This repository contains the Jupyter Notebooks and supporting code for the paper accepted at the 2025 Advances in Science and Engineering Technology International Conferences (ASET).
Authors: Srijesh Pillai, Yodhin Agarwal, Zaheeruddin Ahmed
Paper Link: https://arxiv.org/abs/2509.11075
Audio-based equipment condition monitoring suffers from a lack of standardized methodologies for algorithm selection, hindering reproducible research. This paper addresses this gap by introducing a comprehensive framework for the systematic and statistically rigorous evaluation of machine learning models. Leveraging a rich 127-feature set, our methodology is validated on both synthetic and real-world datasets. Results demonstrate that an ensemble method achieves superior performance (94.2% accuracy), with statistical testing confirming its significant outperformance of individual algorithms. This work provides a validated benchmarking protocol for selecting robust monitoring solutions in industrial settings.
Our framework systematically evaluates six machine learning algorithms across multiple performance dimensions. The ensemble method consistently demonstrates superior accuracy and robustness to noise.
| Overall Performance (F1-Score) | Noise Robustness Analysis (F1-Score) |
|---|---|
 |
 |
This project proposes and implements a complete, end-to-end pipeline for the fair and reproducible benchmarking of classification algorithms for audio-based fault detection.
Clone this repository to your local machine:
git clone https://github.com/YOUR_USERNAME/audio-condition-monitoring-framework.git
cd audio-condition-monitoring-frameworkThe analysis requires the custom_industrial_dataset_127features.csv file. Due to its size, it is not included in this repository. Please place the file inside the data/ directory.
It is recommended to use a virtual environment. Install all required libraries using the requirements.txt file:
pip install -r requirements.txtThe analysis is contained in the Jupyter Notebook: model_analysis.ipynb. This notebook contains the core logic for loading the data, training all six models, evaluating their performance, and generating the primary results tables.