Oil condition monitoring and predicting actions using an Artificial Intelligence technique: Principal Components Analysis algorithm

Abstract

Like other lubricants, oils play a crucial role in providing the space needed to separate moving parts. In addition, the lubricants ensure the suspension and transport of contaminants, neutralize corrosive acids, protect surfaces likely to wear, ensure heat dissipation, and provide the performance and increasing performance characteristics of industrial equipment and beyond. During use, lubricating oils undergo several chemical transformations due to oxidation at high temperatures (regime temperatures) due to degradation and contamination, water, ethylene glycol, coolants, waste residues. Finally, oils reach a lifetime due to a wide variety of degradation mechanisms, which lead to increased oxidation and nitration, base depletion, acid build-up, water contamination, cooling fluids, and viscosity changes. The rather complex nature of lubricants, along with the distinct variety of industrial equipment, especially the latest generation, equipped with high-performance techniques and artificial intelligence, make it very difficult, if not impossible, to predict all possibilities of generating defects. This study was intended to show how to expect action-time series data using Artificial Intelligence techniques on a set of data collected using direct/indirect sensors and computational determinations based on empirical relationships. The algorithm Principal Components Analysis (PCA) has been approached to predict the values of the next steps of any sequence by support vector machines (SVM) models. The PCA approach was considered a favourable experiment. The answers obtained characterize and equate the training sequences with values changed by a step of the time. This means that the data structure learns to predict the next step’s output value at each stage of the input sequence.