| With the extension of service life,thousands of concrete dam have been exposed to diseases and unsafe factors due to aging.The engineering practice has proved that the inspection tour is a reliable method to find the abnormal parts in time.However,on the one hand,the inspection information recorded by dam operation and maintenance managers is mostly non-structural data and no corresponding standards have been established,which is not conducive to the subsequent quantitative evaluation.On the other hand,the traditional inspection is of low efficiency and high risk,and it takes a lot of work to screen and capture important hidden danger information from a large number of inspection images.In order to solve the above problems,this paper proposes the coupled man-machine collaboration and deep learning to carry out inspection methods and evaluation of concrete dam.Firstly,the feasibility of the application of unmanned aerial vehicle in inspection work of concrete dam is explored from man-machine cooperation technology,and the inspection method based on man-machine cooperation technology is proposed.Then the typical defect image features of concrete dam are analyzed and the defect image detection model of concrete dam is constructed.Then design and develop concrete dam inspection information management system;Finally,the above research results are applied to the gravity dam project of Wangfuzhou water conservancy project.The main research contents are as follows:(1)Research on patrol inspection method of concrete dam based on Man-machine collaboration.The inspection content is divided according to human senses,and the UAV is used as human visual extension equipment to assist human inspection,and the key points of man-machine collaborative inspection technology are discussed.Based on the DJI Phantom 4Pro UAV design,the UAV test was carried out under different shooting distance,different wind speed and different lighting conditions and other factors affecting the inspection.The test results show that 3m range can meet the expected requirements of patrol inspection,while the conventional image binarization method is difficult to meet the requirements of crack segmentation,and it will not collision with the photographed object under the unidirectional wind speed of 2.5m/s～4m/s.Finally,the application process of man-machine collaborative inspection method is summarized,which provides guidance for the subsequent system design and practical engineering application.(2)Research on fast detection of concrete dam defect image based on deep learning.The common defect images of concrete DAMS are analyzed and the common cracks and water seepage are selected as the research objects.The simulated defects of concrete slabs were formed,and the illumination influence test was designed to select the optimal environment for image acquisition,and the typical defect data set was collected and constructed under the environment with the highest image quality.Finally,the "two-step method" defect detection model based on Yolov5+U-net network was trained.The evaluation results showed that the defect detection model performed poorly in the identification and segmentation of fine cracks and well in the seepage data set.However,for the purpose of establishing the classification,location and segmentation model of typical defects of concrete DAMS,The "two-step" model can achieve better results.(3)Comprehensive Evaluation of concrete dam inspection condition.In this paper,the comprehensive evaluation idea of foundation evaluation by dam operation and maintenance management personnel is sorted out.Starting from the comprehensive inspection and inspection content of concrete dam,a four-level inspection and inspection evaluation system with scientific and practical features is established layer by layer,and the corresponding evaluation criteria are divided to provide standardized reference for dam operation and maintenance management personnel to carry out "multiple choice" type foundation evaluation.Finally,a comprehensive evaluation model is established from the two aspects of "scoring method" and "scoring method".In order to consider the objectivity of evaluation results,the "scoring method" adopts the combined weight-set pair evaluation model,while the "scoring method" establishes the corresponding score deduction standard from the point of speed and simplicity.The scoring method and deduction method were used to conduct A comprehensive evaluation on the inspection status of gravity dam in Wangfuzhou hydraulic engineering project.The results showed that the inspection status of gravity dam was grade A,that is,in normal condition.The research also shows that the evaluation results are basically consistent with the safety appraisal report and the actual situation,which indicates that the evaluation method is scientific and reasonable.(4)Research on Patrol Inspection Management System of concrete dam.Combined with the proposed "multiple-choice" inspection grading standards and methods,the system development objectives and requirements are set up,the system function modules,database,development and operation environment are designed,the concrete dam inspection information management system is developed based on JAVA language,and the inspection information input and query,comprehensive evaluation,comprehensive management and other functions are realized.Combined with the gravity dam of Wangfuzhou hydraulic project,the interface of each function module of the system is shown. |
