Now showing 1 - 1 of 1
  • Publication
    Profile Calculation and Bridge Damage Detection Using Vehicle-based Inertial Readings and the Fleet Monitoring Concept
    (University College Dublin. School of Civil Engineering, 2022) ;
    0000-0001-6426-9031
    The aim of this research is to use inertial vehicle sensor data to determine road and rail profiles and to monitor bridge condition. A novel fleet monitoring concept is developed to determine profiles and detect bridge damage using a fleet of instrumented vehicles. To improve the robustness of the calculation, a Bayesian updating method is used. To calculate the profile from vehicle response, a novel Inverse Newmark-Beta method is developed. Newmark-Beta allows vehicle acceleration to be calculated in response to an excitation such as a surface profile. Inverse Newmark-Beta finds the excitation corresponding to a known acceleration. For a single vehicle, the profile can be found if the vehicle properties are known. However, for a single vehicle, acceleration by itself is not enough to determine both profile and vehicle properties. Fortunately, a fleet of vehicles provides additional information that can be used to address this problem. To solve the fleet monitoring problem, the Inverse Newmark-Beta method is combined with the Cross Entropy (CE) optimisation method. Here, the road profile is calculated using accelerations from multiple vehicles, without prior knowledge of the vehicle properties. Sprung mass and half-car models are used to represent the vehicle and test this method separately. Numerical results show that the calculated profiles are the same as the ‘true’ profiles. The absolute values of the vehicle properties are not obtained but this algorithm can determine the relative values. Noise added to the accelerations has an influence on the calculated results. The fleet monitoring concept is used again to determine a flexible railway profile. The ‘apparent profile’(AP) of the railway track is defined as the true surface profile plus components of track deflection. Again, the Inverse Newmark-Beta method and CE optimisation are used together to solve this problem. Here, the train is simulated as a 4-axle carriage model and the railway track is represented by a beam supported on spaced sprung masses. The calculated AP of railway track is found to be very close to the true one. Since the previous method is sensitive to noise, the fleet monitoring concept is also solved using a Bayesian Updating method. The road profile is again determined using vehicle measurements. The calculated road profile is close to the true profile and is insensitive to noise in the simulated measurements. In addition, it can determine the relative vehicle properties at the same time. A 3-D ‘carpet’ road profile is also tested and shows good results. This thesis goes on to use similar principles of fleet monitoring to assess bridge condition. Firstly, a novel method is proposed to calculate the moving reference influence line (MR-IL), i.e., the deflection due to a moving (static) unit load at the (moving) location of that load. The results show that the MR-IL can indicate the condition of a bridge. The AP of a railway bridge is used to calculate the MR-IL. This numerical approach is assessed using a blind test operated by an independent research group. In the blind test, a frame structure is used to model the railway bridge and different levels of global damage are simulated. Using a 4-axle train carriage model, the damage levels of the bridge are inferred accurately. When a half car model is used to represent the train bogie, damage levels can be found again with less accuracy. The bridge damage is then detected using the Bayesian Updating method, with drive-by data. For local damage, the second moments of area of each segment of the bridge is updated as data becomes available. It is shown in simulations that estimates of the bridge second moments of area can be found even with local damage. The vehicle mass can be calculated. Bridge bearing damage is also simulated in this section. Using the Bayesian method, the value of bearing rotational spring stiffness, bridge second moments of area and vehicle masses can be calculated.
      10