MBHM: Multimodal Bearing Health Data from 9 Public Sources

MBHM is a multimodal dataset for bearing health management created by SIA-IDE and released in 2024. It integrates vibration signals and a health management text corpus derived from 9 distinct public datasets to support industrial diagnostic research.

Use Cases

• Question answering regarding bearing health using the health management corpus
• Token classification for identifying technical entities in maintenance text
• Multimodal fault diagnosis linking vibration patterns to textual descriptions

Strengths

• First multimodal dataset in the bearing health domain
• Aggregates data from 9 distinct public vibration datasets
• Supports both sensor-based fault diagnosis and NLP tasks

Limitations

• Relies on secondary data sources rather than original sensor collection
• Specific column headers and data schema are not explicitly listed in the metadata

Provenance

Source

SIA-IDE (Arxiv:2408.11281)

Collection Method

Aggregated from 9 public datasets and processed for multimodal alignment

Freshness

Last updated April 2025; based on 2024 research.