A machine learning methodology to predict alerts and maintenance interventions in roads

Abstract

This contribution is about predicting maintenance alerts in roads and selecting the most appropriate type of interventions recommended for preventing the occurrence of future failures. The objective is aligned with that covered by pavement maintenance decision support systems (PMDSS), though the methodology presented can be applied to other non-pavement road linear assets. The purpose is to summarise the main findings in the development of an approach based on testing the four most extended machine learning techniques (ML), namely Decision Trees (DT), K-Nearest Neighbourhood (KNN), Support Vector Machines (SVM) and Artificial Neural Networks (ANN), using data from the historical inventory of inspections and maintenance interventions of a case study to illustrate the potential that such approach can offer to road maintenance managers. The correlation process embodies supervised and unsupervised training of models. The maintenance predictions are presented and compared over various segments corresponding to the real maintenance interventions conducted on an existing road network of a geographical zone.

KEYWORDS:

• Road
• predictive maintenance
• machine learning
• linear infrastructure

Acknowledgements

This study has received funding from the European Union's Horizon 2020 Research and Innovation Programme (grant agreement n° 636496). Some of the authors express their gratitude to the Spanish Ministry of Economy and Competitiveness (MINECO) for the partial subsidy granted under the national R&D programme (TRA2015-65503) and the Torres Quevedo Programme (PTQ-13-06428). The authors acknowledge Infraestruturas de Portugal (IP) for making available the database used in this research. The content reflects only the authors’ view and it is stated that neither the EU nor MINECO is liable for any use that may be made of the information contained therein.

Disclosure statement

No conflict of interest was reported by the author(s).

ORCID

Luis M. Romero http://orcid.org/0000-0001-8963-9879

Francisco G. Benitez http://orcid.org/0000-0002-7712-1666

Additional information

Funding

This study has received funding from the European Union's Horizon 2020 Research and Innovation Programme (grant agreement n° 636496). Some of the authors express their gratitude to the Spanish Ministry of Economy and Competitiveness (MINECO) for the partial subsidy granted under the national R&D programme (TRA2015-65503) and the Torres Quevedo Programme (PTQ-13-06428). The authors acknowledge Infraestruturas de Portugal (IP) for making available the database used in this research. The content reflects only the authors’ view and it is stated that neither the EU nor MINECO is liable for any use that may be made of the information contained therein.