ADAPTIVE CONTROL OF FEED LOAD CHANGES IN ALCOHOL FERMENTATION

ADAPTIVE CONTROL OF FEED LOAD CHANGES IN ALCOHOL FERMENTATION

R. Folly, R. Berlim, A. Salgado, R. França and B. Valdman

Escola de Química/UFRJ - Cidade Universitária, C.T. Bloco E sala 211 CEP- 21945-900,

Rio de Janeiro - Brazil

phone:[phone]/fax:[phone] - E-mail: [e-mail]

(Received: June 11, 1997; Accepted: October 30, 1997)

Abstract - A fed-batch alcohol fermentation on a pilot plant scale with a digital supervisory control system was evaluated as an experimental application case study of an adaptive controller. The verification of intrinsically dynamic variations in the characteristics of the fermentation, observed in previous work, showed the necessity of an adaptive control strategy for controller parameter tuning in order to adjust the changes in the specific rates of consumption, growth and product formation during the process. Satisfactory experimental results were obtained for set-point variations and sugar feed concentration load changes in the manipulated inlet flow to the fermenter.

Keywords: Adaptive control, alcohol fermentation, dynamic.

INTRODUCTION

Biochemical processes in general need some kind of control to maintain the operational conditions within the specific optimal ranges for each kind of process, microorganism and medium. In recent years, great emphasis has been placed on process control issues because of economic pressure to improve process productivity and yield. In spite of this, the control used in the majority of industrial plants is the daily and sequential stage control, such as sterilization, filling, fermenter discharge and the use of simple feedback loops for the direct or indirect control of some variables like temperature, pressure, flow, pH and breathing rate.

The application of control strategies in fermentation processes presents particular problems which make their implementation more difficult than that in chemical processes in general. The most difficult aspect is the lack of appropriate on-line analizers and sensors that permit direct concentration measurements in the fermenter. Another important factor in differentiating these processes is the large number of reactions and interactive transport phenomena caused by the presence of microorganisms. Microorganisms have a complex regulatory mechanism in their own cells and the generally employed control systems manipulate only the extracellular surroundings in an effort to suitably affect the intracellular mechanisms.

The implementation of process control for bioreactors requires an additional design effort in order to compensate for the inherent and nonlinear dynamic characteristics that suffer constant changes during the process. Conventional feedback controllers with constant PID parameters, usually employed to control industrial processes, can not be directly applied to provide good regulation of performance when the physical and biochemical properties are constantly changing. ln these cases, the adaptive control algorithm can better satisfy the variations once the PID parameters are adjusted during fermentation in an attempt to maintain optimum process performance, reducing the costs of the industrial production.

Many studies with adaptive controller applications have been published in the specialized literature (Chen et al., 1991 Vigie et al., 1990), although most of them refer to theoretical aspects or simulated process analysis and control strategies.

This work presents an experimental application of an adaptive control strategy which permits the regulation of broth density in a fed-batch alcohol fermentation by substrate feed flow manipulation, using an industrial PID adaptive controller designed with adaptive rules and tuning algorithm for load and density set-point variations. This same controller was partially applied in previous work on the development stage of the adaptive algorithm (Valdman et al., 1992) and in the proposal of an adaptive control loop strategy for the fermentation (Folly et al., 1994).

EXPERIMENTAL WORK

The experimental investigation done in this work was performed in the Alcohol Pilot Plant of the Instrumentation and Control Laboratory of the Escola de Química, UFRJ (LIC-EQ). The pilot plant contains four units: the fermentation sector, the distillation unit, the utilities sector and the control room, and a general view of the plant can be seen in Figure 1.

The fermentation sector is composed of a molasses storage tank, a series of tanks for molasses preparation and a dilution system. The fermenter is made of glass fibre with a total volume of 210 liters. ln one of the reaction vessels where the fermentation takes place, measurement and actuating instruments, including a continuous density transmitter, level and pH transmitters, a thermocouple transmitter and a control valve that manipulates the feed inlet to the reactor, are installed. These instruments are connected to the industrial controllers in the control room.

The control room, shown in Figure 2, is equipped with four multiloop controllers model CS-500, SMAR, two multiloop adaptive version controllers CS-600, SMAR, and two programmable logical controllers from ALTUS and MAXITEC. The controllers are interfaced with a microcomputer, where the supervisory system SMARCON 2.0 is installed. This permits the supervision and control of the operational variables, using preestablished monitor screens, from the measuring transmitters and actuating valves installed in the different sectors of the pilot plant and connected to the controllers in the control room.

Figure 1: General víew of the alcohol pilot plant-LIC/EQ.

Figure 2: Control room.

A set of experimental runs was achieved to verify the efficiency of the adaptive controller to load and set point changes for the control loop of the fed-batch alcohol fermentation. The strategy of adaptive control was used in order to adjust the fermented broth density by the manipulation of the inlet feed to the fermenter. To do so, a continuous density transmitter was used which permits a continuous monitoring of the broth density in the range of 0-13^o Brix, developed by the group in a previous study (Valdman et al,, 1993). The adaptive controller used was the adaptive version of the CD-600 controller present in the control room of LIC/EQ.

The complete operating process and all the equipment used in this work, along with the measurement and control instrumentation, is represented in the diagram shown in Figure 3.

