Please use this identifier to cite or link to this item:
Title: Applications of Deep Learning: Sequence Modelling in Industrial Sound Analytics
Authors: Borwankar, Prabhanjan
Dept. of Data Science
Keywords: Data Science
Machine Learning (ML)
Neural Network Architectures
RNN or Recurrent Neural Network
CNN or Convolutional Neural Network
LSTM or Long Short Term Memory networks
Confusion matrix
Issue Date: Jun-2021
Citation: 48
Abstract: In today's world of data and information, deep learning has proven to be the ace in data science applications. While we hear much of deep learning concerning computer vision, NLP (Natural Language Processing), audio analysis and its usefulness to ML have been recognised in the past decade. Audio data analysis analyses and understands audio signals captured by digital devices such as microphones. It has widespread applications, including automatic speech recognition (ASR), digital signal processing, medical diagnostics, music classification, tagging and generation. Factory machinery is prone to malfunction or breakdown, resulting in a significant loss of time, money and resources. So, there is a need for an efficient and affordable solution to this problem. So, there is a rising interest in building deep learning models for monitoring machines. The Industrial Internet of Things (IoT) and data-driven techniques have revolutionised the manufacturing industry. Many new and innovative approaches have been attempted to build algorithms to monitor machinery. Many new and innovative approaches have been attempted to build algorithms to monitor machinery. Sound Classification is one of the most popular applications of Deep Learning. It involves classifying sounds based on various frameworks like music genre prediction, urban sound classification, identification of speaker or tone from sound clips. It has huge applications, especially in manufacturing industries, like predicting machine failures, spotting faulty pieces of machinery.
Description: The purpose of this project is to try and build a neural network architecture that can classify different sounds accurately. This classification will, in turn, be used to classify the sounds produced by machines at different conditions, like when it is functioning well when it starts to dis-function to the point where it is not at all working what it is required to perform. This particular application of ML is known as Predictive Maintenance of Machines. Specific to A2IoT, the architecture built will be used to classify sounds from a factory environment among the background industrial noise. The condition is to detect whether there is a gas leakage or not at the desired location.
Appears in Collections:MS THESES

Files in This Item:
File Description SizeFormat 
20141150_Abhishek_Choudhary_MS_Thesis.pdfMS Thesis1.66 MBAdobe PDFView/Open

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.