Holistic Simulation Model of the Temporal Degradation of Rolling Bearings

Data-driven diagnostic and prognostic methods for engineering systems, especially those employing machine learning, have gained prominence due to their reliance on data rather than physical system understanding.However, industrial applications often face challenges like unbalanced data distributions or limited data availability, as acquiring data is costly and time-intensive.Although some synthetic data sets and simulation models are publicly available, they often do not represent industry-relevant scenarios.Therefore, this work introduces a simulation model for generating representative run-to-failure data, focusing on rolling bearings.The model comprises three modules: the first determines the bearing life and fault type; the second simulates the degradation progression up to the point of failure; the third generates vibration signals reflecting operating conditions and bearing degradation.Each module is designed as a random process and reproduces the inherent variation of, for example, the life under a given load.As a novelty, the model simulates the vibration signals over the entire life of bearings.Furthermore, it is publicly available and can be used to generate arbitrary data.An initial data set is also published and publicly available.