Cobot-assisted capacitive sensing for in-process, micrometre-scale thickness estimation of liquid films

In-process measurement of coating and liquid film thickness on conductive and non-conductive substrates is essential for quality assurance in industries such as automotive painting, pharmaceutical coating, and consumer product finishing. However, existing inspection methods face significant challenges: first, vision-based and statistical approaches are highly sensitive to environmental conditions and material reflectivity, resulting in unstable measurements. Second, current capacitive or contact-based techniques require static setups or pre-calibrated dielectric properties, limiting adaptability during dynamic spreading. Third, inadequate spatial resolution prevents early detection of localised defects. These limitations often lead to undetected non-uniformities such as thin spreading, centred spreading, or empty spreading, causing material waste, costly rework, and product assessment failures. To address these issues, this paper proposes a robot-assisted capacitive sensing framework for non-destructive, in-process thickness estimation on mimic substrates. A 6-DoF collaborative robot (UR10e) executes structured multi-point scanning, while a capacitive sensor captures dielectric responses correlated with micrometre-scale layer thickness. A calibrated linear regression model maps sensor signals to thickness values, enabling detection of thin films, pooling, and voids in-process. Experimental validation across four target thickness levels (15–90 µm) under uncontrolled conditions achieved an R² of 0.915 and a mean squared error of 0.00699. The proposed system provides a Cobot-assisted, interpretable, automated sensing solution suitable for in-process quality control of liquid films in industrial spreading, painting, and coating applications.