The use of near infrared hyperspectral imaging for the prediction of processing parameters associated with the pelleting of biomass feedstocks

• Hyperspectral imaging was used to predict processing parameters of a pellet mill.
• Moisture content was accurately predicted with an R2 value of 0.94.
• Specific energy and feed rate models allow approximate predictions to be made.
• Image analysis ensures the efficient mixing of biomass feedstocks is maintained.

Near infrared hyperspectral imaging combined with chemometrics was used to assess the potential for the prediction of the moisture content, specific energy and the feed rate of the feedstock into the pellet die. Samples were produced from a diverse set of agricultural products and wood chips, with a range of moisture contents. Image analysis was also utilised to assess the efficient mixing of biomass feedstocks prior to pelleting in a multi biomass stream. The moisture content (%), specific energy (kWh kg− 1) and feed rate (kg min− 1) were predicted with root mean square errors of prediction of cross validation of 1.11% (R2 = 0.94), 0.12 kWh kg− 1 (R2 = 0.64) and 0.20 kg min− 1 (R2 = 0.70), respectively. The results of this study indicate that near infrared hyperspectral imaging has the potential to be incorporated into a biomass pelleting facility to improve the efficiency of the system.