---
license: mit
task_categories:
- time-series-forecasting
- tabular-classification
- robotics
tags:
- predictive-maintenance
- motor
- fault-diagnosis
- industrial-iot
- twinnable-agent
pretty_name: MCC5-THU Motor Replay Dataset
---

# MCC5-THU Motor Replay Dataset

This Praxis mirror contains processed replay Parquet files derived from the author-maintained MCC5-THU Motor dataset mirror.

## Source and Attribution

- Author HuggingFace mirror: `Samlzy/MCC5-THU-Motor`
- Source card Mendeley DOI: `10.17632/6s3dggj9mw.1`
- Source title: Multi-mode Fault Diagnosis Datasets of Three-phase Asynchronous Motor Under Variable Working Conditions

The full-rate raw archives remain available from the author mirror and are backed up in the Praxis Google Drive dataset archive. This repository stores the normalized replay representation used by TwinnableAgent ingestion tests and state-tick demos.

## Processing

The source CSVs are full-rate 12.8 kHz headerless recordings. The Praxis replay representation samples one row per second from every source run, preserving all run/fault/condition coverage while keeping ingestion lightweight.

## Files

- `data/train.parquet`: deterministic split by source run.
- `data/val.parquet`: deterministic validation split by source run.
- `metadata.json`: source, schema, row-count, and split metadata.

## Schema

Replay columns:

- `unit_id`
- `cycle`
- `source_archive`
- `source_file`
- `fault_type`
- `operating_mode`
- `torque_setpoint_nm`
- `rpm_setpoint`

Feature columns:

- `speed_key_phase`
- `torque_nm`
- `motor_vibration_x_0_1g`
- `motor_vibration_y_0_1g`
- `motor_vibration_z_0_1g`
- `motor_current_a_0_1a`
- `motor_current_b_0_1a`
- `motor_current_c_0_1a`

## Praxis Role

Dataset 9 / KAIST replacement: industrial three-phase motor fault-diagnosis coverage with torque, current, and vibration signals under variable operating conditions.