Data-driven Transport Infrastructure Maintenance

Abstract

What we did This report assesses the potential of data-driven approaches to improving transport infrastructure maintenance. It looks at trends in maintenance strategies, explores how the targeted use of data could make them more effective for different types of transport infrastructure, and looks into implications for policy. The report builds on discussions held during workshops with members of the International Transport Forum's Corporate Partnership Board. What we found Maintenance constitutes an inevitable, albeit often invisible, part of countries' transport policies. Increased demand for transport infrastructure accelerates infrastructure's ageing. The effects of climate change further aggravate this. Unsurprisingly, many governments look for transport infrastructure maintenance policies that provide better value for money than current practices offer. Infrastructure maintenance strategies are gradually shifting towards data-driven approaches. They exploit the power of digital technologies, Big Data analytics and advanced forecasting methodologies. Data-driven approaches have gained momentum in transport infrastructure maintenance as a result of four simultaneous technological innovations. First, the development of digital technologies has resulted in the digitalisation of society, industry and transport, which facilitates data sharing. Second, computing technologies have provided the necessary horsepower for running the digital infrastructure. Third, the Internet of Things and sensor technology have increased the potential for automating reporting from sensors that capture and measure new phenomena and provide data sets that flow through digital infrastructures. Fourth, artificial intelligence (AI) has helped to extract information from vast amounts of data, recognising patterns beyond the capacity of individual observation and exploiting digital infrastructure and computing power. Policy makers are beginning to leverage these developments in various ways. Data-driven maintenance is becoming common in many parts of the transport industry. Railroads collect massive amounts of inspection data from different sources using various methods, such as track inspection cars and drones that gather data to model track degradation. However, the rail sector faces numerous challenges for applying Big Data analysis: a lack of specific data analysis tools, high cost of involving stakeholders and heterogeneous data sources. Also, the algorithms currently used to predict the wear of rail infrastructure only work under lab conditions. For road infrastructure, various automated inspection methods exist. These include vision-based methods, laser scanning, ground penetration radar and a combination of these. All are accurate and effective but usually costly. As a result, the coverage and collection frequency can prove insufficient for detectingchanging road conditions. Several pilot studies have tried to use smartphones to collect data on the state of roads to reduce deployment costs for data-driven maintenance. At airports, the demand for accurate real-time data has spawned systems that automatically acquire and process infrastructure data. Advanced technologies now register when deformities develop on runways. They accurately measure moisture levels, temperature, strain and other factors relevant to wear and degradation. Several airports have built, or plan to build, concrete pavements with embedded strain gauges and other sensors to monitor the stress in the material caused by aircraft. Overall, data-driven approaches to infrastructure maintenance promise to enhance fact-based decision making and capabilities to predict the remaining useful life of assets. They can also improve cost efficiency and environmental sustainability. However, some new challenges need to be addressed, notably for the use of AI. AI predicts future behaviour based on historical data. Yet all predictions can prove incorrect where events do not follow past trends. What we recommendScale up and speed up the deployment of data-driven approaches to transport infrastructure maintenance Transport infrastructure maintenance could benefit from a broader and accelerated roll-out of data-driven approaches. These could improve the quality of assets, enhance the life cycles and save costs - especially when the relevant technologies are well-known, such as sensor technologies. In some cases, more tests and pilot projects will be useful, notably where leveraging data technologies for more effective maintenance policies poses specific challenges, as is the case of artificial intelligence in the railway sector. Update regulation and guidelines for transport infrastructure maintenance to facilitate the introduction of more data-driven approaches Current regulations and guidelines apply to condition-based maintenance strategies. These may set requirements that are ill-adapted to data-driven approaches to maintenance and may hamper their roll-out. Policy makers should ensure that the policies applied to data-driven approaches do not stifle their potential benefits. Ensure data-driven infrastructure maintenance approaches follow good practices in data governance The use of data in infrastructure maintenance must be in line with privacy protection laws and regulations. All data should be anonymised and encrypted. Location and trajectory data should be covered by the most robust protection methods, as they create the severest vulnerabilities for citizens. Tools to limit privacy risks include non-disclosure agreements between data users and providers, the involvement of trusted third parties to conduct the data collection and the development of "safe answers" approaches, in which only query results are exchanged instead of raw data. Governments could also broker data-sharing partnerships for the purpose of data-driven maintenance, for instance, between data providers and infrastructure managers. However, it may want to limit such partnerships to data of public interest and require purpose specificity and data minimisation.