Paper Key : IRJ************433
Author: Shravani Shreedhar Nagtilak
Date Published: 18 Oct 2023
Abstract
In this research, we present a transparent approach to using Convolutional Neural Networks (CNNs) for classifying vibration signals. The process starts by converting vibration data into image-like representations through the Short-Time Fourier Transform (STFT). We employ a CNN as the classification model, and to better understand how the model works, we use Gradient Class Activation (Grad-CAM) to visualize where the model focuses its attention during analysis. Vibration analysis has long been a valuable tool for diagnosing such faults, and the integration of Explainable Artificial Intelligence (XAI) techniques promises to enhance the accuracy, interpretability, and practicality of this critical task. This research paper proposes a novel approach that combines advanced machine learning models with XAI methods to analyze vibration signals for bearing fault diagnosis. By doing so we aim to not only improve the diagnostic accuracy but also provide insights into the reasoning behind the models decisions.