Information shared fairly with all teams
Send technical or administrative questions to the HUMS2027 Committee. Anonymised questions and answers may be published here.
Where do I register for the Data Challenge?
Register through the official DSTG HUMS2027 Data Challenge page. The DSTG page is the authoritative portal for registration and dataset access.
Where do I download the dataset?
The restricted dataset is not provided as a public download from the HUMS conference website. After registering and accepting the Terms and Conditions through the official DSTG page, participants will receive a personalised dataset-download link.
Is there a fee to enter the Data Challenge?
There is no separate Data Challenge entry fee. However, at least one member of each team must register for and attend HUMS2027 to present the team’s results.
How do I ask a question?
Email DSTG.HUMSConference@defence.gov.au. Questions of general value may be answered on this page.
Has this dataset been publicly available before?
No. The test data were not released publicly and only a limited number of researchers had previously examined them. A curated dataset is now available to registered HUMS2027 Data Challenge participants through the official DSTG portal.
What must a team submit?
Each team will submit a concise results summary using the official template, including its methodology, earliest convincing detection, progression trend and supporting figures. Presentation slides will also be required for the conference session.
Can a team include members from several organisations?
Yes. Teams may be formed within one institution or across multiple institutions. One team may submit one entry.
Do all team members need to attend HUMS2027?
No. At least one team member must attend the conference and present the team’s results.
Can we use AI or machine learning?
Yes. There is no restriction on methodology. Physics-based, signal-processing, data-driven, AI and hybrid methods are all permitted.
Which tachometer should be used with each vibration channel?
Tach1 aligns with RR-1 and IP-2. Tach2 aligns with RF-3 and RL-4. The two tachometer channels are not mutually aligned.
The planet bearing defect frequencies seem inconsistent with the bearing dimensions. Which values are correct?
Thank you for pointing this out. In version 1 of the dataset description (April 2026), Eqs. (1) and (2) were written in terms of radii while the worked example used diameters, and the raceway diameters (48.97 and 61.94 mm) did not match the 12.97 mm roller. Version 2 (2 November 2026), now on the Dataset page, corrects this: the equations use the roller diameter and the pitch diameter, the raceway diameters are 42.49 mm (inner) and 68.43 mm (outer), and the pitch diameter is 55.46 mm, with 12 rollers. The defect frequencies are 120.0808 Hz (20.9404 carrier orders) for the inner race and 74.5613 Hz (13.0025 carrier orders) for the outer race, compared with 120.0828 Hz and 74.5593 Hz in version 1. The change is 0.002 Hz, well below the frequency resolution of a 15-second record (about 0.067 Hz), so it does not affect any analysis already done. The dataset is unchanged.