Pantograph-Scanner for Predictive Maintenance Management in Railway Operation Use-cases.

Pantograph-Scanner to extract features from a specific component of a pantograph from within an image to provide measurement data to be integrated into the operational business process.

Context

In an integrated Railway System, such as at Swiss Federal Railways (SBB), there exist use-cases in the interaction of rolling stock with rail infrastructure, which are of high interest to monitor the condition of the involved technical systems. Condition Monitoring as a discipline helps to manage failure-modes appropriately and helps to prevent failure-modes to occur, by applying specific counter measures (maintenance tasks) to them. Predicting a certain failure-mode to happen is the capability to execute the maintenance tasks predefined at the right moment, yet before the failure-mode is occurring.
Those concepts are well known for decades. The challenge lays therein, that operators, manufacturers, OEMs, maintenance have not yet agreed in a failure-data set for the Rail industry, something that in contrary exists for oil and gas, wind turbines, nuclear facilities, etc. So generally spoken, we are on a journey toward fulfilling that vision for Rail. But – we would like to share our experience to a broader community and start with a first example and model(s) that we are going to contribute.

Content

We provide a set of images taken from a roof top of a specific rolling stock fleet, which have been taken using high resolution cameras; the current use-case under discussion is about the sliding element of the pantograph of the coach of a trainset, which interacts with the catenary copper overhead line from the Rail infrastructure.
We managed in a two years field test to proof of concept using an in-house engineered system to measure the thickness of sliding element reliably. The goal of the ongoing initiative is not only to show, that it works, but provide ready-to-build-upon models for the Rail industry, being operators, maintainers, etc. The model under discussion has been trained for a specific type of pantograph and sliding elements. Publishing of the model is planned in Q1 2024 for peer-review and for a broader public in Q2 2024.

Urs Gehrig, March 2024