Implementation of multiple linear regressions in lubricant degradation prediction algorithm

This paper presents the development of the prediction algorithm of lubricant degradation based on Beer Lambert's transmittance theory by using Multiple Linear Regressions (MLR). Recently, an increasing amount of wasted lubricant has been due to the unnecessary changing of lubricant even though the lubricant still remains its lubrication behavior. Therefore, a condition based technique is introduced to monitor the degradation parameters in lubricating oil by using optical approach. This work focuses on Total Acid Number (TAN) that has been identified as the main parameter in determining the lifetime of lubricant and it occurred at band location from 1,050-1,250cm
^-1
and 1,700-1,730cm
^-1
. The best input parameter has been identified for sensor development and signal processing. Then, the prediction model is used to validate the measured and the predicted value of degradation. The high correlation between the predicted and measured data shows the prediction algorithm can be used for prediction purposes efficiently.