Structure Identification And Performance Analysis Under Uncertainty

Uncertainty in civil engineering is inevitable.the US Defense Advanced Research Projects Agency launched the “Quantization Uncertainty in Physical Systems” project in 2015.The goal is to make uncertainty quantification as an integral part of simulation and modeling process.After decades of development,structural health monitoring technology has made great progress.Advanced sensor technologies such as fiber optic sensing,wireless sensing,and microwave radar technology have emerged and rapidly promoted.The results of structural identification and performance analysis based on monitoring data have begun to be integrated into the normal maintenance management of civil engineering.However,the results of structural identification at this stage generally only are deterministic results,and the reliability of the results are unknown.At the same time,it also brings inconvenience to the structural performance evaluation based on probability and statistics theory.In the structural design,the influences of various uncertain sources such as materials and structural systems are fully considered.Therefore,the author believes that the uncertain effects should also be paid attention in structural health monitoring.In this paper,we carry out the following research work in the aspect of structural identification and performance analysis considering uncertainty:(1)A subspace flexibility identification methdod which is adaptive to different types of input forces is developed.The flexibility of a structure is the inverse of the stiffness and can predict the displacement of the structure under any load,so the flexibility identification can be related to the structural performance analysis.The bridge rapid impact vibration testing technology developed by our group can simultaneously measure the input and output responses of a structure,and has the unique advantage of being able to identify deep structural parameters.This paper analyzes the influence of the types of input force on the structure identification algorithm,and then proposes a time domain based subspace flexibility identification method.The new identification method is applicable to various input force types,so it is more conducive to rapid impact vibration test.(2)Based on long gauge strain measurement data,a methodology is developed for simultaneously identifying strain flexibility and displacement flexibility.The long gauge strain sensor has the unique advantage of reflecting both the macroscopic corner and the microscopic strain of a structure.This paper further develops the macro strain mode theory based on long gauge strain.The developed algorithm adapts the state space model to the long gauge strain,and uses the conjugate beam method as a bridge to simultaneously identify the strain flexibility and displacement flexibility of the structure.The developed method is verified by simulation example and experimental example.(3)The uncertainty quantification method for flexibility identification under ambient vibration test is established,and the structural performance analysis is carried out based on the identified flexibility and its corresponding confidence intervals.There are uncertainties in the identification algorithm based on monitoring data.Firstly,through the first-order perturbation theory,the uncertainty is transmitted from the monitoring data step-by-step to the modal parameter to estimate the confidence interval of the modal parameters.Combined with the mass matrix provided by the finite element and its coefficient of variation,the confidence interval of the identified modal flexibility is derived.Finally,the predicted deflection and its confidence interval are calculated,thus the uncertainty quantification from monitoring data to predicting deflection is completed.In order to carry out the performance analysis based on the identified results,the reliability calculation of the displacement serviceability limit state is carried out and the bridge performance evaluation under structural overload is analyzed.(4)Uncertainty quantification of input-output subspace identification is established.In this paper,the uncertainty quantization of subspace identification technology under input and output data is proposed firstly.By analyzing the uncertainty sources under input and output measurement dada,the first-order perturbation theory is applied to the complete system matrix,and then the structural frequency response function,modal scaling factor,structural flexibility and their corresponding confidence intervals are further obtained.The theoretical formulae involved are validated by the Monte Carlo sampling method in the simulation example,and finally the proposed method is applied to a benchmark experiment example.(5)The confidence intervals of identified results are integrated into the uncertainty of finite element model,and then quantifies the uncertainty of the model to further perform probabilitybased structural performance analysis.When performing model updating,there are problems in which multiple finite element models match the structural identification results.In order to solve the problem that the optimal model is not unique,this paper takes the model parameters as the sampling target,the identified results and their confidence intervals as the observation values,and then conduct the Bayesian inference to obtain the posterior probability of the finite element model through the Markov chain Monte Carlo sampling method.In order to further solve the computational efficiency problem,this paper introduces the surface response model for sampling calculation,and conducts the structural probability response prediction based on first-order perturbation theory.In this paper,the Nanjing right-hand suspension bridge is taken as an example to analyze the uncertainty of the identified results and the bridge performance.Finally,a structural performance analysis method based on local and global hybrid simulation is proposed as an extension.