A set of fermentations was achieved in order to verify the adaptive PI controller performance. ln the experimental assays, operational conditions similar to those in industry were reproduced. During fermentations, and using diluted molasses as the substrate and bread yeast as the agent, disturbances were introduced to the set point of the fermented medium density through the controller and a load disturbance in the molasses density in the feed flow. The operational conditions used are shown in Table 1. The initial tuning parameters of the adaptive controller for Case I were obtained from simulations of the process and adaptive control closed-loop studies obtained in previous work (Folly et al., 1994). For Case 2, the optimal values obtained from Case 1 were used as the initial tuning. The initial values of the controller parameters are shown in Table 2.

Figure 3: Process and instrumentation diagram.

FERMENTATION OPERATIONAL CONDITIONS Case 1 Case 2 Initial Volume 40 liters 40 liters Final Volume 160 liters 160 liters Initial Ethanol Concentration 0 g/l 0 g/l Initial Cell Concentration 40 g/l 40 g/l Feed Density 18 ° Brix 11 ° Brix Temperature * 36 ° C 36 ° C

FERMENTATION OPERATIONAL CONDITIONS
		Case 1		Case 2
Initial Volume		40 liters		40 liters
Final Volume	160 liters		160 liters
Initial Ethanol Concentration	0 g/l		0 g/l
Initial Cell Concentration	40 g/l		40 g/l
Feed Density	18 ° Brix		11 ° Brix
Temperature *	36 ° C		36 ° C

Table 1: Operational conditions of the experimental process

RESULTS

Figures 4a, ^b, ^c, ^d show the experimental results obtained in Case 1 fermentation which was performed in accordance with the conditions shown in Tables 1 and 2. During the fermentation, two consecutive step disturbances were performed at set point values from 4.0 to 5.5 and from 5.5 to 6.5, and another step disturbance in the molasses feed density in order to verify the process variable response with the adaptive PI controller. ^{Figures 4b} and ^4c show the volume and the controller output signal variation, respectively. ^{Figure 4d} shows the ethanol and the total reduced sugar concentration profiles during the fermentation obtained by laboratory analyses.

The fermentation of Case 2 was operated in accordance with the initial conditions shown in Tables 1 and 2 with two consecutive step disturbances from 4.0 to 5.4 and from 5.4 to 6.4 values at set point and one positive load step disturbance, changing the feed density from 11 to 18^o BRIX. Figure 5 a, ^b, ^c, ^d shows the experimental results obtained with this fermentation.

CONTROLLER PARAMETERS Case 1 Case 2 Proportional Gain - Kp 6,0 9,0 Integral Time- Ti 4,8 2,2 Set Point - SP 4,0 4,0

CONTROLLER PARAMETERS
		Case 1		Case 2
Proportional Gain - Kp		6,0		9,0
Integral Time- Ti	4,8		2,2
Set Point - SP	4,0		4,0

Table 2: Initial tuning parameters of the controller

Figure 4a: Density response.

Figure4b: Controlled output variable response.

Figure 4c: Volume profile.

Figure 4d: Fermenter concentration response.

Figure 4: Experimental results of process variables with the adaptive PI controller - Case 1.

Figure 5a: Density response.

Figure 5b: Controlled output variable response.

Figure 5c: Volume profile.

Figure 5d: Fermenter concentration response.

Figure 5: Experimental results of process variables with the adaptive PI controller - Case 2.

CONCLUSIONS

An adaptive control strategy was proposed and analyzed for the medium density control of a fed-batch alcohol fermentation. Experimental studies in a pilot plant operation showed satisfactory results in the behaviour and evaluation of the controller performance with the adaptive algorithm in response to disturbances in the set point and in the fermenter feed substrate concentration. The adaptation time in control parameters and in the controlled variable recuperation diminished considerably during the process, after its starting point in the automatic mode operation, reaching 20 minutes for a fermentation with an effective response time characteristic of 2.0 hours.

• Chen, L.; Bastin, G. and Breusegem, V., Adaptive Nonlinear Regulation of Fed-Batch Biological Reactors: An Industrial Application, Tec. Rep. 91, 25, Cesame, Belgique (1991).
• Folly, R.O.M.; Ramirez, N.B. and Valdman, B., Process Productivity Increase by lntegrated Process Operation and Control Design, Proc, PSE'94, 5th lntem. Symp. on Process Systems Engin., p. 717, Korea (1994).
• Valdman, B.; Folly, R.0.M.; Arias, M.Z. and Ramirez, N.B., Use of Continuous Density Transmitter for Inferential Control of a Fed-Batch Alcoholic Fermentation Process, Proc. 11^o lnternational Congress of Chemical Engineering, D9-72, T1 85, p. 1, Prague (1993).
• Valdman, B.; Ramirez, N.B. and Valdetaro, E.D., Controle Adaptativo de uma Fermentaçăo Alcoólica em Batelada Alimentada, INSTEC - lnstrum. e Controle de Proc, VI, 58, P.93 (1992).
• Vigie, P.; Goma, G. and Renauld, P.Y., Adaptative Predictive Control of Multistage Fermentation Process, Biotechnol. and Bioeng., 35, p.217 (1990).

Publication Dates

• Publication in this collection
  06 Oct 1998
• Date of issue
  Dec 1997

History

• Accepted
  30 Oct 1997
• Received
  11 June 1997