Bridge Structure Damage Detection Based On Convolution Neural Network Combining With Dragonfly Algorithm

Bridge is a kind of important infrastructure in the transportation network.It is of great social and economic significance to ensure the safe operation of the bridge.In actual operation,bridges will face many potential threats,such as environmental erosion,material aging and heavy load.In order to ensure the normal operation of the bridge structure,it is very important to diagnose the structural damage and health status of the bridge in time.Structural damage detection(SDD)is a core content of bridge structural health monitoring.Structural damage detection is an interdisciplinary application field.With the development of structural damage detection,advanced numerical analysis methods and calculation theory are introduced,which greatly broaden the method of structural damage detection.Deep learning and swarm intelligence optimization algorithms have strong numerical analysis and data processing capabilities,and have received a lot of attention and application in the field of damage detection in recent years.However,existing studies have found that both of them have their own advantages and disadvantages when applied separately to structural damage detection.Therefore,it is a meaningful research topic to combine them to achieve complementary advantages.However,the existing studies have found that there are advantages and disadvantages of both methods in structural damage detection.Therefore,it is a meaningful research topic to combine the two to achieve complementary advantages.In order to achieve this goal,this paper combines convolution neural network and dragonfly algorithm to study the location and quantitative detection of bridge damage.The specific research contents are as follows:(1)The research status of structural damage detection at home and abroad is comprehensively described.At the same time,the research status of convolution neural network and dragonfly algorithm in structural damage detection is introduced in detail.The main research direction and innovation of this paper are determined.(2)A detection method based on convolutional neural network is proposed for structural damage location recognition.Firstly,the structure and algorithm principles of the convolutional neural network are introduced in detail.Subsequently,the mathematical principle of structural damage detection is explained,and it is connected with the characteristics of convolutional neural network,and the theoretical feasibility of using convolutional neural network for damage detection is explained.The process of damage detection using convolutional neural network is introduced.At the end,a numerical simulation of multi-structure damage location based on convolutional neural network is also carried out.The results show that the proposed method can effectively locate structural damage and exhibits good robustness.(3)Aiming at the damage unit identified by the convolutional neural network,the dragonfly algorithm is proposed to further complete the quantitative detection of damage.First,the mathematical model and optimization steps of the dragonfly algorithm are introduced,which shows the calculation principle and usability of the dragonfly algorithm.Then according to the calculation principle of the dragonfly algorithm and the characteristics of the dynamic characteristics after the structure damage,the frequency residual and the modal guarantee criterion are selected as the objective function of the damage index to construct,and the objective function is constructed for these two damage indexes and using them.The process is introduced in detail.Finally,a numerical simulation is carried out on the quantitative damage detection of the structure,and the damage detection method mentioned in this article is compared with the method of damage detection using the dragonfly algorithm alone.The results show that the method proposed in this paper has more advantages in calculation speed and calculation accuracy than the structural damage recognition using dragonfly algorithm alone.(4)In order to verify the practical engineering applicability of the structural damage detection method combining convolutional neural network and dragonfly algorithm proposed in this paper,a simple beam test model was built in the laboratory for actual test verification.First,establish a finite element model of the test structure,and then modify the finite element model according to the collected test beam data to minimize the difference between the two,and then build a convolutional neural network based on the structure to be damaged.Finally,the test beam data under different working conditions are input into the network and algorithm for damage detection.For comparison,the dragonfly algorithm is used alone to identify the damage of the same structural working condition.The results show that the damage detection method proposed in this paper can more accurately identify the real damage of the test beam,and has better performance than the method of damage detection using the dragonfly algorithm alone,and has certain engineering practical value.