Application Of Convolutional Neural Network In Structural Fault Diagnosis

Civil engineering structures are easily affected by many uncertain factors during its operation,such as external over limit load,fatigue and corrosion.These factors will lead to serious potential safety hazards within the life span and threaten the users' lives and properties.Therefore,necessary measures shall be taken to conduct regular or real-time structural health detection.However,Traditional bridge health diagnosis methods are still insufficient and limited.With the advancement of science and technology,bridge health detection also presents a trend of big data,with a view to mining the features from big data for damage diagnosis.Meanwhile,the neural network diagnosis technology based on deep learning algorithms has great advantages in terms of big data processing,associative reasoning,adaptive study,faults tolerance and so on.Therefore,the convolution neural network algorithm is adopted to realize the intelligent diagnosis of structural damage.The main research contents include the following points:(1)This thesis studies the necessity of health diagnosis of civil structures,summarizes the existing methods of structural damage detection,compares the unique advantages of deep learning algorithm in the field of structural damage detection,and highlights the advantages of deep learning in the field of intelligent damage detection.(2)A method of structural damage detection based on convolution neural network is proposed.The original vibration time history signal of the structure is obtained through the finite element simulation test as the input of the network.The theoretical exploration test is carried out for the proposed method,and the influence of different types of excitation on the structure on the recognition accuracy is focused.(3)Design a test structure for example analysis and verify the damage detection method based on convolutional neural network.(4)On the basis of(2),the finite element software was used to carry out numerical simulation tests on the prefabricated simple supported beam bridge structure and the combined arch bridge structure,and preliminary study on the damage in the horizontal and vertical distribution was carried out.It was found that the staged learning method adopted in this paper has good feasibility in the preliminary damage localization.The results show that the damage detection method based on convolution neural network algorithm can effectively diagnose the structural health status.The type of stimulus exerted on the structure is not the decisive factor that affects the accuracy of its recognition,and through example analysis,it proves that the convolutional neural network performs well in practical applications.On the other hand,through the staged learning method used in this paper,both horizontally and vertically distributed structural damage can be identified and initially located.