Sensitivity Analyses And Experimental Validations Of Structural Damage Detection Indicators Based On Steel Frame

In order to avoid catastrophic accidents caused by structural damage of large civil engineering structures and infrastructures,and to guarantee the safety of human life and property,researchers put forward a variety of methods to detect the damage of structures in service,so as to repair and reinforce the structures as early as possible.In these studies,mode shapes,modal strain energy,strain modes have been widely studied.This paper also studies these three indicators,hoping to explore the sensitivities of the three damage indicators and improve the accuracy of the future structural damage detection.In this paper,the sensitivity part first introduces the derivation of the eigen(mode shape)sensitivity,then based on the eigen sensitivity,the analytical expressions of the first-order modal strain energy sensitivity and first-order modal strain sensitivity are derived.Then,a spatial steel frame numerical model is established to analyze the sensitivities of the three indicators.The relationship between the sensitivity of three damage indicators,the relationship between the sensitivity of the damage indicators and the degree of damage,the relationship between the sensitivity of the damage indicators and the location of damage,the relationship between the sensitivity of the damage indicators and the type of damage elements,and the impact of multiple damages on the sensitivities of the damage indicators are emphatically discussed.Finally,the first order modal difference of three indexes is used to identify the damage in the numerical model to verify the sensitivity analysis results.In this paper,the theory of strain modal analysis is introduced before the experimental tests,and the frequency response function is derived by using the vibration theory of beam.Compared with the frequency response function in the existing mechanics textbooks,it has more explicit and intuitive physical meaning and does not involve complex matrix calculation.Then,the experimental equipment of a steel frame structure is established.Acceleration sensors and strain gauges are used to obtain acceleration-time curves and strain-time curves respectively,and the required modal parameters were obtained under the analysis of frequency response function and strain frequency response function respectively.Using the measured modal parameters,the modal shape difference,the modal strain energy difference and the modal strain difference are calculated respectively.The results were compared with the damage detection sensitivity of the three indicators to verify the conclusions of the sensitivities of the three indicators.Based on the experimental results,it is found that the three damage detection indicators adopted had misjudgment of damage under multiple damage conditions.Therefore,a BP neural network is proposed to combine the three damage indicators for damage detection.The purpose is to combine the advantages of three kinds of damage indicators to identify the damage,so as to improve the damage detection effect,predict the damage degree and reduce the signal interference.The specific steps include the introduction of damage detection process,network design,result judgment standard and network training.Finally,the experimental data are input into the trained network to automatically obtain the predicted damage detection results.