Luis M. Romero

Automatic Prediction of Maintenance Intervention Types in Roads using Machine Learning and Historical Records

A methodology to support and automate the prediction of maintenance intervention alerts in transport linear-asset infrastructures can greatly aid maintenance planning and management. This paper proposes a methodology combining the current and predicted conditions of the assets, and unit components of the infrastructure, with operational and historical maintenance data, to derive information about maintenance interventions needed to avoid later severe degradation. By means of data analytics and machine learning techniques, the proposed methodology generates a prioritized listing, ranked on severity levels, corresponding to the pre-alerts and alerts generated for all assets of the transport infrastructure. The methodology is applied and tested in a real case consisting of a road network with different section classes. The analysis of the results shows that the algorithms and tools developed have good predicting capabilities.

Historical maintenance relevant information road-map for a self-learning maintenance prediction procedural approach

A large percentage of transport infrastructures are composed of linear assets, such as roads and rail tracks. The large social and economic relevance of these constructions force the stakeholders to ensure a prolonged health/durability. Even though, inevitable malfunctioning, breaking down, and out-of-service periods arise randomly during the life cycle of the infrastructure. Predictive maintenance techniques tend to diminish the appearance of unpredicted failures and the execution of needed corrective interventions, envisaging the adequate interventions to be conducted before failures show up. This communication presents: i) A procedural approach, to be conducted, in order to collect the relevant information regarding the evolving state condition of the assets involved in all maintenance interventions; this reported and stored information constitutes a rich historical data base to train Machine Learning algorithms in order to generate reliable predictions of the interventions to be carried out in further time scenarios. ii) A schematic flow chart of the automatic learning procedure. iii) Self-learning rules from automatic learning from false positive/negatives. The description, testing, automatic learning approach and the outcomes of a pilot case are presented; finally some conclusions are outlined regarding the methodology proposed for improving the self-learning predictive capability.

Bilevel O/D Matrix Adjustment Formulation Using High Convergence Assignment Methods

The Frank-Wolfe algorithm has been for years the most widely used method for solving the traffic assignment problem (TAP). In the last decade there have been new proposals for the resolution of the TAP. It has been shown that these algorithms are feasible for large scale problems with very high convergence, much higher than the achieved by the Frank-Wolfe algorithm. The O/D matrix adjustment problem based upon traffic counts can be formulated as a bilevel optimization problem in which the TAP is the lower level. The convergence of the TAP and the computational cost can be critical because the number of TAPs to be solved during each step of the process is very high. This paper exploits the possibilities offered by new TAP methods in the O/D matrix adjustment problem. Numerical examples on medium-sized networks using the new proposed methods are presented.

Adjustment of Origin-Destination Matrices Based on Traffic Counts and Bootstrapping Confidence Intervals

Mobility studies require, as a preliminary step, that a survey of a sample of users of the transportation system be conducted. The statistical reliability of the data determines the goodness of the results and the conclusions that can be inferred from the analyses and models generated. Because of the high costs of collection, data are partially reused in either a disaggregated or an aggregated manner. In the first case, statistical reliability is not always guaranteed; this condition affects the results that will be derived from projections and estimates of future hypothetical scenarios. A methodology is presented: it is based on bootstrapping techniques and is used for robust statistical estimation of mobility matrices. Confidence intervals of travel between origin–destination pairs defined by each matrix cell derived from a survey are generated. The result is applicable to defining the dimensions of certainty for matrix cells and subsequent adjustment by techniques based on aggregate data (e.g., traffic counts, cordon line matrices, paths). A statistically reliable data mobility study conducted in Spain at a regional level is used. Results derived from disaggregating data at an interprovincial level are presented, along with an application to the posterior mobility matrix adjustment based on traffic count data. The study results demonstrate the potential of the methodology developed and the usefulness of the conclusions.

Traffic Flow In Dense Urban Areas By Continuum Modelling

This paper considers a new approach for modelling the tratlic flow in dense urban areas by using a mathematical continuum model, which needs information data from a very limited number of points from the network. T’he purpose of this communication is to provide a mathematical foundation for the study of this approach. This methodolo~ can SPA up the computation of relevant values concerning the time evolution of traffic on a city network.