Moisture content estimation in ceramic tiles and CaCO3-based feed materials using surface hyperspectral imaging

Drying is a crucial processing procedure in the production of a plethora of materials, including ceramic tiles and CaCO3-based animal feed. To achieve a more accurate production process, methods to assess material Moisture Content (MC) in these scenarios are essential as they can support the development of closed-loop controls that aim at process optimization. Non-contact measurement solutions represent valuable options to guarantee process monitoring flexibility and avoid material alteration as well as process interruption. To compensate for the lack of off-the-shelf solutions, a hyperspectral imaging (HSI) based method is proposed in this paper. Full size imaging and Digital Image Correlation region matching approaches are utilized to reduce measurement uncertainties by direct single-point reflection and to meet the requirements of industrial process monitoring. The relationships between the Reflection Rate (RR) ratio and the material MC are found in multiple series of tests. A consistent inverse relationship between the RR ratio at 1490 nm/1050 nm and surface moisture content was observed for both applications, in the investigated moisture ranges. Comparable trends were obtained for ceramic tiles and CaCO3-based powders, despite differences in their physical structure such as pore size and pore-size distribution. Quantitatively, the root mean square error and mean absolute error between the reference and estimated MC values were 0.77% and 0.66%, respectively, for the ceramic-tile samples, and 0.47% and 0.33%, respectively, for the CaCO3 samples. These results indicate a strong and consistent relationship between the RR index and MC for both materials. The proposed approach demonstrates stable and repeatable moisture content estimation across different samples and hardware configurations, achieving surface moisture sensitivity suitable for industrial process monitoring and control-oriented applications.