Food model exploration through evolutionary optimisation coupled with visualisation: Application to the prediction of a milk gel structure

• We propose a methodology for the analysis of complex food models.
• The methodology couples evolutionary exploration of the model and visualisation.
• Case study is an established model of milk gel competition at oil–water interface.
• We find visual evidence for a correlation between parameters of the model.
• Visual evidence is backed up by theoretical analysis; model complexity is reduced.

Obtaining reliable in-silico food models is fundamental for a better understanding of these systems. The complex phenomena involved in these real-world processes reflect in the intricate structure of models, so that thoroughly exploring their behaviour and, for example, finding meaningful correlations between variables, become a relevant challenge for the experts. In this paper, we present a methodology based on visualisation and evolutionary computation to assist experts during model exploration. The proposed approach is tested on an established model of milk gel structures, and we show how experts are eventually able to find a correlation between two parameters, previously considered independent. Reverse-engineering the final outcome, the emergence of such a pattern is proved by physical laws underlying the oil–water interface colonisation. It is interesting to notice that, while the present work is focused on milk gel modelling, the proposed methodology can be straightforwardly generalised to other complex physical phenomena.Industrial relevanceSustainability is nowadays at the heart of industrial requirements. The development of mathematical approaches should facilitate common approaches to risk/benefit assessment and nutritional quality in food research and industry. These models will enhance knowledge on process–structure–property relationships from the molecular to macroscopic level, and facilitate the creation of in-silico simulators with functional and nutritional properties. The stochastic optimisation techniques (evolutionary algorithms) employed in these works allow the users to thoroughly explore the systems: when coupled with visualisation, they make it possible to provide the experts with a restricted set of significant data, helping them to highlight eventual issues or possible improvements in the model. With regard to the complexity of the food systems and dynamics, the challenge of the mathematical approaches is to realise a complete dynamic description of food processing. In order to reach this objective, it is mandatory to use innovative strategies, exploiting the most recent advances in cognitive and complex system sciences.