Optimizing viscosity measurement: an automated solution with YOLOv3

Abstract

Abstract Measuring viscosity can be done using either Tracker software or a digital viscometer. However, the Tracker software proved to be ineffective due to the need to manually set the object’s center point for obtaining its final velocity. On the other hand, the digital viscometer was costly. Hence, a novel approach is needed to measure viscosity with high precision, efficiency, and affordability. To tackle these concerns, the study combined measuring instruments and computer programming with YOLOv3. The YOLOv3 model was applied to measure the B30 Biodiesel viscosity in a falling ball viscometer. The stages were: using a 10.07 mm ball size, the YOLOv3 model tracked the iron ball in the experimental videos to obtain the velocity. Next, based on the velocity, B30 biodiesel viscosity can be obtained. The formula of velocity and viscosity of the falling ball viscometer is integrated into the algorithm. The results are then compared to the reference data. The result showed that the final velocity and viscosity error relative were 1.30% and 2.04%. With an error relative below 5%, The data indicates that the algorithm effectively measures the velocity and viscosity of B30 biodiesel. This study was provided as a foundation for automatization in the quality control process for the biodiesel industry.