Oil Condition Monitoring and Remaining Life Prediction using Classification Learner Technique, an AI Application

This study uses an estimation and prediction method using characteristic techniques to predict the remaining life of the lubricating oil based on data collected directly from the hydraulic system of a mechanical processing machine. The data collected represent the measured values of the 19 lubricating oil condition parameters. The measurements were made online, on an experimental stand built and operated in conditions similar to those of a mechanical processing company. The data were recorded, for six months, in 258646 courts for all 19 operating parameters. To predict the values of the next steps in a sequence, the Classification Learner Technics has been approached by support vector machines (SVM) models. The generated output values characterize and equate the training sequences with values modified by a step of time, implying that the data structure learns to anticipate the output value at the next time step at each stage of the input sequence. To avoid the forecast from diverging, the training data has to be standardized to attain a zero mean and unit variance. In addition, the test data set was normalized in the same way that the training data was.