Published October 20, 2026 | Version new version, fixed undetectable fault labels during inactive state of the component

HSU TwinFlow: A Living Benchmark for Evaluating Anomaly Detection and Diagnosis in CPPSs Using Digital-Twin-Generated Data

  • 1. Helmut Schmidt University, Hamburg, Germany
  • 2. Digital Twin Factory GmbH, Bad Oeynhausen, Germany
  • 3. Novatio Solutions GmbH, Monheim, Germany
  • 4. Institute of Software Engineering and Artificial Intelligence, TU Graz, Graz, Austria

Description

HSU TwinFlow is a digital-twin-generated dataset for benchmarking anomaly detection and fault diagnosis in cyber-physical production systems. It contains nominal and faulty simulation runs, including training and test scenarios, system measurements, fault labels, and configuration parameters.

Code, prior knowledge, baseline methods, and evaluation scripts are available on GitHub: https://github.com/imb-hsu/twinflow

This benchmark is published here:

Nemanja Hranisavljevic, Alexander Diedrich, Lukas Moddemann, Domenic Schaeffer, Frank Marek, Ingo Pill, and Oliver Niggemann.

“HSU TwinFlow: A Living Benchmark for Evaluating Data-Driven Methods for Anomaly Detection and Diagnosis in Cyber-Physical Production Systems Using Digital-Twin-Generated Data.”

In Proceedings of the 37th International Conference on Principles of Diagnosis and Resilient Systems (DX 2026), Cork, Ireland, 2026.

Files

test.zip

Files (91.7 MB)

Name Size
md5:15ea7162bed8c68fbf8000fa31e61d6f
46.6 MB Preview Download
md5:98d1bd82ec419b5df364ac3b16e900c3
45.1 MB Preview Download

Additional details

Related works

Is published in
Conference paper: 10.4230/OASIcs.DX.2026.7 (DOI)

Software