State Identification and Control Strategy in Penicillin Fermentation

Abstract

The most existing industrial fermentation processes of antibiotics are controlled manually and empirically, which results in relatively low and fluctuated fermentation levels. For increasing the fermentation level, on-line optimal computer control is a very important technique. In this work a new technique based on process modelling and computer control has been developed for industrial penicillin fermentation. As a basis of computer control three models should be established beforehand monitor model - to relate the on - line measurable variables to the biological process variables; kinetic model - to decide the consumption rate of nutrient components and the formation rates of biomass and secondary metabolite product; and control model - to determine the control variables. A lab-scale computer coupled fermentation system (5L) was developed, which can on-line monitor pH, dissolved oxygen level (D. O), temperature, aeration rate, agitation speed, antifoam, pressure, concentrations of oxygen and carbon dioxide in effluent, and can calculate up to 12 variables in real time, including the carbon dioxide production rate [CPR], oxygen uptake rate (OUR), respiratory quotient (R.Q), oxygen transfer coefficient (ka) etc. A feeding device was controlled by this computer system, which can control the feed-rates of sugar, nitrogen source and precursor according to a built-in model. This system has been employed to simulate an industrial penicillin fermentation process. Through investigation of the time courses it was found that the variation of carbondioxide production rate with time has a close relation to the fermentation states, which are characterized as fast growing phase, transient phase and production phase. Referring to the variation of pH, the fermentation stages can be determined from the changing rate of [CPR]. During the production phase the level of sugar should be controlled strictly. It was first found that there exists a proportional relation between the total amount of sugar consumed and the total amount of carbondioxide released, the slope of the regression line being 1.55 g sugar/n-L CO2. Consequently, the feeding rate of sugar can be estimated by the on-line measured carbon dioxide production rate. A pseudohomogeneous kinetic model was set up. The relation among sugar consumption rate, nitrogen source consumption rate and product accumulation rate was described by this model. The feeding rates of sugar, nitrogen source and precursor in different fermentation stages can thus be resolved. By this technique based on the above models the fermentation level was increased from 30,000 u/mL to 38,000 u/mL in lab-scale study, the best result was 45,200 u/mL. In a pharmaceutical factory a two stage computer system was set up, controlled by the same model, the fermentation level in a 100 M3 fermentor was increased by 25%.