Convolutional Neural Networks for Real-Time and Wireless Damage Detection
Structural damage detection methods available for structural health monitoring applications are based on data preprocessing, feature extraction, and feature classification. The feature classification task requires considerable computational power which ma
- PDF / 911,644 Bytes
- 8 Pages / 595.516 x 790.987 pts Page_size
- 77 Downloads / 257 Views
Convolutional Neural Networks for Real-Time and Wireless Damage Detection Onur Avci, Osama Abdeljaber, Serkan Kiranyaz, and Daniel Inman
Abstract Structural damage detection methods available for structural health monitoring applications are based on data preprocessing, feature extraction, and feature classification. The feature classification task requires considerable computational power which makes the utilization of centralized techniques relatively infeasible for wireless sensor networks. In this paper, the authors present a novel Wireless Sensor Network (WSN) based on One Dimensional Convolutional Neural Networks (1D CNNs) for real-time and wireless structural health monitoring (SHM). In this method, each CNN is assigned to its local sensor data only and a corresponding 1D CNN is trained for each sensor unit without any synchronization or data transmission. This results in a decentralized system for structural damage detection under ambient environment. The performance of this method is tested and validated on a steel grid laboratory structure. Keywords Convolutional neural networks · Real-time damage detection · Structural health monitoring · Structural damage detection · Wireless sensor networks
17.1 Introduction The civil infrastructures are inevitably aging and engineers have always been interested in the level of aging by looking at the visible damage and/or trying to detect the invisible damage. Systematic monitoring of infrastructure has become a norm in time with the simultaneously emerging profession of Structural Health Monitoring (SHM) [1]. From visual inspection to sophisticated sensor usage, the field of Structural Damage Detection (SDD) techniques within the SHM context enabled engineers and facility owners make healthy decisions on infrastructure, for multiple disciplines. In parallel to improvements in sensor technology [2], the use of wireless sensors is adopted to create Wireless Sensor Networks (WSN) in SDD and SHM applications [3–8]. The mainstream “centralized” methods in WSN applications involve processing large amount of data which is infeasible. In this paper, the authors are introducing the use of One Dimensional Convolutional Neural Networks (1D CNNs) to create a “decentralized” technique for WSNs for efficient SDD applications in ambient dynamic environment [9]. In this technique, each sensor unit is trained locally in a decentralized way, running the raw ambient acceleration data on a steel laboratory structure at Qatar University.
17.2 Convolutional Neural Networks and the Laboratory Structure The extensive vibration analysis [10–16], serviceability [17–25], suppression [26–32] and optimization [33–37] work conducted by the authors has motivated them to utilize their experience in SDD/SHM applications. The authors have been focused on vibration based SDD studies [38–43], and have lately introduced the use of CNNs in vibration based SDD field [44–47]. The structure of this study presented here is arguably the largest stadium structure built and instrumented in a
O. Avci () · O
Data Loading...
