Knowledge-Based Systems for Formulation Optimization, 1995 Hot Melt Symposium Proceedings

William H. VerDuin

AI WARE Incorporated

Customers require faster delivery of increasingly customized products. New environmental constraints must be met. Producers of formulated products must design better products faster. The challenge is to meet these requirements without additional staffing or capital costs.

New computer technologies called knowledge-based systems address this need, enabling computers to be more effective in roles beyond their traditional strengths. Some knowledge- based systems capture problem-solving knowledge acquired from experts. Others discover relationships within data.

This knowledge may be used, for example, to predict the effect of formulation changes on properties and costs, or the effect of process changes on performance and cost. Knowledge-based system modeling capabilities surpass alternatives such as statistics in data acquisition requirements and self-improvement capabilities. Models capturing both formulation and processing provide a broader view of design and processing opportunities and costs. The most advanced knowledge-based systems incorporate sophisticated optimization and sensitivity analysis capabilities, enabling users to find product formulations and processing conditions that best meet multiple ranked objectives.

This paper describes applications of knowledge-based design methods for formulation development and process optimization. Benefits, including reduced experimentation, reduced reliance on analytical skills, and cost reductions without degradation of product properties are described.