Data-based multivariate modeling of a grain comminution process

A grain comminution process is based on a gradual size reduction approach, through repeated milling and sieving units that increase the flour yield. In this study we use latent variable modeling techniques to link process parameters and grain properties to the final product quality. Using experimental data, it is shown how the use of models in their direct form allows one to improve process understanding and to predict the product quality from the process settings and grain properties. Additionally, it is shown that, by inverting the latent variable models, the optimal combination of process parameters and grain properties leading to a desired product quality can be determined.