Font Size: 

This study explores the application of principal component analysis and singular spectrum analysis for the purposes of fault identification in rolling element bearings. Rolling element bearings are one of the most frequently used ones in a lot of machinery parts. They suffer damage due to different reasons and there are different fault types that can appear depending on their location and also on the type of damage- e.g distributed and concentrated (crack). These faults usually affect the vibration signature measured on the machine and there are specific frequencies which have been identified characteristic for different fault types. Unfortunately these faults are high frequency faults which possess rather little energy and as such are difficult to detect and identify from the raw signal spectrum. Thus a number of methods have been developed for the purposes of rolling element bearing (REB) fault detection and identification. Principal component analysis (PCA) is a data analysis method that can be used to reduce the data dimension and for the case of categorical data it can be used to bring the data from the same category closer while in the same time increasing the distance between data from different categories. In this study PCA is used for analysis of signatures from different categories of REB , namely healthy ones and bearings with different fault types and dimensions. The raw signals are subjected to pre-treatment and subsequently to PCA which transforms the signals into new variables just several of which can be used for the further identification process. This research explores the number of components to be used in order to achieve better identification and their usefulness and application as factors in the further process.