The artificial intelligence embedded in the sensor provides the following advantages:
Yumain’s fully scalable automated pantograph inspection system, using an AI (Deep Learning) based architecture, provides real-time alerts of faulty pantographs to any rail network operator. At each inspection point, the fixed sensor(s) capture high-resolution digital video images of each piece of equipment. At the same time, the engine number is also read automatically. Real-time analysis (< 5 minutes) is then performed, and an alarm is then generated if necessary, and transmitted by email.
The device for automatic pantograph analysis is based on
on combining the use of robust and mature algorithms (RCNN mask),
as well as filters applied by image processing.
The Yumain solution is designed to be robust and industrial in order to respond to the major problem of overhead lines being torn out by defective pantographs.
Due to the compactness and simplicity of our solution, installation, calibration and implementation can be carried out by 2 people, and does not require any special equipment or machinery. This allows to reduce
considerably the cost of installation.
As the tariff per measuring point is up to ten times lower than more complex solutions, it allows a multiplication of control points (e.g. station entrances) on the national rail network.
To detect the non-conformity of a pantograph (read or not) on a train running under a sensor, the system will rely on the combination of image analysis and on-board algorithms based on artificial intelligence.
The arrival of the train will be identified either by image analysis (Sigma/Delta technology) or by an external sensor (the device allows 2 inputs per dry contact).
The train can be identified either by:
- Reconciliation of the time-stamped event with the operator’s train schedule database.
- The location and reading of the train identification plate (if located on the front)
- Any other (radio) signal to authenticate the train
The location and analysis of all the pantographs (deployed or not) present on the engines is fully automatic and in real time (< 5 minutes). If an anomaly is identified, an alarm is immediately generated in the form of an alarm:
- dry contact (2 outputs) to supply a local PLC.
- a time-stamped email with photos of the faults
- a complete message via FTP and Ethernet or wifi connection.
This information can also be stored and time-stamped for statistical purposes, but also to feed useful indicators to the QSE manager.
This solution, due to its fanless design, does not require any maintenance.
This device includes all the elements necessary for its operation and installation: “Edge Computing” solution. In its basic version, it only requires an external power supply + connection to a 4G modem. A local connection to a customer’s processing unit is also possible.
Due to the different possible layout configurations, the devices can be powered in several ways:
- By permanent electrical network (240 VAC)
- By a continuous power supply
- By solar panel
The sensor is intended for outdoor use (IP 57).
Night use of the device is not yet possible at this time
Thanks to technical developments, and the choice of a universal support type Ram-Mount™ (https://www.ram-mount.fr), the sensor can be installed on various supports.
Installation on a mast or on a bridge are the most commonly adapted solutions. Installation on a signalling gantry is also possible.
Hello, how can we help you?
A project, a question, our AI specialists are there to bring you the solutions
We will contact you immediately.