This is the authors' abstract. We don't add key points for this paper.
The growing demand for efficiency in production chains has centered an increasing search to enhance the reliability of complex systems in industry. Rotating machines are crucial in generating and converting energy, which is essential for manufacturing. With the emergence of artificial intelligence, several techniques for qualitative diagnosing machines have been successfully applied. However, some issues, such as the models' lack of explainability, interpretability, and the determination of fault severity, still need to be addressed to allow their implementation in high-risk situations. This work proposes improving the monitoring of rotor systems through a diagnostic method that quantifies fault severity through parameters (fault parameters). This procedure favors estimating the remaining operation lifetime and even indicates the need and instructions for maintenance procedures. The proposed method uses surrogate models allied with epistemic uncertainty quantification in interpretable and explainable models. Epistemic uncertainty is useful to enhance the models' diagnostic trustworthiness. To this end, Bayesian deep operator networks were used to create surrogate models to quantify fault severity in rotating machinery using a direct inference algorithm in unbalance situations combined with coupling misalignment. A two-step search algorithm was used to quantify fault parameters using Bayesian optimization and a gradient-based algorithm to overcome problems regarding nonunique solutions and nonconvex optimization. As a result, it is possible to map the error levels of the models by quantifying the uncertainty and enhancing their trustworthiness. The surrogate models showed less than 1.2% error and demonstrated a 13-fold reduction in computational cost.
Answers are grounded in this paper and cite their sources.
No takes yet. Share an insight, caveat, or question.
Belli et al. (2025) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: