; Kosaric et al., 1987]. Ethanol used mainly as the carbon source in A. calcoaceticus RAG-1 has an inhibitory effect on the cell growth of A. cal- coaceticus [Abbott, 1973; Abbott et al., 1973; Schaefer, 1985; Choi et al., 1996a]. Emulsan production is affected by the e- thanol concentration in culture medium since emulsan is pro- duced with mixed growth-associated pattern and the cell growth is directly dependent on the ethanol concentration [Gutnick et al., 1980; Schaefer, 1985; Choi et al., 1996a]. Fed-batch cul- tivation with continuous feeding of carbon source was per- formed as one of the most feasible operating strategy for the enhancement of emulsan production as well as cell growth by reducing the catabolite repression effect of ethanol [Choi et al., 1996a,c]. In this study, the fuzzy control algorithm was applied to the control of ethanol concentration in fed-batch cultivation ofA. calcoaceticu,~ RAG-1 for emulsan production. The mere-