Search

Industrial fed-batch yeast fermentation process is a typical nonlinear dynamic process that requires good control technique and monitoring to optimize the yeast production. This chapter explores the applicability of Q-learning in determining the feed flow rate in a fed-batch yeast fermentation process to achieve multiobjectives optimization. However, to develop such control system, the complex nature of the yeast metabolism that will affect the system stability has to be considered. Q-learning is well known for its interactive properties with the process environment and is suitable for the learning of system dynamic. Therefore, the utilization and performance of Q-learning to seek for the optimal gain for the controller is studied in this chapter. Meanwhile, the performance of Q-learning under the process disturbance is also tested.

The aim of this chapter is to optimise the productivity of an exothermic batch process, by maximising the production of the desired product while minimising the undesired by-product. During the process, heat is liberated when the reactants are mixed together. The exothermic behaviour causes the reaction to become unstable and consequently poses safety issues. In the industries, a dual-mode controller is used to control the process temperature according to a predetermined optimal reference temperature profile. However, the predetermined optimal reference profile is not able to limit the production of the undesired by-product. Hence, this work proposed a genetic-algorithm-based controller to optimise the batch productivity without referring to any optimal reference profile. From the simulation results, the proposed algorithm is able to improve the production of the desired product and reduce the production of the undesired by-product by 15.3 and 34.4 %, respectively. As a conclusion, the genetic-algorithm-based optimisation performs better in raw materials utilisation as compared to the predetermined optimal temperature profile method.