Predictive High-Speed Network Maintenance pilot strives to provide functionality that predicts the failure of mainline railway assets with sufficient foresight that preventative maintenance can be scheduled and performed. Improving the maintenance of rail assets consistently and predictably improves the safety case and other sub goals of the project such as cost efficiency and minimising disruption.
Unlike conventional “static” maintenance models in use today, the new proposed model shall be dynamically updated. The updates will be based on:
- information provided by traditional inspections (but with the aim of reducing them to a minimum);
- indirect information provided by traffic control systems (e.g. from the number of train passing on a given switch in a certain time it is possible to infer the switch stress);
- information provided by vehicles for track analysis and inspection;
- information related to weather (and possibly other environmental) conditions;
- information from capacity and/or infrastructure managers about utilization, traffic and route planning, possessions etc.;
- information from asset managers about maintenance and renewal plans and realization;
Furthermore, organizing and planning maintenance activities in the most efficient and safe way requires up to date information about traffic and capacity planning which implies close cooperation with infrastructure managers and train operating companies. In fact the mentioned information will be used in forecast models that will allow to establish more accurately the maintenance intervals and will also help to determine the consequences of any failure.
Finally, some of them will contribute to the analyses of possible step changes in component degradation that will be performed considering for example changes in traffic patterns, rolling stock, line speeds, weather conditions, and so on.
The project will showcase:
- Availability and accessibility of reliable and up to date information of infrastructure for operational purposes as capacity assignment, traffic planning, maintenance planning and preparation.
- Implementation of predictive models of decaying infrastructure valid for life cycle management and intelligent maintenance planning.
- Prediction of track profiles degradation:
Different sources from track profile degradation variables are used in order to collect the most important information that affects the track structure. These variables come from the analysis that the different maintenance companies perform, highlighting dynamic inspection, geometric inspection and maintenance task. Two different severity thresholds will be obtained from this analysis.
- Prediction of the degradation of point machines:
Different sources from point machine degradation variables are used in order to collect the most important information that affects the point-machines structure. These variables come from the analysis that the different maintenance companies perform, highlighting movements time, maintenance task and characteristic data.
- Optimization of railway operation in the Rail Traffic Management System.
By using the predictions obtained in the points descibed above, new data will be available to modify and optimize the railway traffic.
Indra has developed a new module to improve the infrastructure maintenance based on predictive information about failures probabilities. The benefits of predictive maintenance are wide-reaching and have positive repercussions on the safety, reliability and condition of the infrastructure.
Furthermore, the reduction in the number of intervetions passing form a preventive to a predictive model, has a direct impact in the reduction of maintenance costs and the operation disruptions. Other main benefit related to the end-user of the railway services is the improvement on passengers comfort.
Big Data Technology, Techniques and Algorithm used for the presented project are based on the SOFIA2 Platform (by Indra Sistemas, S.A.), a platform that brings an open source toolset for the Big Data exploit, based on the main standards and tools like Hadoop.
Sofia 2 community version provides a set of open APIs based on the main standards. As communication protocols Sofia 2 uses MQTT, RESTful, Ajax Push, Websocket, AMQP and JMS.
The interchange of information is based in the definition of ontologies, a semantic solution to define the information semantically, and to develop the Knowledge Processor from the data source. Sofia2 uses SSAP-Json and SSAP-XML as standards for the exchanging of information.
The modules of the platform Sofia 2 used in the construction of the pilot are:
- Storage - The information modelled in the Platform is stored in the Big Data Repository included in the platform. The reference implementation of this repository that is supported on Hadoop is used. Apache Hadoop is an open-source framework that allows the distributed processing of large amounts of data and working with machine clusters in a distributed way.
- Data Flow - is the main entry of data and information to the platform. This module is used as an ETL (Extract, Transform and Load) for both purposes: to intake data as for complex transformations within the platform and/or to export data involving intermediate transformations.
- The Notebook is an interactive and intuitive module from Sofia2 that allows to show the data and to facilitate its analysis.
- Machine Learning Module – is a platform that allow apply different learning techniques (predictions and regression). In the The Mahcine Learning platform, different learning technics have been studied to perform the best solution to determine the railway assets degradation. Some of the analytic models studied are Decision Tree Classifier, Random Forest Classifier, Extra Trees Classifier, Gradient Boosting Classifier,SVM or Neural Network Multi Layer Perceptron.
Using standard technology, we ensure the project evolution, without the need of change or modify the basic architecture to adapting to new communication mechanism.