A new approach is presented for health monitoring of structures using terrestrial laser scanning (TLS). Three-dimensional (3D) coordinates of a target structure acquired using TLS can have maximum errors of about 10 mm, which is insufficient for the purpose of health monitoring of structures. A displacement measurement model is presented to improve the accuracy of the measurement. The model is tested experimentally on a simply supported steel beam. Measurements were made using three different techniques: (1) linear variable displacement transducers (LVDTs), (2) electric strain gages, and (3) a long gage fiber optic sensor. The maximum deflections estimated by the TLS model are less than 1 mm and within 1.6% of those measured directly by LVDT. Although GPS methods allow measurement of displacements only at the GPS receiver antenna location, the proposed TLS method allows measurement of the entire building's or bridge's deformed shape, and thus a realistic solution for monitoring structures at both structure and member level. Furthermore, it can be used to create a 3D finite element model of a structural member or the entire structure at any instance of time automatically. Through periodic measurements of deformations of a structure or a structural member and performing inverse structural analyses with the measured 3D displacements, the health of the structure can be monitored continuously.
In this study, a structural response recovery method using a convolutional neural network is proposed. The aim of this study is to restore missing strain structural responses when they cannot be collected due to a sensor fault, data loss, or communication errors. To this end, a convolutional neural network model for data recovery is constructed using the strain monitoring data stably measured before the occurrence of data loss. Under the assumption that specific sensors fail among the multiple sensors installed on a structure, the structural responses of these specific sensors are intentionally excluded and the remaining structural responses are set as the input data of the convolutional neural network. In addition, the intentionally excluded structural responses are set as the output data of the convolutional neural network. In case of a sensor fault, the trained convolutional neural network is used to recover the missing strain responses using functional sensors alone. The applicability of the proposed method is verified by a numerical study on a beam structure and an experimental study on a frame structure. The data recovery performance of the proposed convolutional neural network is discussed according to the number of failed sensors and the types of structural members with the failed sensors. Finally, the field applicability of the proposed method is examined using strain monitoring data measured from an overpass bridge in use over a long period of time.
30 January 2020 in Structural Control and Health Monitoring
This study presents a convolutional neural network (CNN)‐based response estimation model for structural health monitoring (SHM) of tall buildings subject to wind loads. In this model, the wind‐induced responses are estimated by CNN trained with previously measured sensor signals; this enables the SHM system to operate stably even when a sensor fault or data loss occurs. In the presented model, top‐level wind‐induced displacement in the time and frequency domains, and wind data in the frequency domain are configured into the input map of the CNN to reflect the resisting capacity of a tall building, the change in the dynamic characteristics of the building due to wind loads, and the relationship between wind load and the building. To evaluate stress, which is used as a safety indicator for structural members in the building, the maximum and minimum strains of columns are set as the output layer of the CNN. The CNN is trained using measured wind and wind response data to predict the column strains during a future wind load. The presented model is validated using data from a wind tunnel test of a building model. The performance of the presented model is verified through strain estimation with data that were not used in the CNN training. To assess the validity of the presented input map configuration, the estimation performance is compared with a CNN that considered only the time domain responses as input. Furthermore, the effects of the variations in the configuration of the CNN on the wind response estimation performance are examined.