Sensor self-diagnosis method based on a graph neural network

Abstract

Many types of sensors are used in industrial processes, and their reliability is high. However, the traditional method of regularly detecting and evaluating their health status is time-consuming and laborious, and is not suitable for the development of intelligent sensors. In this work, the relative entropy method is first used to quantitatively evaluate the redundancy relationship between sensors, and a sensor graph network is established based on this relationship. Secondly, an unsupervised multi-sensor self-diagnosis model, called attention-based pruning graph convolutional network, is proposed. In order to capture the strong redundancy among sensors by the attention mechanism, multi-sensor timing prediction is realised using a graph convolutional neural network, and the health status of each sensor can be independently judged by the changes in redundancy among the sensors. Finally, a temperature measurement system in a nickel flash furnace is considered as a case study to verify the feasibility and effectiveness of the proposed method.

Assessing Sensor Integrity for Nuclear Waste Monitoring Using Graph Neural Networks

“…The novelty of the work presented in this paper lay in the combination of these three elements: using a graph neural network for anomaly detection in a sensor network by employing node classification models. Similar work was achieved in Jiang et Luo 2023 [ 54 ], where a model for sensor self-diagnosis was used. In Deng et Hoooi 2021 [ 39 ], a GNN was developed for the detection of anomalies in time series, which were measured using sensors.…”