A novel multi-level data fusion and anomaly detection approach for infrastructure damage identification and localisation

Abstract

In this paper, a new damage identification and localisation framework utilising multi-level data fusion and anomaly detection techniques is introduced. The performance of the framework is presented via case studies of a simply supported bridge and a continuous bridge, whose data were numerically generated through a surrogate model. Accelerations, deflections and bending moments obtained at multiple sensor locations when the bridge was subjected to a moving vehicle were used as the input. A damage-sensitive feature was established via coupling principal component analysis and Mahalanobis distance, allowing for initial data dimensionality reduction and information integration. Anomaly detection using a deep convolutional autoencoder was performed to identify the presence of single and multiple damages on the bridge. It is demonstrated that the proposed approach is independent of the mass and speed of moving vehicles, and is applicable to different types of bridges (i.e. simply supported bridge and continuous bridge) with only minor modifications to the data augmentation/preparation processes. The accuracy of damage identification for the simply supported bridge is shown to be consistently greater than 93%, even for extremely small damage severity (e.g. 1% reduction in local stiffness), and multiple damages. While the damage identification accuracy for the continuous bridge is greater than 94.9% for the majority of the tested cases. The proposed damage localisation approach can accurately determine damage location(s) for both single and multiple damage cases. It is shown how the proposed framework can be adopted for early damage detection to enable reliable decision-making and maximise structural safety.