Damage Detection And Identification Of Steel Truss Bridge Based On Improved Artificial Neural Network And Genetic Algorithm

Railway bridges are susceptible to the problems of fracture,corrosion and fatigue during its intended service life owing to environmental and operational conditions.This may results to stiffness reduction in steel truss joints of the truss bridges and decrease the operational effectiveness of the bridge.The steel truss bridge is a commonly structural form of railway bridges,while bolted connections are the most widely adopted type of joints.Structural health monitoring has become the most efficient tool for diagnosing damages in civil engineering structures.Damage detection is taken as one of the most essential parts of Structural health monitoring.Without taking environmental conditions into account,damage detection may be inefficient and inaccurate for practical applications.Temperature is taken as one of the most vital and influential environmental effects on structures,particularly in bridges.In this research,a hybrid algorithm based on Improved Artificial Neural Network(IANN)coupled with Genetic Algorithm(GA)is proposed to evaluate the stiffness reduction of truss joints in a railway bridge.GA is utilized to provide training parameters and addresses the local minimum problems of IANN.Modal strain energy damage index and natural frequency are utilized as input data,whereas output data is damage characteristics locations and levels of truss joints.The members with joint damage are subdivided into three regions in the numerical model and the two regions with a length of 1/10 of the total member length(L)adjacent to the joints are designated as the end elements.IANN-GA accurately predicts vibration behavior and damage of the structure,including single and multiple stiffness reduction scenarios of truss joints.The main work and research conclusions are as follows:(i)For single damage detection scenario,the regression line is near the 45-degree line and the R-values exceed 0.98.As a result,the network has received good training and can be used to detect fatigue damage in the bridge under consideration.The R-value computed by ANN-GA(0.985)is higher than that computed by IANN(0.982).The difference between computed and real values(error)of IANN-GA is lowest,at 1.008,whereas those of IANN and GA are is 1.201 and 4.501,respectively.In terms of computational time,IANN and IANN-GA spend 105 s and 119 s,respectively to look for the best solution,whereas GA expends 55423 s for this process.(ii)For multiple damage detection,the regression line is near to the 45-degree line and Rvalues are greater than 0.97.This indicates that there is a good agreement between the computed and observed findings.The R-value calculated by IANN-GA(0.9758)is greater than that computed by IANN(0.9765).The difference between computed and real values(error)of ANN-GA is lowest,at 3.0503,whereas those of IANN and GA are is 3.2134 and 4.831,respectively.In terms of computational time,IANN and IANN-GA spend 1342 and 1465 s to look for the best solution,whereas GA expends 56324 s for this process.The findings presented in this thesis highlight the potential implications of incorporating IANN and GA approach into current structural Health Monitoring practices due to are accurate and need a lower computational time than ANN,and evolutionary algorithm(EA)alone in terms of structural damage localization and quantification.With the help of recommendations for proactive maintenance,this study aims to provide the necessary information for well-informed decisionmaking by continuously gathering and evaluating information in virtually real-time.