Limitations of RUL estimation methods

The accuracy of remaining useful life (RUL) estimates and probabilistic failure predictions depends on a number of determining factors relating to the data, the methods used and the operating context. Firstly, data quality plays a fundamental role in the reliability of predictions. Secondly, modeling and the choice of algorithms directly influence the accuracy of estimates. The adoption of physical, statistical or artificial intelligence-based models must be adapted to the volume and diversity of the data available. Machine learning approaches, although powerful, require a representative learning base to avoid bias and guarantee reliable predictions.

Furthermore, the temporal evolution of systems requires regular updating of models to avoid obsolescence. On-line learning enables predictions to be adjusted in line with new data and changing operating conditions....