Parametric Generalized Adaptive Moment Features (PG-AMF) for Bearing Fault Diagnosis and Machine Health Monitoring

2026-06-24 · Source: Artificial Intelligence · Field: Manufacturing & Industrial — Smart Manufacturing & Industry 4.0, Manufacturing Operations & Management, Artificial Intelligence & Machine Learning · Depth: Intermediate, quick

Summary

Parametric Generalized Adaptive Moment Features (PG-AMF) is a novel framework designed for accurate fault diagnosis of rolling element bearings in rotating machinery, addressing limitations of conventional statistical methods and deep learning approaches. Unlike predefined descriptors, PG-AMF learns feature characteristics directly from vibration data, extracting complementary representations such as absolute features for signal energy, signed moment features for waveform asymmetry, and AC-coupled moment features for dynamic fluctuations. It also incorporates a structured fusion mechanism to model interactions across multiple sensor channels. Evaluated on a benchmark gearbox bearing dataset with five health conditions, PG-AMF demonstrated improved classification performance, strong generalization capability through cross-validation, and enhanced feature separability, making it practically applicable in industrial monitoring systems.

Key takeaway

For Machine Learning Engineers tasked with improving bearing fault diagnosis accuracy and interpretability, PG-AMF offers a robust, data-driven alternative. This approach bypasses the fixed configurations of traditional methods and the high data demands of deep learning, providing adaptive feature extraction and strong generalization. You should consider integrating PG-AMF to enhance the reliability and efficiency of your industrial machine health monitoring systems, particularly for critical rotating machinery.

Key insights

PG-AMF learns adaptive features from vibration data for superior bearing fault diagnosis, overcoming limitations of fixed descriptors and deep learning.

Principles

• Feature characteristics can be learned directly from data.
• Complementary representations enhance fault diagnosis.
• Structured fusion improves multi-sensor fault representation.

Method

PG-AMF extracts absolute, signed moment, and AC-coupled moment features from vibration signals. It then uses a structured fusion mechanism to model interactions between multiple sensor channels for enhanced fault representation.

In practice

• Apply PG-AMF for predictive maintenance systems.
• Use PG-AMF for real-time industrial monitoring.
• Improve bearing fault classification accuracy.

Topics

• Bearing Fault Diagnosis
• Machine Health Monitoring
• Predictive Maintenance
• Feature Extraction
• Vibration Analysis
• Adaptive Moment Features
• Sensor Fusion

Best for: Machine Learning Engineer, AI Scientist, Research Scientist

Related on AIssential

• A Multi-Plant Machine Learning Framework for Emission Prediction, Forecasting, and Control in Cement Manufacturing — Data-driven ML models can significantly reduce cement plant NOx emissions and costs without hardware changes.
• Towards Unified and Data-Efficient Prognostics and Health Management with Tabular Foundation Models — Tabular Foundation Models, via tabularization and in-context learning, offer a data-efficient and robust solution for diverse industrial PHM tasks.
• Enabling Predictive Maintenance in District Heating Substations: A Labelled Dataset and Fault Detection Evaluation Framework based on Service Data — A new public dataset and evaluation framework enable reproducible early fault detection in district heating substations.
• Hybrid Neural Network and Conventional Controller Approach for Robust Control of Highly Unstable Systems: Application to Tilt-Rotor Control — Direct neural network input-output control fails for highly unstable systems; hybrid SMC with NN-enhanced dynamics is robust.
• Varifold Moment Invariants for Sustainable and Explainable Contour Feature Extraction — Varifold Moment Invariants unify existing moment invariants, providing high-power, geometrically meaningful contour features with low computational cost.
• Adaptive generative moment matching networks for improved learning of dependence structures — Adaptive bandwidth selection in GMMNs significantly improves learning of complex dependence structures and model prediction.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Artificial Intelligence.