A New Framework for Isolating Sensor Failures and Structural Damage in Noisy Environments Based on Stacked Gated Recurrent Unit Neural Networks

To address the problem of sensor faults and measurement noise being misinterpreted as structural damage in structural health monitoring (SHM), this paper proposes a new framework for distinguishing sensor faults and structural damage based on stacked gated recurrent neural networks (S-GRU NN) that considers measurement noise. In this framework, the sensor signal reconstruction model was constructed by learning and training the S-GRU NN. The sensor fault threshold was determined based on a statistical analysis of the response reconstruction error between the true and reconstruction values. The sensor fault and structural damage are then distinguished by the fact that the sensor fault is independent and the structural damage is global. The framework is compared with other isolation frameworks based on traditional deep learning models through numerical simulations of a three-span continuous beam and laboratory steel frame experiments. The results show that the S-GRU NN has better reconstruction effect and isolation performance of sensor faults and structural damage in noisy environment.

Dieses Werk wurde unter der Creative-Commons-Lizenz Namensnennung 4.0 International (CC-BY 4.0) veröffentlicht und darf unter den Lizenzbedinungen vervielfältigt, verbreitet, öffentlich zugänglich gemacht, sowie abgewandelt und bearbeitet werden. Dabei muss der Urheber bzw. Rechteinhaber genannt und die Lizenzbedingungen eingehalten werden.