(763d) Food Product Design: A Hybrid Machine Learning and Mechanistic Modeling Approach

Zhang, X., Hong Kong University of Science and Technology
Zhou, T., Max Planck Institute for Dynamics of Complex Technical Systems
Zhang, L., Dalian University of Technology
Fung, K. Y., Hong Kong University of Science and Technology
Ng, K. M., Hong Kong University of Science and Technology
A food product is often designed by trial and error based on experience. In this study, a hybrid machine learning and mechanistic modeling approach, formulated as a grey-box optimization problem, is proposed to expedite new food product design. The objective is to maximize food product quality quantified by sensorial ratings, which is predicted using machine learning models. The requirements on food property and structure are represented as design constraints which are decided using mechanistic or machine learning models. Moreover, a set of food ingredient candidates and the key operating conditions in food processing are identified based on heuristics, databases, etc. Design variables include the selection and composition of food ingredients as well as the key operating conditions. To solve the problem, genetic algorithm is utilized where constraints are handled as penalty functions. A chocolate chip cookie example is provided to illustrate the applicability of the hybrid modeling framework and solution strategy.