Acoustic Analysis of Cutting Tool Wear

This repo contains the code used to analyse the Cutting Tool Wear Dataset. You can find the dataset on Kaggle titled, "Cutting Tool Wear Audio Dataset".

The dataset is a collection of 1488 audio samples of End Mill Cutters of different levels of wear cutting through a piece of mild steel workpiece at two different RPMs using a Vertical Milling Machine. For more information about the dataset in particular, please refer to the Kaggle link. Please note, there was no addition of background noise to the dataset.

Steps to use the code

The code was tested with Python3.9 as the interpreter.

• Clone this repo.
• Install dependencies using the below command:

pip install -r requirements.txt

• Download the dataset from Kaggle.
• Make changes to the baseline.yaml file related to the location of the dataset and neural network features.
• run baseline.py with the following command:

python baseline.py

• After completion, a results.yaml will be generated in the results directory and an ROC Curve graph will be generated inside the model directory.

References and Citation

The code is heavily inspired by the MIMII_Baseline.

If this work renders helpful to you, please consider citing our paper:

@INPROCEEDINGS{10461855,
  author={Soni, Nachiket and Kumar, Amit and Patel, Hardik},
  booktitle={2023 IEEE 11th Region 10 Humanitarian Technology Conference (R10-HTC)}, 
  title={Acoustic Analysis of Cutting Tool Vibrations of Machines for Anomaly Detection and Predictive Maintenance}, 
  year={2023},
  pages={43-46},
  doi={10.1109/R10-HTC57504.2023.10461855}}