Road infrastructure is one of the country’s most important and valuable assets. It is essential for the efficient growth of the economy and has to meet many requirements: On the one hand, citizens demand a reliable infrastructure to reduce the cost of running their cars, and on the other hand, public administrations try to manage their road budgets. as efficiently as possible. In addition, the quality of road infrastructure also affects tourism, transport and supply, urban development and citizen satisfaction.
Currently, this budget planning is done according to outdated methods and follows the logic of “preventive maintenance”, where the road maintenance intervention is planned manually by experts in the field based on information such as the date of the last intervention, available budget, manually detected road quality, road characteristics (geometry, height, slope). Against the backdrop of the challenging and important planning of road maintenance measures by decision-makers, the use case focuses on the implementation of a “predictive maintenance” model based on available data. Smart planning or smart maintenance of road infrastructure is therefore important to optimise road quality and allocate the available budget. The system collects data from various data sources that provide data on road quality levels and road usage. Current market solutions for predictive maintenance have been developed especially for the individual needs of clients and the type of data available.
Recent developments in maintenance modelling based on data-driven approaches such as machine learning (AI-based software) have enabled a wide range of applications. In the automotive industry, ensuring functional safety throughout the entire product life cycle while reducing maintenance costs has become a major challenge. One key approach to achieving this goal is predictive maintenance. Since modern vehicles come with a large amount of operational data, machine learning is an ideal candidate for predictive maintenance. Subsequently, we identify open challenges and discuss possible research directions.
We conclude that:
- Publicly available data would lead to enhanced research activities
- Most works rely on controlled methods requiring labelled data
- Combining multiple data sources can improve accuracy
- Using a deep learning method will continue to expand, but they require efficient and interpretable methods and the availability of large amounts of data
