Reliability Analysis And Life Prediction Of PC Continuous Bridge Based On GA-SVR

As a critical project of highway traffic engineering,bridges have an extraordinary significance for the normal use and safe operation of highway.Therefore,reliability analysis and life prediction related to bridges have important scientific value and application prospects.At present,the emphasis of reliability analysis in bridge structure is mainly about how to determine the limit state function appropriately.However,bridges are affected by many random factors during their service life,which means that the limit state function of each failure mode is usually a highly nonlinear implicit function.Furthermore,bridges are quite complicated structural systems and therefore there are considerable failure modes in the structure.Meanwhile,the correlation between various failure modes cannot be ignored.To solve the above mentioned problems in reliability analysis,the decision function in GA-SVR algorithm was applied to approximate the real limit state function.And the correlation coefficient in reliability index vectors was used to reflect the correlation between failure modes.Finally,the system reliability of bridge was solved by GA-SVR algorithm and reliability index vectors,which provided the approach to perform the time-dependent reliability analysis and predict the life of bridge during the normal use stage.The main research of this paper was as follows:(1)The theory related SVM was introduced and deduced,and the mathematical expression of SVR decision function was obtained.To develop an accurate model,Genetic Algorithm was adopted to optimize SVR parameters and corresponding algorithm—GA-SVR algorithm was designed and written.(2)The reliability calculation program based on GA-SVR algorithm,FORM method and JC method was written by MATLAB.The accuracy and rationality of the reliability calculation program were verified by two examples.Meanwhile,the performance of SVR model impacted by sample number,different ranges of data normalization and SVR parameters was also discussed.(3)Taking a PC continuous bridge as the engineering background,the finite element model of bridge during normal use stage was established by Midas/Civil.The samples composed of stochastic variable samples and corresponding finite element calculation were acquired to train SVR models of failure modes.GA-SVR algorithm was employed to make the limit state function of main failure modes explicit,and the correlation coefficient matrix of main failure modes was deduced.The calculation program of system reliability based on GA-SVR and reliability index vectors was designed and written,which offered calculation method for the time-dependent reliability analysis of bridge system.(4)The sensitivity analysis of reliability in each failure mode was performed by correlation analysis of mathematical statistics.Then the key stochastic variables with great influence on the failure modes of structure were obtained,and timevarying probability models of corresponding stochastic variables were established to calculate the time-dependent reliability of bridge system.Then the life of the bridge was predicted according to the result of time-dependent reliability and target reliability index.Moreover,the influence of key stochastic variables on life prediction was also discussed.Finally,it came to the conclusion,which summed up the study and put forward the problem that needed to be further studied.