Bearing Remaining Useful Life Prediction Using Feature Extraction

The patent extracts only the significant features from the hundreds of columns in condition monitoring data. It uses correlation analysis and feature selection techniques to identify and extract the most relevant features for RUL prediction, discarding noise and irrelevant information. This extraction process transforms the raw high-dimensional data into a concentrated set of meaningful features that can be effectively used for prediction.

The patent performs preliminary action by predicting the remaining useful life of equipment before actual failure occurs. It uses condition monitoring data and machine learning models to forecast when equipment will fail, enabling maintenance to be scheduled just in time. This preliminary prediction allows operators to replace equipment only when necessary, avoiding both premature replacement and unexpected failures.

The patent uses copying by creating simplified representations of the complex data through feature extraction. Instead of directly processing hundreds of raw data columns, it creates a copy in the form of selected significant features that capture the essential information. This copied feature set can be processed by less complex models while maintaining prediction accuracy, reducing computational burden.