Bridges are critical to economic and social development of a country. In order to ensure the safe operation of bridges, it is of great significance to accurately predict displacement of monitoring points from bridge Structural Health System (SHM). In the paper, a CEEMDAN-KELM model is proposed to improve the accuracy of displacement prediction of bridge. Firstly, the original displacement monitoring time series of bridge is accurately and effectively decomposed into multiple components called intrinsic mode functions (IMFs) and one residual component using a method named complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN). Then, these components are forecasted by establishing appropriate kernel extreme learning machine (KELM) prediction models, respectively. At last, the prediction results of all components including residual component are summed as the final prediction results. A case study using global navigation satellite system (GNSS) monitoring data is used to illustrate the feasibility and validity of the proposed model. Practical results show that prediction accuracy using CEEMDAN-KELM model is superior to BP neural network model, EMD-ELM model and EMD-KELM model in terms of the same monitoring data.
Monitoring the dynamic responses of bridge structures has received considerable attention. It is important to synchronously measure both the quasi-static and dynamic displacements of bridge structures. However, the traditional accelerometer method cannot capture the quasi-static displacement component, although it can detect the dynamic displacement component. To this end, a novel composite instrument of a smartstation was proposed to monitor vibration displacements of footbridges. Full-scale experiments were conducted on a footbridge to validate the feasibility of the composite instrument-based monitoring method. A Chebyshev filter and wavelet algorithms were developed to process the composite instrument measurements. It was concluded that the measurement noise of the composite instrument was mainly distributed in a frequency range of 0–0.1 Hz. In two case studies with displacement peaks of 5.7–10.0 mm and 1.3– 2.5 mm, the composite instrument accurately identified the quasi-static and dynamic displacements. The composite instrument will be a potential tool for monitoring structural dynamics because of its enhanced overall performance.
To meet the demand for high-precision recognition and real-time processing when using intelligent detection methods in daily concrete bridge crack detection tasks, a high-precision lightweight crack image recognition method based on knowledge distillation is proposed. Through semisupervised learning, the information embedded in unlabeled data is used to train a lightweight network to obtain high-precision crack recognition results. First, by comparing the recognition accuracy of lightweight network under different training strategies, the effectiveness of the knowledge distillation method in improving the accuracy of lightweight network is verified. Second, a label-free knowledge distillation method for engineering detection is proposed. Finally, using MobileNetv3 as the student model, under the self-built three-classification dataset of nearly 700,000 images, the accuracy reached 99.84% through the proposed knowledge distillation method, which is very close to the teacher model ResNeXt101 accuracy (−0.08%), and MobileNetv3 compared with ResNeXt101 its model inference speed increased by more than 13 times. At the same time, the MobileNetv3 model improves the inference speed by more than 13 times and reduces the number of model parameters by more than 35 times compared to the ResNeXt101 model. The research results fully demonstrate that under the premise of having a large amount of unlabeled data, using the proposed knowledge distillation method, the recognition accuracy of the trained lightweight model has a significant advantage over the existing lightweight network training strategies.