| dc.description.abstract | The need for assessment of structural safety through the estimation of parameters, mainly stiffness, has been acknowledged in the literature. Their demand is due to an increasing number of aging structures, some of which may be reaching the end of their projected life. In this context, many methodologies have been explored to characterize structural parameters using measurements of various natures as their inputs. From the point of view of measurements, a basic division can be drawn between methods using static and dynamic data. The latter are common in structural health monitoring systems due to the fact that measurements can be easily obtained. On the other hand, static tests require simpler mathematical models than dynamic data, but measuring in the field can be more challenging, requiring a more strict control of the tests. There are numerous algorithms, including neural networks, Bayesian updating or genetic, devoted to estimate structural parameters of a model from measurements. A common denominator to all methodologies is the negative influence of noisy measurements on the results. In practice, the application of static measurements to bridge safety has been mostly limited to proof load tests. Thus, the main motivations behind this thesis are threefold; first, to evaluate the potential of using and extending surveying techniques as a means of obtaining measurements for structural assessment purposes; second, to move forward methodologies for estimating structural parameters from static measurements; and third, to contribute to a better understanding of changes in frequencies of a vehicle-bridge system when employing dynamic measurements. Surveying techniques offer a series of advantages when measuring the response of a structure: they are almost entirely non-contact, easily implemented and low-cost. However, they are only equipped to obtain static measures and their accuracy may be limited, especially for small displacements. Modern surveying techniques allow measuring a high number of points in a short amount of time, but their inaccuracy can cause small changes due to structural damage to go unnoticed. Here, two novel applications of surveying techniques for estimating boundary conditions and material loss in two beam-and-slab bridges are introduced. Both bridges are damaged or suspected to have deterioration. In one case, an electronic tachymeter is employed to obtain deflections of the structure in a series of points; which are used to evaluate the stiffness of the bridge and the state of their supports. In the other case, high-definition survey is used to evaluate the damage of a structure and a methodology is tested to obtain a continuous deflection of the bridge After demonstrating direct applications of surveying to bridge inspection, the thesis focuses on the development of a cross-entropy algorithm for obtaining structural stiffness using static measurements. In contrast to other deterministic approaches, a positive feature of cross-entropy lies in the evaluation of stiffness as a random variable, thus, generating results consisting of a numerical value for stiffness plus an indicator of the reliability of the estimation. The cross-entropy algorithm is tested in an experimental set-up and numerical simulations that include a noise model based on surveying measurements. A modification of the algorithm including random field theory is also investigated. Some issues concerning noise remain, and for this reason, a regularization method is proposed to mitigate its effect. The impact of regularization on the results is assessed with surveying data from the loading of a concrete beam in a laboratory test. Finally, changes in forced frequencies have implications for structural health monitoring systems based on recording accelerations uninterruptedly. Therefore, a dynamic investigation of the variation in bridge frequencies due to vehicles, as a result of their characteristics and position on the bridge, is carried out. | spa |
