Supervisory Expert System-Based Intelligent Optimization of a Microbioreactor

Abstract

Microbioreactors with immobilized yeast cells are conventionally packed uniformly. A recent study has shown, however, that a topologically optimized distribution of cells yields much greater outputs of the desired product. Because topology optimization is a complex method requiring a good mathematical model, artificial intelligence (AI) has been employed here as an alternative method. For the same system-in other words, immobilized genetically modified yeast cells-an expert system selected online the better of two AI methods-a fuzzy neural network (FNN) and a genetic algorithm (GA)-according to the output of the product recombinant glucoamylase. Progressing in short time intervals enables the expert system toshift continually between the FNN and the GA, thereby maintaining optimal performance at all times. This method is more robust than topology optimization, easier to implement, does not require a mathematical model, and improves glucoamylase output even further.