View in EDS

Enhancing non-stationary rotodynamic monitoring through current signature analysis.

Saved in:
Bibliographic Details
Title: Enhancing non-stationary rotodynamic monitoring through current signature analysis.
Authors: Cruz-Rangel, David1,2 (AUTHOR) david.antonio.cruz@upc.edu, Ocampo-Martinez, Carlos1 (AUTHOR) carlos.ocampo@upc.edu, Diaz-Rozo, Javier2 (AUTHOR) jdiaz@ainguraiiot.com
Source: Expert Systems with Applications. May2026, Vol. 310, pN.PAG-N.PAG. 1p.
Subjects: Signal frequency estimation, Fault diagnosis, Rotating machinery
Abstract: • A novel online NILM methodology based on MCSA is proposed for non-stationary industrial environments. • The proposed method was validated in multiple industrial testbed scenarios. • A PSD estimator tailored for non-stationary signals outperforms ANF and CEEMDAN in frequency tracking. • Superior performance is achieved in key metrics such as monotonicity and robustness for fault progression monitoring. • The method enables reliable early fault detection and feature-level monitoring under variable-speed operating conditions. Early detection and primary monitoring of rotodynamic degradation in industrial machinery require robust methodologies capable of accurately analyzing non-stationary signals, particularly under variable-speed conditions where conventional techniques often fail. This study introduces a novel online Non-Intrusive Load Monitoring (NILM) methodology based on advanced Motor Current Signature Analysis (MCSA), specifically designed to improve frequency estimation and fault progression tracking in rotating machinery operating under variable-speed conditions. The proposed approach integrates a power spectral density (PSD) estimation algorithm optimized for non-stationary environments and is benchmarked against the Adaptive Notch Filter (ANF), Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN), and conventional vibration-based techniques. The methodology operates at the level of degradation-related feature extraction and early fault indication, without developing explicit prognostic models or decision-level maintenance strategies. In particular, remaining useful life (RUL) estimation, health index-based prognostics, and maintenance optimization frameworks are outside the scope of the present study. The methodology is validated using two accelerated bearing run-to-failure experiments conducted on a single industrial testbench, incorporating variable-speed profiles and progressive bearing degradation. Cross-asset, cross-plant, and large-scale industrial generalization are not addressed in this work and remain topics for future investigation. Results demonstrate superior accuracy in frequency tracking, achieving significantly lower mean squared error (MSE) compared to ANF- and CEEMDAN-based techniques. Moreover, the extracted features exhibit greater sensitivity to early fault development, while monotonicity and robustness metrics confirm the reliability of the proposed methodology for fault progression monitoring within the considered experimental scenarios. [ABSTRACT FROM AUTHOR]
Copyright of Expert Systems with Applications is the property of Pergamon Press - An Imprint of Elsevier Science and its content may not be copied or emailed to multiple sites without the copyright holder's express written permission. Additionally, content may not be used with any artificial intelligence tools or machine learning technologies. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)
Database: Engineering Source
Description
Abstract:• A novel online NILM methodology based on MCSA is proposed for non-stationary industrial environments. • The proposed method was validated in multiple industrial testbed scenarios. • A PSD estimator tailored for non-stationary signals outperforms ANF and CEEMDAN in frequency tracking. • Superior performance is achieved in key metrics such as monotonicity and robustness for fault progression monitoring. • The method enables reliable early fault detection and feature-level monitoring under variable-speed operating conditions. Early detection and primary monitoring of rotodynamic degradation in industrial machinery require robust methodologies capable of accurately analyzing non-stationary signals, particularly under variable-speed conditions where conventional techniques often fail. This study introduces a novel online Non-Intrusive Load Monitoring (NILM) methodology based on advanced Motor Current Signature Analysis (MCSA), specifically designed to improve frequency estimation and fault progression tracking in rotating machinery operating under variable-speed conditions. The proposed approach integrates a power spectral density (PSD) estimation algorithm optimized for non-stationary environments and is benchmarked against the Adaptive Notch Filter (ANF), Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN), and conventional vibration-based techniques. The methodology operates at the level of degradation-related feature extraction and early fault indication, without developing explicit prognostic models or decision-level maintenance strategies. In particular, remaining useful life (RUL) estimation, health index-based prognostics, and maintenance optimization frameworks are outside the scope of the present study. The methodology is validated using two accelerated bearing run-to-failure experiments conducted on a single industrial testbench, incorporating variable-speed profiles and progressive bearing degradation. Cross-asset, cross-plant, and large-scale industrial generalization are not addressed in this work and remain topics for future investigation. Results demonstrate superior accuracy in frequency tracking, achieving significantly lower mean squared error (MSE) compared to ANF- and CEEMDAN-based techniques. Moreover, the extracted features exhibit greater sensitivity to early fault development, while monotonicity and robustness metrics confirm the reliability of the proposed methodology for fault progression monitoring within the considered experimental scenarios. [ABSTRACT FROM AUTHOR]
ISSN:09574174
DOI:10.1016/j.eswa.2026.131258