Disease Analysis And Intelligent Maintenance Of Small And Medium Span Bridges In Dry And Cold Regions

With the gradual improvement of China’s infrastructure construction,bridge construction has also gradually entered the buffer period,China is now facing no longer the vigorous development of bridge construction,but the whole life cycle of maintenance of bridges in service,especially in the dry cold regions of China,due to the region’s bridges for a long time by dry and cold,strong ultraviolet light and another harsh climatic environment,bridge maintenance practitioners to bridge safety diagnosis and disease prevention The level of bridge maintenance practitioners is reduced.Therefore,it is important to study the diseases of small and medium-span bridges in dry cold regions.Based on an overview of the development status of the bridge maintenance field and bridge disease research,this paper focuses on the analysis of safety state analysis technology and bridge apparent disease detection technology for small and medium-sized span bridges in dry cold regions and proposes an intelligent maintenance evaluation and detection method based on bridge disease conditions.This study can improve the national capacity of inspection and supervision,forecasting and early warning,decision-making and dispatching,and emergency response for highway bridges,and then promote the development of bridge structures in the direction of high strength,green,high toughness,and high reliability.The main research and findings of this paper are as follows:(1)Through field research on bridges in dry and cold regions,bridge inspection reports and literature review,and according to the specifications such as "Evaluation of Technical Condition of Highway Bridges" and "Specification for Maintenance of Highway Bridges and Culverts",a region-specific safety level evaluation index system is constructed through analysis of bridge disease mechanisms in dry and cold regions;At the same time,2543 images of small and medium span bridges in dry cold regions containing four types of diseases: cracks,spalling,exposed tendons,and water seepage were collected,and Label Img software was used to label the disease categories and construct a bridge apparent disease target detection dataset.(2)Forty small and medium-span bridges were selected as training samples,and the key factors affecting bridge safety were first analyzed by Principal Component Analysis(PCA),and further the Radial Basis Function(RBF)neural network training was optimized by Particle swarm optimization(PSO)to establish an intelligent maintenance evaluation model for small and medium span bridges in dry cold regions.The experimental results show that the model constructed by using PCA-PSO-RBF neural network has a training time of 74 s and an accuracy rate of 95.1%,which largely improves the network operation speed and has a smaller mean square error compared with the traditional RBF neural network and PSO-RBF neural network models.By using four bridges inspection data in Gansu Province as test samples for example analysis,the intelligent identification results of the model are consistent with the manual inspection results of each bridge,which verifies the reliability and accuracy of the model in this paper for intelligent maintenance evaluation of small and medium span bridges in dry cold regions.(3)For the problems of a large number of training model parameters and difficult to distinguish disease types,this paper proposes a lightweight model of bridge intelligent maintenance inspection based on YOLOv5＿FC algorithm for target detection of bridge apparent diseases.The model introduces a new SPP-Fast network structure on the backbone network,which can further improve the feature extraction capability;at the same time,it incorporates the Coord Att attention mechanism,which can effectively use the global information without changing the depth of the model,and then accurately obtain the target location information,making the target detection have better performance;this paper introduces a multi-layer fusion mechanism,which fully integrates the shallow and deep features and improves the target detection speed.The experimental results show that the Mean Average Precision(m AP)of the YOLOv5＿FC model is 90.2%,which is 26.6%、7.3%、11.7%,and 6% higher than the SDD model、Faster R-CNN model、YOLOv3 model and YOLOv5 s model.Finally,the effectiveness of the improved method is verified by ablation experiments,which makes the method achieve the expected goal and has an important reference value in the field of bridge apparent disease detection.