Run-to-failure vibration dataset of 3D-printed robotic grippers

Vibration recordings of the complete life cycle of three 3D-printed robotic grippers (a parallel gripper fitted to EEZYbotARM Mk2 arms) under accelerated degradation, collected on a laboratory production line at the Faculty of Engineering of the University of Porto (FEUP). Each gripper was monitored by a different sensor: an ADXL345 accelerometer (3200 Hz), an MPU9250 accelerometer (4000 Hz), and a piezoelectric vibration sensor (~2800 Hz). The data covers 22 sessions of about one hour (~21 h of operation): 16 healthy sessions, five sessions in which faults were progressively introduced, and the complete failure. Files: one Parquet file per session and sensor (ARM1_adxl345.zip, ARM2_piezo.zip, ARM3_mpu9250.zip), per-session metadata (sessions.csv), and the 1-s window features used in the accompanying work (features.zip). README.md describes the protocol, the file formats, and the known issues of the recordings (in particular the replacement of the piezoelectric sensor after session 14); please read it before using the data. Code, trained models, and results: https://github.com/pedrodsmota/iiot-autonomous-maintenance