Most bridge condition assessment methods available in the literature are based on deflection, acceleration or strain measurements. Although current technology enables the measurement of rotational deformation on a bridge structure with high precision, the potential of rotation measurements as a main parameter to identify damage has not been thoroughly investigated. This is the main focus of this dissertation.
Initially, the sensitivity of rotation as the main parameter to identify damage is investigated through numerical analyses carried out on a 1-D simply supported bridge model. It is shown that when local damage occurs, even when it is remote from a sensor location, it results in an increase in the magnitude of rotation measurements, proving that rotation is a parameter that is sensitive to damage. This fact is exploited in a Bridge Weigh-In Motion (B-WIM) algorithm to develop a novel bridge damage detection methodology. Broadly speaking, in the proposed method, damage manifests itself as an increase in mean rotation. However, rather than looking directly at rotation values which are subject to natural variations, in this method, the rotation measurements are used in a B-WIM algorithm to predict the axle weights of the passing vehicles. This has the effect of separating the contributions of individual axles to the rotation signal, thereby removing the influence of vehicle configuration (axle spacings and relative axle weights). Providing that the influence line is not recalibrated after damage takes place, the damaged bridge gives an increase in the inferred weight. Unlike directly measured rotations where the ‘healthy-bridge answer’ is unknown, when working with a traffic population, the healthy bridge answer can be found – it is the statistical distribution of vehicle weight data for the site. This is generally repeatable – for example, the mean and standard deviation of fully loaded 5-axle trucks tend to remain the same from month to month. The capability of the proposed method to detect local damage on a bridge structure is demonstrated through numerical analysis carried out on a simply supported bridge model using real ambient traffic data obtained from 5 years of Weigh-in-Motion data from a site in the United States. The results and the specific details of the proposed methodology are elaborated in Chapter 2.
The next chapter develops a new bridge damage detection methodology using direct rotation measurements. Through initial numerical analysis carried out on a 1-D bridge model, it is shown that the difference in the rotation responses of a bridge to a moving vehicle obtained for healthy and damaged bridge states can identify damage and its location. Subsequently, a sophisticated 3-D dynamic Finite Element (FE) model of a 20 m long simply supported bridge structure is used to test the robustness of the proposed damage detection methodology under more realistic conditions. An extensive range of damage scenarios is investigated using the FE model. These simulations were performed as ‘blind’ tests, i.e., the damage states were not available to the candidate before submitting the damage detection findings to a collaborator. The results of these studies are presented in Chapter 3.
The next part of the thesis seeks to validate the robustness of the proposed damage detection methodology through experimental studies. Initially, a laboratory experiment is conducted on a 5.4 m simply supported model bridge structure loaded with a 4-axle model vehicle. In this study, rotation information is extracted from the acceleration data obtained from uniaxial accelerometers oriented in the bridge longitudinal direction. The methodology of obtaining rotation measurements using acceleration data is further elaborated within the scope of this study. To investigate a wide range of test scenarios without damaging the model bridge, it is stiffened locally by clamping steel plates onto the main girder flange, a process equivalent to applying ‘negative damage’. The results obtained from this study demonstrate that the proposed methodology can successfully identify ‘damage’ for all test cases. Having seen that it works well under laboratory conditions, the capability of the proposed damage detection methodology is tested on a full-scale railway bridge through a field-testing campaign, again using clamped plates to generate negative damage. Similar to the laboratory experimentation, the rotation response of the bridge to passing trains was recorded using accelerometers placed at the supports. At the test site, the axle weights of the passing trains were unknown. A site-specific procedure is developed which uses additional strain sensors to compensate for the contribution of different axle weights in the trains that crossed the healthy and stiffened bridges. The findings from the study show that the difference between rotation responses to a single axle load, normalised with respect to recorded strains for the healthy and stiffened states, suggests the location of the negative damage. The results obtained from the laboratory and field testing are presented in Chapters 4 and 5, respectively. -> Read more about the research by Farhad in TRUSS