DigiFoodTwin: Digital Biophysical Twins Combined with Machine Learning for Optimizing Food Processing

Abstract

Production processes must allow high flexibility and adaptivity to ensure food supply. This includes reacting to disruptions in the supply of ingredients, as well as the varying quality of ingredients, e.g., seasonal fluctuations of raw material quality. Digital twins are known from Industry 4.0 as a method to model, simulate, and optimize processes. In this vision paper, we describe the concept of a digital food twin. Due to the variability of these raw materials, such a digital twin has to take into account not only the processing steps, but also the chemical, physical, or microbiological properties that change the food independent of the processing. We propose a model-based learning and reasoning loop, which is known from self-aware computing (SeAC) systems in the so-called learn–reason–action loop (LRA-M loop), for modeling the input for the LRA-M loop of food production, not as a pure knowledge database, but data that are generated by simulations of the bio-chemical and physical properties of food. This work presents a conceptual framework on how to include data provided by a digital food twin in a self-aware food processing system to respond to fluctuating raw material quality and to secure food supply and discusses the applicability of the concept.