An auto-regulating expert system strategy for dynamic intelligent system selection for on-line optimization of a bioprocess

Abstract

Mathematical models derived through mechanistic information and validated under laboratory conditions do not always portray satisfactorily the behavior of complex microbial processes under realistic conditions. Models based on artificial intelligence (AI) then offer a viable alternative. However, AI models also have limitations, and sometimes different AI methods may be the most effective at different points in time or in different regions of the operating space. The present communication, therefore, presents a supervisory expert system that receives performance data continually and selects that AI module from an on-line library, which maximizes a specified performance index. This ensures that the most efficient AI system is functional at all times. The concept is presently being applied to glucoamylase production in continuous cultivation of a recombinant strain of <i style="mso-bidi-font-style:normal">Saccharomyces cerevisiae</i>, and initial results support its feasibility and effectiveness.