Using fuzzy-based association rule mining to improve production systems for chemical product development

Abstract

In chemical product development, challenges lie in the determination of appropriate ingredients and parameter settings that lead to the desired product attributes. This relies heavily on the past knowledge and experience of the domain experts to generate feasible product candidates for verification. In this paper, a fuzzy-based association rule mining model (FbARM) is developed to provide knowledge support during chemical product development. Fuzzy-based association rule mining is applied to discover hidden relationships between parameters and the resultant product quality, followed by the use of fuzzy logic to generate recommendations on parameter settings. The feasibility of the FbARM is verified by means of a case study in a personal-care products manufacturing company. The results demonstrate the practical viability of the FbARM, while the learning ability of the FbARM allows a continuous improvement of the fuzzy rules, which is of paramount importance in responding to the changing requirements of the chemical industry.