Skip to content

Function-Informed Adaptive Graph Spatiotemporal Network for Remaining Useful Life Prediction of Aeroengines

Nov 2026 · IEEE Sensors Journal · Vol 26, pp. 34598-34608 · 0 citations · 39 references

Abstract

In this article, a function-informed adaptive graph spatiotemporal network (FI-AGST) has been proposed to improve the accuracy of remaining useful life (RUL) prediction for aeroengines. A dual-graph adaptive fusion architecture is adopted in FI-AGST. Static thermodynamic prior knowledge is encoded by a sensor functional graph (SFG). Window-conditioned data-driven correlations are captured by an adaptive learned graph (ALG). The two graphs are then integrated at the sample level by a graph-level attention fusion (GAF) module, producing an interpretable and flexible adjacency matrix. The proposed GAF method alleviates the limitation of single or static graph topologies under variable operating conditions. To improve joint spatiotemporal feature extraction, a graph-LSTM-dual-attention fusion (DAF) backbone combining graph-constrained temporal units with sensor and temporal dual-attention mechanisms is adopted. The proposed approach is evaluated on the benchmark C-MAPSS dataset from NASA (FD001–FD004). The results demonstrate that the proposed method achieves competitive prediction accuracy and stable performance across the four subsets.

View source

We use cookies to run the site and, with your consent, for analytics and to show ads. See our Cookie Policy.