Context
Train doors are an important part of the passenger experience. But this equipment is put to the test by passengers and the weather.
After repeatedly observing malfunctions during the opening and/or closing of doors on several TGVs, the company’s agents expressed the wish to be able to automate the detection of faults on these key elements.
Indeed, if a fault occurs during door opening or closing, this represents a risk to passenger safety and requires rapid, targeted intervention by maintenance staff.
Need
Following this observation, the customer wanted to be able to generate an anomaly detection model for train doors that would be able to identify the origin of faults and give the root causes, enabling maintenance operators to intervene quickly and efficiently.
The information provided by the detector will thus help to improve the safety and reliability of the equipment, enhancing user comfort while optimizing maintenance operations.
Solution
DiagFit software was used to analyze the time series extracted from the position of the door, and the current and voltage sensors of the door motors. The data came from a sample of 84 trains, each comprising 5 vehicles equipped with 4 doors.
Following the analysis, it was possible to establish a database representing the nominal behavior of the train doors, and thus generate a fault detection model.
Resultat
DiagFit analyzed opening and closing cycles to highlight abnormal ones. Thanks to these results, it was possible to identify suspicious systems.
In addition, thanks to the ability to compare results from several doors on the same wagon, it was possible to discriminate between failures linked to an external event common to all 4 doors (bad weather, for example) and failures caused on a single door (by a user, for example).
To date, some of DiagFit’s results have yet to be compared with the reality on the ground as reported by TGV agents, in order to refine the detection model and associated metrics.
