Indirect approach for measurement of main physico-chemical parameters of engine oils using optical methods

The article presents an indirect approach for measuring two main physicochemical parameters – kinematic viscosity and total alkalinity of motor oils based on digital analysis of oil images and regression models containing the color features of their images, which were developed based on reference values obtained from laboratory analysis. Two types of oils with viscosity class – XW-30 and XW-40 – were analyzed. Ten regression models were developed, which were estimated using the two errors RMSE and MAE. The smallest average values of RMSE and MAE are obtained when the images are captured under homogeneous illumination without external light. For viscosity measurement with the highest R 2 , the regression models are linear SVM and fine tree, and for total alkalinity – stepwise regression and fine tree.