Skip to content

IoT based TinyML anomaly detection for pumps using acoustic signals

Notifications You must be signed in to change notification settings

0xbadc0fee/ECE551B

Repository files navigation

ECE551B

Mini Pump Acoustic Dataset

This is a collection of audio recordings for use in acoustic classification and anomaly detection of for a submersible water pump.

Author: Silas Curfman
Course: ECE-551, University of New Mexico, Spring 2023
email: sgc98944@protonmail.com

Contents

File/Folder Description Items Format Size
0db No background noise, general information 5 n/a 531 MB
/data sub folders 5 n/a 527 MB
/data/raw original recordings, unedited, untrimmed 96 wav/audio 157 MB
/data/10_sec_splits all recordings, trimmed into 10 sec segments 5 wav/audio 212 MB
/data/train (14) rec of classes: idle, speed1, speed2, and dry_running 56 wav/audio 94.2 MB
/data/test (4-5) rec of classes: idle, speed1, speed2, and dry_running 19 wav/audio 28.2 MB
/data/anomaly (21) rec of class: cavitation for anomaly detection 21 wav/audio 34.9 MB

Hardware

  • Pump:
    • Mini Submersible Water Pump for Pond, Aquariums, Fish Tanks.
    • 50 GPH
    • Variable Speed
    • 3W, 110v / 60hz
  • Microphone:
    • Model: ONN USB Cardioid Microphone
    • Frequency: 20 Hz - 20 KHz
    • Sampling rate: 48 KHz / 16bit
    • Data Connection: USB
    • Mounting: Fixed Tripod, 5-8 inches from audio source

License

The data and associated code are being released under the Open Use of Data Agreement, with the intention of promoting open research using this dataset.

About

IoT based TinyML anomaly detection for pumps using acoustic signals

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published