Linking human and machine - towards consumer-driven automated manufacturing

In this paper we establish a link between linguistic descriptors describing food preferences and product manufacturing processes. We show how this is achieved using a model-based methodology that translates consumer preferences into product and process specifications. The ultimate goal is the large scale personalization of formulated food product manufacture where consumers are also the co-creators of the food products they wish to buy. Firstly, we investigate how those sensory attributes for such products can map onto product and process specifications. Fuzzy set modelling is used to capture the preferences and perception for these attributes by different groups of people. Specifically, type-1 fuzzy sets are generated from interval-valued survey data for the linguistic descriptors (i.e., thin, thick, smooth, pulpy) and the sensory indicators (i.e., smoothness, roughness and orange flavor) that describe how consumers perceive and select orange based beverages. Then, the models are employed to establish the links between such product attributes and the actual formulation parameters to make the product. We demonstrate the manufacture of the desired orange beverage that emerged from the modelling approach by deploying the process parameters, which map onto those descriptors, on the controller of a continuous food formulation system which was selected due to its flexibility and its computer controller that provides the ability to redeploy new formulation specifications rapidly. With this overall methodology we demonstrate for the first time the digital, on-demand manufacture of soft beverages with targeted attributes, selected directly by a consumer group.