Virtual Sensor for Real-Time Bearing Load Prediction Using Heterogeneous Temporal Graph Neural Networks

TL;DR

The paper tackles the challenge of real-time bearing load monitoring without invasive sensing by introducing HTGNN, a heterogeneous temporal graph neural network that fuses temperature and vibration signals while incorporating rotational-speed context. By modeling temperature and vibration as distinct node types and leveraging both same-type (GCN) and cross-type (GATv2) interactions across time, HTGNN delivers accurate axial and radial load predictions via a final MLP. The approach demonstrates superior performance over a 1D-CNN baseline, especially under representative training conditions, and highlights the value of physical priors encoded through the bearing sensor network connectivity. This virtual sensor enables continuous load estimation beyond the battery life of sensor-rollers, facilitating proactive maintenance and extending bearing health management in industrial settings.

Abstract

Accurate bearing load monitoring is essential for their Prognostics and Health Management (PHM), enabling damage assessment, wear prediction, and proactive maintenance. While bearing sensors are typically placed on the bearing housing, direct load monitoring requires sensors inside the bearing itself. Recently introduced sensor rollers enable direct bearing load monitoring but are constrained by their battery life. Data-driven virtual sensors can learn from sensor roller data collected during a batterys lifetime to map operating conditions to bearing loads. Although spatially distributed bearing sensors offer insights into load distribution (e.g., correlating temperature with load), traditional machine learning algorithms struggle to fully exploit these spatial-temporal dependencies. To address this gap, we introduce a graph-based virtual sensor that leverages Graph Neural Networks (GNNs) to analyze spatial-temporal dependencies among sensor signals, mapping existing measurements (temperature, vibration) to bearing loads. Since temperature and vibration signals exhibit vastly different dynamics, we propose Heterogeneous Temporal Graph Neural Networks (HTGNN), which explicitly models these signal types and their interactions for effective load prediction. Our results demonstrate that HTGNN outperforms Convolutional Neural Networks (CNNs), which struggle to capture both spatial and heterogeneous signal characteristics. These findings highlight the importance of capturing the complex spatial interactions between temperature, vibration, and load.

Figures (6)

• Figure 1: Architecture of the proposed Heterogeneous Temporal Graph Neural Network (HTGNN) for Load Prediction.
• Figure 2: Test-rig configuration.
• Figure 3: Train-test split of bearing load conditions for vibration data analysis (55% training, 45% testing)
• Figure 4: Heterogeneous graphs for bearing sensor network relationship modeling. (a) T-T (b) V-V (c) T-V (d) V-T (e) connectivity across two test rig bearings.
• Figure 5: Examples of input signals and load prediction performance. Shaded areas indicate unseen conditions.