Performance Analysis Of Existing Masonry Structures Based On The Bayesian Inference And Dynamic Tests

Because of the good physical properties of its materials, simple construction method and low cost, masonry materials are still the dominant building materials in China. Masonry structures have a long history in our country. With the times going, these existing masonry structures have a lot of potential threats, so the performance test of the existing masonry structures have become particularly important. The in situ testing or core sampling method is not desirable because of the new damages to the structures. So the structure performance detection methods based on the non-destructive dynamic test become more and more popular. The basic frequency of existing masonry structures is a key indicator of their safety, its measured values are of high diversity and inadequate to make convincing estimations because of the uncertain nature of the existing damages and structural configurations. The traditional regression analysis method will totally depend on the measured values, if the measured values are discrete, it will lead to serious deviation of results. To address the issue, the Bayesian inference was applied through a Markov Chain Monte Carlo(MCMC) sampling method, combined with the prior judgments, the basic frequency was taken as an uncertain variable and inferred using statistical methods.First the sampling frequency and an accelerometer layout was determined based on 3 dimensional numerical simulations, and then the accelerometers were mounted. The acceleration time history values of the specimens were measured through ambient vibration tests and forced vibration tests, and the basic frequencies were calculated by Fast Fourier Transform(FFT). A Bayesian inference for the basic frequencies was applied through a Metropolis-Hasting(MH)sampling, and combined with the prior judgments, the posterior probability density functions(PDFs) of the basic frequencies were established. Estimations on the basic frequencies were made based on the PDFs, and estimations on the equivalent elasticity modulus of the structure were calculated by the finite element simulation. The proposed method was demonstrated to be robust for limited and noisy measured basic frequencies, and would readily lead to stable and appropriate posterior PDFs, which were strong bases for elasticity modulus estimation.Feasibility of the proposed method was illustrated through the dynamic test on an existing masonry building. Combined with the priori judgment, the posterior PDFswere obtained, and estimations on the equivalent elasticity modulus of the structurewere calculated by the finite element simulation. The method can be economicallyand effectively used to monitor the performances of existing masonry structures. Key Word: masonry structure, Bayesian inference, dynamic test, basic...