Multi-Level Temporal Graph Networks with Local-Global Fusion for Industrial Fault Diagnosis

arXiv cs.AI / 4/22/2026

💬 OpinionIdeas & Deep AnalysisModels & Research

共有:

Key Points

The paper proposes a structure-aware multi-level temporal graph neural network with local-global fusion aimed at improving industrial fault detection and diagnosis.
It dynamically builds a correlation graph between process variables using Pearson correlation coefficients, then extracts temporal behavior with an LSTM-based encoder and spatial sensor dependencies with graph convolution layers.
A multi-level pooling strategy progressively coarsens the graph to learn higher-level patterns while preserving fault-relevant details.
A fusion module combines fine-grained local features and coarse global patterns for final fault prediction.
Experiments on the Tennessee Eastman process (TEP) show the method outperforms multiple baseline approaches, especially on complex fault cases.

Abstract

Fault detection and diagnosis are critical for the optimal and safe operation of industrial processes. The correlations among sensors often display non-Euclidean structures where graph neural networks (GNNs) are widely used therein. However, for large-scale systems, local, global, and dynamic relations extensively exist among sensors, and traditional GNNs often overlook such complex and multi-level structures for various problems including the fault diagnosis. To address this issue, we propose a structure-aware multi-level temporal graph network with local-global feature fusion for industrial fault diagnosis. First, a correlation graph is dynamically constructed using Pearson correlation coefficients to capture relationships among process variables. Then, temporal features are extracted through long short-term memory (LSTM)-based encoder, whereas the spatial dependencies among sensors are learned by graph convolution layers. A multi-level pooling mechanism is used to gradually coarsen and learn meaningful graph structures, to capture higher-level patterns while keeping important fault related details. Finally, a fusion step is applied to combine both detailed local features and overall global patterns before the final prediction. Experimental evaluations on the Tennessee Eastman process (TEP) demonstrate that the proposed model achieves superior fault diagnosis performance, particularly for complex fault scenarios, outperforming various baseline methods.