Spatio-Temporal Attention Graph Neural Network for Remaining Useful Life Prediction

TL;DR

The paper tackles RUL prediction in complex industrial systems by integrating spatial and temporal features through a cascade STAGNN that combines a GNN for spatial relations and a TCN for temporal dynamics, augmented with multi-head spatio-temporal attention. It also investigates data preprocessing via unified versus clustering normalization based on operating conditions, showing that clustering normalization improves performance in multi-condition settings while STAGNN achieves state-of-the-art results even without clustering. Empirical evaluation on the C-MAPSS dataset demonstrates strong predictive accuracy and enhanced explainability through attention visualizations and feature representations. These contributions advance accurate, interpretable RUL prognostics and highlight normalization as a key factor when operating conditions vary in real-world deployments.

Abstract

Remaining useful life prediction plays a crucial role in the health management of industrial systems. Given the increasing complexity of systems, data-driven predictive models have attracted significant research interest. Upon reviewing the existing literature, it appears that many studies either do not fully integrate both spatial and temporal features or employ only a single attention mechanism. Furthermore, there seems to be inconsistency in the choice of data normalization methods, particularly concerning operating conditions, which might influence predictive performance. To bridge these observations, this study presents the Spatio-Temporal Attention Graph Neural Network. Our model combines graph neural networks and temporal convolutional neural networks for spatial and temporal feature extraction, respectively. The cascade of these extractors, combined with multi-head attention mechanisms for both spatio-temporal dimensions, aims to improve predictive precision and refine model explainability. Comprehensive experiments were conducted on the C-MAPSS dataset to evaluate the impact of unified versus clustering normalization. The findings suggest that our model performs state-of-the-art results using only the unified normalization. Additionally, when dealing with datasets with multiple operating conditions, cluster normalization enhances the performance of our proposed model by up to 27%.

Figures (5)

• Figure 1: The RUL prediction workflow and the spatio-temporal attention graph neural network
• Figure 2: Normalization results based on two methods. Sequences with different colors represent the $N$ records of the 7th sensor.
• Figure 3: True RUL and the predicted RUL by the proposed approach on the two test datasets.
• Figure 4: The t-SNE-based visualization of the feature representations for STAGNN and STAGNN*. The color and size represent the degradation of RUL
• Figure 5: Samples of spatial and temporal attention matrix.