1992

Aarts, Robert

Bioprocesses are characterised by a high degree of uncertainty, incomplete knowledge and a lack of quantitative knowledge and abundant experience. Traditional, mathematical methods used for control and optimization of bioprocesses utilize only a small subset of the available knowledge and are therefore often rather uneffective. To investigate the possibilities of a knowledge-based approach to productivity improvement, knowledge-based systems for the diagnosis, control and planning of bioprocesses were constructed. An expert system shell with capabilities for real-time, on-line process monitoring and control was constructed. The necessary knowledge was represented by objects, fuzzy sets and rules for effective computer processing. The shell was used to develop expert systems for the diagnosis and control of glucoamylase production by Aspergillus niger and flat-bread extrusion cooking. Experience with these first systems led to the development of a simple formalism for improved representation of structural knowledge of the process. The performance of the system developed for glucoamylase production could be improved by the use of such structural knowledge. Development of bioprocesses can be viewed as a very complex planning, or constraint satisfaction, problem. An extensive and complex knowledge-base was shown to be useful in bioprocess development. Research on the planning of process parameters resulted in the construction of a system for the planning of mashing temperature profiles. This system uses fuzzy sets and rules to generate constraints. The tolerance propagation approach used for constraint processing was extended to use fuzzy sets. Laboratory test results indicated that worts obtained by mashing of malts using temperature profiles proposed by the planner, were indeed within the required specifications.