Control of a Food Process Based on Sensory Evaluations

Sensory evaluation is often the ultimate measure of food quality, but food process control relies on instrumental measurements. In this research, a strategy for using sensory evaluations in food process control was explored on the basis of a rice cake production process. Important sensory attributes were first modeled as functions of instrumental variables (manipulated or instrumentally measurable variables) to determine the potential controllability of the sensory variables. Neural networks were used to convert sensory quality targets into process control set points. Trained neural networks effectively predicted the instrumental measurements or process set points that resulted in certain sensory qualities. The neural networks were used to determine the best process set points for a given set of sensory targets, and the set point values were used to produce new product samples. The sensory attributes of the new rice cakes were close to those defined as the targets. The results demonstrated the effectiveness and potential of the method.