Estudo de caso sobre avaliação da condição de óleos lubrificantes em uso mediante modelo de regressão logística

Abstract

The lubricant is a material inserted between the parts of the surface that have relative movement to each other in order to reduce the friction between them. Lubricating oils are mostly used in mechanical systems because they have greater penetration power and are able to perform secondary functions, such as the dissipation of heat that arises from the movement between parts. It has become a common practice to monitor the quality of lubricating oil in use through oil analysis. However, interpretation of the reports is tough, requiring the person in charge of this task to have considerable empirical knowledge about the lubricating oil analyzed and its application. Therefore, the present work aims to propose a framing model of the condition of oil in use, applying the statistical technique of logistic regression in a real database. We chose Logistic regression due to the characteristics of the problem in question: the variable of interest was binary and dependent on a set of other variables. The model proposed in this work presented an excellent success rate, a value close to 95.5%, and used only four explanatory variables. Our results showed that for the classification of lubricating oil, some properties are more important than others, including viscosity, the level of acidity, and the number of particles present in the analyzed fluid.