Measuring workshop effectiveness at DB Regio
August 9, 2021
Emerson Hsieh
Felicitas Köck (Frachtwerk)

DB Regio, a division of Deutsche Bahn, serves over 1.9 billion passengers annually across its rail division. With a growing passenger volume and number of rolling stock in daily service, DB Regio faces the challenge of optimizing maintenance across its entire rail fleet to ensure timely operation across Germany.

Palifer worked with Frachtwerk GmbH on a project for DB Regio to demonstrate how door and air conditioning system defects can be predicted through insights derived from SAP maintenance data. We were able to develop specific maintenance recommendations and strategies from long-term trends that disrupt service.

Deriving Insights from SAP Maintenance Data

First, specific vehicle fault codes related to door and air conditioning failures were identified from data exported from maintenance workshops. The fault code data was cleaned and prepared using standard open-source libraries in Python. A fault code hierarchy was used to filter over four million data points, which include sensor logs and operator-generated reports, to ensure that only defects related to this project were selected. Frachtwerk enlisted the help of maintenance experts from DB Regio workshop to identify relevant faults where maintenance strategies can be applied.

The Palifer classifier was used to classify unstructured reports to their relevant fault codes and derive the mean-time between failures of specific components. For a better evaluation of the types of errors generated by the rolling stock, the team clustered data points based on fault code descriptions. The cluster analysis helped direct focus on consistent patterns across different workshops. One of the first significant findings was that fault code patterns of the same series of rolling stock result in different error patterns across different workshops, despite similar maintenance procedures and equipment across the same series.

Standardizing Maintenance Procedures Across Workshops

Based on the results of the data analysis, the project team was able to recommend targeted measures for optimizing processes and infrastructure for vehicle maintenance. The proposals were communicated with DB Regio workshops to ensure that the anomalies uncovered by the analysis are also reflected in the workshop’s current procedures. 

We were also able to identify specific reasons behind the inconsistent results of workshops. Specifically, we determined that the workshop in Munich outperformed other workshops due to how they handled sorted and organized fault codes from different rolling stock. We formulated additional measures for standardizing maintenance procedures across different workshops.

Failure Mode Patterns Across Workshops

Air conditioning malfunctions had evident seasonal fluctuations up to the end of 2019 when there was a change in maintenance strategy in summer 2020. The change can be seen in the graph above.


The door malfunction patterns can often be attributed to the weather and the number of passengers. Frachtwerk recognized an increasing trend of door malfunctions from 2018 to 2020. The COVID-19 pandemic also had an impact on decreased rail traffic and door failures.


A measure that can be implemented immediately is to have workshops exchange best practices among plants at which similar rolling stock are maintained. Recommendations for more efficient vehicle maintenance will be logged and shared on a common web-based platform. The implementation of these recommendations will also be monitored, after which the procedure with the most identifiable improvements will be scaled.

The goal of our cooperation with DB Regio is to predict fault code patterns with a greater degree of accuracy in the future from organized fault data, from which we can derive insights on standardizing and improving maintenance procedures just as we did with door and air conditioning faults in this project.

Contact us
Palifer helps operators evaluate maintenance approaches, predict material usage, and to benchmark supplier performance. We achieve this by using AI and natural language processing on your existing maintenance data.