ABSTRACT
Limited maintenance budgets, coupled with an ageing bridge-stock, mean that there is much appetite for efficient and inexpensive techniques for monitoring and detection of damage in bridges. It is not feasible or cost-effective to install individual Structural Health Monitoring systems on every bridge on the transport network and, as such, the concept of using sensors located within vehicles to monitor the condition of bridges has become the focus of much attention in recent years. This paper presents a drive-by bridge monitoring concept which utilizes in-vehicle measurements to infer bridge behavior, allowing changes in bridge condition to be monitored over time. Machine learning techniques are leveraged to allow the influence of vehicle speed to be considered. The proposed approach is shown to be capable of detecting midspan cracking and increased rotational stiffness at the support due to seized bearings. The use of a data-driven machine learning algorithm provides much more robust damage detection capabilities compared to existing drive-by approaches. The application of machine learning shows promising results, with a primary advantage of such an approach being the ability to train the algorithm to recognize and account for various external and environmental factors which typically cause difficulty when attempting to identify the dynamic properties of a bridge from a measured in-vehicle response. It is expected that this approach can be implemented on a large scale and act as an early warning tool for infrastructure owners to identify bridges which may be presenting signs of distress or structural deterioration.
