Method of Operational Loading Process Augmentation Based on a Given Damage Criterion

An iterative method for augmenting operational loads to generate synthetic time series with a specified fatigue life is presented. This method addresses the problem of insufficient data for training machine learning models in structural life prediction problems. The method combines classical time series augmentation with an optimization algorithm, specifically modifying the signal shape or, if necessary, its duration. Validation on front-end loader frame stress data demonstrates that target damage is achieved with a 36-fold signal extension while preserving the key physical characteristics of the original process, such as distribution, frequency content, and statistical and spectral metrics.