Resumen de Resilience oriented decision-making support technologies for infrastucture safety
Transport is essential to facilitate the movement of people, services and goods in all countries. This important role makes it essential to have policies that develop robust infrastructure networks, such as the Trans-European Transport Networks (TEN-T), considering all the infrastructures that comprise it as Critical Infrastructure Systems (CIS). As they are interdependent, a failure in one of them can trigger cascading failures in others. Therefore, these infrastructure systems must have the ability to resist, prevent and withstand potential hazards, absorb initial damage and recover to function normally, known as resilience. Ensuring the resilience of CIS involves assessing risks, threats, failures and breakdowns through analyses such as RAMS (Reliability, Availability, Maintainability and Safety). Recent important advances reflect the growing trend in the study of infrastructure resilience. Nevertheless, there is still a significant gap in the analysis of transport infrastructure performance by assessing the requirements of each of its levels (i.e., network, system and object) and identifying the specific components that compromise the resilience of the entire network. The general objective of this doctoral thesis is aimed to enhance transportation infrastructure resilience and ensure their impact on society through RAMS analysis, assessing requirements at all levels of the infrastructure using a top-down approach by applying current technology trends such as Artificial Intelligence (AI) to integrate data from numerous sensors and sources. The knowledge generated is intended to serve as a support base for improve decision-making in resilience-oriented maintenance for infrastructure safety. To achieve this, the thesis proposes six methodologies. The first one defines the general structure of the whole thesis, evaluating the requirements at all levels of the infrastructure (i.e., network, system and object) from a top-down approach. This methodology identifies risk areas in infrastructures through a RAMS analysis, taking into account various hazards. The other five methodologies, using the results of the first methodology, go deeper into each of these levels: three of them focus on the system level and two on the object level. Two of the system-level methodologies analyze how landslide hazards affect road safety by using an automated method to identify and inventory slopes along the road, assessing the areas that would be affected in the event of a rockfall, and defining a qualitative fault tree to analyze the loss of slope stability. The third of these system-level methodologies examines road safety through the development of accident prediction models. The two object-level methodologies focus on leveraging Digital Twin technology to enable proactive safety measures by assessing in parallel potential vehicle collisions with vulnerable road users, while also proposing the extraction of both trajectories using open-data from traffic cameras with AI vision technologies. The data used in all methodologies of this dissertation come from various levels of the infrastructure and from numerous sensors, with a focus on ensuring that all this information is openly accessible. The methodologies have been developed to be as standardized as possible, allowing them to be applied to any network, system or object level of any infrastructure. In addition, following current trends in artificial intelligence, technologies have been employed within the context of Intelligent Transportation Systems. This includes the use of Digital Twins enhanced with predictive models and AI vision frameworks. The workflows and methods presented have been validated using different datasets and tested in various real infrastructure scenarios. The results have contributed to the state of the art, leading to the publication of peer-reviewed scientific papers in three different scientific journals indexed in the Journal Citation Report (JCR) and three international congresses.
