Damage Identification Of Long-Span Continuous Rigid Frame Bridges Based On Support Vector Machines

The damage and local failure of bridge has occurred because of frequent load action, overloading, the unfriendly natural environment and the damage caused by traffic accident during the bridge operation. Moreover, along with the bridge service life growth, there will be more and more damage types and the locations emerged. What is more the damage degree will become more and more serious. In order to guarantee the operational safety and the durability of bridge structures, and give the appraisal to the stated of health and bearing capacity of bridge, it is extremely essential to carries on the health monitoring of bridge structures. As the core of bridge health monitoring system, the damage identification theory has already become the focus problem which many engineers and technicians concerned.The state of arts and purposes of bridge damage identification were introduced firstly, and the support vector machines theory is systematic discussed. Then the support vector machines method was adotped in the damage indentificaiton of long-span continuous rigid frame bridges and good identification effect was obtained. The specific studies were listed as following:1. The state of arts and purposes of bridge damage identification was outlined, and the contemporary damage identification method was introduced as well. Then the brief introduction for the support vector machines method and vulnerability zone of long-span continuous rigid frame bridges was carried out.2. Several dynamic fingerprints of pattern recognition are introduced, the deduction of each dynamic fingerprints was figured out. And the advantages and disadvantages of each fingerprint was briefly introduced as well.3. The theory and method of support vector machines were introduced, and the software - LibSVM adopted in the article was briefly discussed.4. The long-span continuous rigid frame bridges were taken as the example, different feacture vectors were constructed, and the support vector classification and support vector regression were adopted respectively to locate and quantify the damage. The damage identification resule which under the different indexs is compared with each other and the influence of different noise level is also considered.Finally, the main works and the research results are summarized, and the content which needs further research is pointed out.