Research On Bridge Structure Damage Identification Based On Deep Learning Theory

As an important project in transportation engineering,bridges are key nodes connecting different lines.During the service period of the bridge structure,due to external loads,natural environment,material properties and other factors,different degrees of fatigue and damage will inevitably occur,which will seriously affect the safety,applicability and durability of the structure,and affect the lives and property safety of the people.Therefore,it is necessary to adopt accurate damage identification methods to discover the damage status of the structure in time.Traditional recognition methods have insufficient computing power and low recognition accuracy in processing massive data.Deep learning can process high-dimensional massive data by virtue of its multi-level perceptrons.Therefore,in the field of damage recognition,it is compared with traditional The pattern recognition method has great advantages.The main research work of this article includes:(1)Describes the theoretical development of deep learning,introduces three current mainstream deep learning network models,compares their performance,and selects Deep Belief Network as the method of bridge damage identification through comparison.(2)Propose a bridge structure damage identification method based on the Deep Belief Network,and build a Deep Belief Network suitable for the characteristics of bridge damage identification.(3)Taking a simply supported steel beam as an example,the proposed method is used to do damage identification analysis on the vertical acceleration response of the structure directly extracted,and the identification results about the damage location and damage degree of the structure are obtained,and compared with the traditional SVM support vector machine,BP The neural network comparison shows that the recognition performance of this method is significantly better than the traditional recognition method,and the anti-noise performance is excellent,and the recognition accuracy rate is above 85%.(4)Taking a prestressed concrete continuous beam as an example,setting multiple damage conditions,comparing the proposed method with the recognition results of SVM support vector machine and BP neural network,the results show that the recognition performance of this method is compared with traditional methods under multiple damage conditions It has advantages.It has certain anti-noise performance under multi-damage conditions.Under 30%noise conditions,the accuracy of damage location recognition is over 85%,and the accuracy of quantitative recognition is slightly lower,floating around 70%.(5)A recognition method based on wavelet transform and Deep Belief Network is proposed,and the multi-damage conditions of cable-stayed bridges are recognized.The comparison results show that the method has obvious anti-noise performance.It is better than the recognition method of Deep Belief Network.(6)Using the proposed recognition method based on wavelet transform and deep belief network,the test verification of damage recognition of I-beam is carried out.The test results show that the method can identify the damage of the structure under actual conditions,and the recognition accuracy rate is relatively high.High,the results of damage location and damage quantification are both above 80%.In summary,the damage identification method based on deep learning theory can be used for bridge structures with good performance,but how to implement the method under actual conditions needs more in-depth research.