CAF – Inspection of vehicles in operation from infrastructure
CAF
Sector: Railway
Business Case
The control of the condition of some suspension parameters and rail rolling stock requires periodic reviews at a significant labour cost. Sensorising all trains is not a viable cost/return solution. A solution is proposed in which the sensorised track in a location where the trains pass through in operation collects data that, when processed, generates the KPIs of the status of these elements.
Objetivos
Design an installation that can be located next to the track, that is connected to a Cloud infrastructure and that allows the integration of additional functions on the same electronics base. Minimise the information to be recorded, the raw data are preprocessed in the electronics and only the KPI data are collected in the cloud infrastructure, traceability of the elements associated with these data, date and time.
Use case
Requirements: the measurement parameters to be obtained, accuracy, sensor strategy in the installation are defined. A digital twin of the installation is developed: mechanical model of the track, train model, virtual sensors integrated in the model. Scenarios are defined to generate synthetic signal to train the models. We work on methodologies that maximise accuracy at minimum cost
Infraestructura
Hw: for sensor data logging and preprocessing. Cloud: Pre-weighed information recording and dashboards
Tecnologías utilizadas
Automatic or Deep Learning
Datos utilizados
Synthetic time series by digital twin. Time series on prototype installation Validation of KPIs obtained from manual measurements (original methodology)
Recursos utilizados
CAF R&D: methodology design and technical leadership of the project CETEST: implementation of the instrumentation, electronic integration, software and digitisation solution CEIT: doctoral thesis researcher for the University of Navarre to solve the technological challenges of the project.
Dificultades y aprendizaje
The most complex aspect was to define a methodology that achieves sufficient quantitative accuracy with a limited number of sensors while minimising the design and maintenance of the rail infrastructure. Once the methodology was defined, the training of the KPI estimation models under study was resolved by generating synthetic signals using the physical model and a combination of different scenarios.
KPIs (impacto en el negocio y métricas del modelo)
A single installation can measure all trains (whether from the same fleet or from different fleets) passing through that infrastructure. Continuous monitoring by train or fleet trends, in addition to still photographs obtained through periodic inspections. Saves time and resources by avoiding repetitive maintenance tasks through automation and digitisation of data recording.
Financiación
Industrial Doctorate for basic research activity and methodology proposal financed by the Bikaintek program of the Basque Government.
Colaboradores
Metro de Bilbao provided its infrastructure to validate the prototype in exchange for access to the results. Multidisciplinary work team: CAF Group (CAF R&D, CETEST) and CEIT.
