Abstract
The milling industry envisions solutions to become fully compatible with the industry 4.0 technology where sensors interconnect devices, machines and processes. In this contest, the work presents an integrated solution merging a deeper understanding and control of the process due to real-time data collection by MicroNIR sensors (VIAVI, Santa Rosa, CA)-directly from the manufacturing process-and data analysis by Chemometrics. To the aim the sensors were positioned at wheat cleaning and at the flour blends phase and near infrared spectra (951-1608 nm) were collected online. Regression models were developed merging the spectra information with the results obtained by reference analyses, i.e., chemical composition and rheological properties of dough by Farinograph(®) (Brabender GmbH and Co., Duisburg, Germany), Alveograph(®) (Chopin, NG Villeneuve-la-Garenne Cedex, France) and Extensograph(®).(Brabender GmbH and Co., Duisburg, Germany) The model performance was tested by an external dataset obtaining, for most of the parameters, R(PRED) higher than 0.80 and Root Mean Squares Errors in prediction lower than two-fold the value of the reference method errors. The real-time implementation resulted in optimal (100% of samples) or really good (99.9%-80% of samples) prediction ability. The proposed work succeeded in the implementation of a process analytical approach with Industrial Internet of Things near infrared (IIoT NIR) devices for the prediction of relevant grain and flour characteristics of common wheat at the industrial level.