Intelligent Innovative Smart Maintenance of Assets by integ
Intelligent Innovative Smart Maintenance of Assets by integ. Rated Technologies FINAL EVENT, Naples, 10 th October 2019 TRACK CIRCUITS MONITORING SYSTEM This project has received funding from the Shift 2 Rail Joint Undertaking under the European Union’s Horizon 2020 Research and Innovation Programme (Grant Agreement No. 730569)
Demonstrator Objectives The objective of the work of STS inside IN 2 SMART scope is the application of the Intelligent Assets Management System (IAMS) to the support and the improvement of the Track Circuit Systems (TCS) maintenance process. In the analysed scenario, TCS maintenance is so far based on corrective activities and preventive actions carried out periodically. On one side, corrective maintenance aims to restore functionalities of an asset only once it failed, on the other side, preventive maintenance consists in a set of scheduled actions in order to avoid asset degradation. The introduction of the new platform enabled the collection of assets data from the field and the development of datadriven models enabling diagnostics and prognostics functionalities. In other words, the deployment of IAMS allowed a first shift from a standard maintenance approach to a reactive/predictive one. More in details, assets involved belong to a urban line where Hitachi Rail STS developed the signalling system and is in charge of maintenance management. In this context, the developed application addresses the mitigation of false track occupancies. The functional architecture of the developed platform is based on 4 levels: • Gateways and data collection: interfaces and procedures for data acquisition within the platform. • Data processing: operations for data cleaning, standardization, filtering and storage. • Data analytics: analysis algorithms with the aim of extracting knowledge from the data. • Visualization: HMI to provide extracted information to the final user in an intuitive way. Track Circuits False Occupancies One of the main issues related to track circuits is the false track occupancy phenomena, during which the track is erroneously considered in occupied state, due to malfunctions or external conditions that may affect the correct behaviour. The false track occupancy is a critical failure for a railway line because it has relevant consequences for trains’ circulation: this event usually implies the reduction of service availability. Alternatively, a massive use of preventive maintenance to avoid false occupancies occurrences. For this reason, the main goal of the new platform is to support maintenance decisions: on one side, the aim is the prioritization of track circuits’ maintenance interventions on the basis of the status computed through the acquired data in order to act before assets fails, on the other side, the aim is to support the diagnosis process after a fail occurs. Thus, the new maintenance framework is able to maximize the service reliability and to optimize both usage of resources and available time, avoiding contractual penalties and delays for passengers.
Data Sources and Collection • TCS devices parameters. Collection of measures of TCS functional (e. g. Shunt Level, Variance and Receiver Level). The data are acquired every minute. • CCS logs. Collection of events and alarms logs from the central system. These large sets of logs allow to extract information about alarms related to TCS, trains movements events and alarms related to other components. The data are acquired every 5 seconds. • Maintenance Reports. • Weather Data. historical data collected, a data-driven approach has been used to develop a model in order to identify anomalous behaviour of the assets. In this context, the anomaly is represented by a particular TCS status in which the asset is characterized by some malfunction but still not affected by false track occupancy phenomena. Thus, the proposed anomaly detection model should be able to predict in reasonable advance if a track circuit will go in a false occupied state. The exploited approach automatically generated a specific model instance for each track circuit, fitted to its specific behaviour, characteristics and working conditions. Final Workflow Data Processing and Models for Predictive Maintenance For all considered data sources, extraction, cleaning, aligning and formatting steps have been developed. Moreover, exploiting the All the functionalities described above are deployed on the platform following the architecture depicted in the picture above and characterized by a structure in layers: • Connection layer: deployed gateways to pull data from the systems on the field. • Speed layer: near real-time operations including cleaning, aligning and formatting steps, TCS status computation and anomaly detection output computation. • Batch Layer: batch operations for additional data standardization, low priority processes, historical data storage and data aggregation. • Serving Layer: data is prepared to be consumed with a set of specific operations in order to push data with the required format on a specific database.
The Human Machine Interface developed allows the visualization of the data collected and the analysis output described in the previous sections; different levels of user groups have been defined according to their roles. More specifically: • Operative: the user can monitor the status of the assets involved in real time, through the display of the current values of the parameters involved and their recent performance. From this view a specific field identifies the presence of an anomalous status for a track circuit. • Tactical: the user can dynamically access the data log of the track circuit parameters for tactical level analysis. The historical trend of the parameters of interest can be displayed, and filtered by track ID, date and at station level. The view also allows comparisons between assets. • Strategic: the user views the distribution of alarms and events of interest generated by CCS; the information is updated in real time and allows for considerations to be drawn in relation to alarms on different levels: § distribution throughout the line; § distribution by asset, date, time, station; § frequency of the different alarm types.
- Slides: 4