| This study is focused on performance evaluation of 5 different modal parameter estimation methods by vertical ambient vibration response and 4-year System Identification based on obtained modal parameters by ANN method.;In chapter 2, short descriptions of signal processing (filtering and windowing) are denoted and 5 modal parameter estimation methods (PP, FDD, ITD, ERA, SSI) are explained in detail for the sake of its real application in the later chapters.;In chapter 3, general information of Jamboree Road Overcrossing (JRO) has been introduced in the beginning. Then, modal parameter estimation has performed in 2 different ways. One is the modal parameter estimation with 1 average data by 5 different methods in order to compare frequency and mode shape differences. The other way is the estimation also by 5 different methods; however, analysis has performed with respective 30 different data for performance evaluation of 5 methods based on statistical view (Standard Deviation). According to 1 average data analysis, it was found that frequency domain methods had detected higher natural frequency values in both first and second mode than time domain method. From the second analysis, SSI-BR showed the best performance in time domain and FDD was the best in frequency domain according to deviation.;In chapter 4, long term modal frequency change detection by using FDD method with 30 selected data from 2002 to 2005 in summer and winter seasons were performed. Annually 0.02 Hz in first mode and 0.00575 Hz in second mode decrements were detected. Based on obtained mode frequencies and its corresponding mode shapes, Back Propagation Neural Network (BPNN) was performed. Qausi-Newton algorithm was selected for training function and total 2744 input-output pairs from FE model were used for network to be trained. Through the trained network, stiffness decrement of column, superstructure and abutments were detected and formularized. Those annual decrements were 1.19%, 1.512% and 0.652% respectively starting from the year of 2002. |
