Login | Register

A Comprehensive Study on Tool Condition Monitoring Using Time-Frequency Transformation and Artificial Intelligence

Title:

A Comprehensive Study on Tool Condition Monitoring Using Time-Frequency Transformation and Artificial Intelligence

Soltani Rad, Javad (2015) A Comprehensive Study on Tool Condition Monitoring Using Time-Frequency Transformation and Artificial Intelligence. Masters thesis, Concordia University.

This is the latest version of this item.

[thumbnail of SOLTANIRAD_MASc_S2015.pdf]
Preview
Text (application/pdf)
SOLTANIRAD_MASc_S2015.pdf - Accepted Version
11MB

Abstract

Tool failure is one of the probable faults during machining process which may cause unscheduled downtime and damage of tools, machines and work pieces. Therefore, developing an accurate and reliable online tool condition monitoring (TCM) system is in high demand. This research investigates TCM using time-frequency transformation methods and artificial intelligence. Multi-sensory monitoring systems and sensor fusion are investigated in the first step. Many different sensors at various locations are tested to determine the best input sets with most complimentary information. Three data fusion techniques 1) feature level, 2) score level, and 3) decision level are implemented and compared in this step. The result suggests that score level data fusion is superior for this application. Moreover, five advanced time-frequency transformation methods are employed due to superior ability of time-frequency transformation to reveal time variant characteristics of a signal as well as its frequency components. S-transform demonstrates the most accurate results among these methods. This research also proposes a novel feature extraction method to select the most discriminative information and reduce data’s dimensionality and calculation cost. This method selects a local region of data in time-frequency domain using genetic algorithm optimization. The proposed method is also combined with 2D principal component analysis which has improved the systems in terms of accuracy and performance. Finally, three well-known artificial intelligence methods 1) multi-layer perceptron artificial neural network, 2) radial basis function artificial neural network and 3) adaptive neuro-fuzzy inference system are applied to find a model between extracted features and system fault. Based on the results, radial basis function has the minimum mean error and adaptive neuro-fuzzy produces the lowest maximum error.

Divisions:Concordia University > Gina Cody School of Engineering and Computer Science > Mechanical and Industrial Engineering
Item Type:Thesis (Masters)
Authors:Soltani Rad, Javad
Institution:Concordia University
Degree Name:M.A. Sc.
Program:Mechanical Engineering
Date:15 April 2015
ID Code:980007
Deposited By: JAVAD SOLTANI RAD
Deposited On:13 Jul 2015 18:51
Last Modified:18 Jan 2018 17:50

Available Versions of this Item

All items in Spectrum are protected by copyright, with all rights reserved. The use of items is governed by Spectrum's terms of access.

Repository Staff Only: item control page

Research related to the current document (at the CORE website)
- Research related to the current document (at the CORE website)
Back to top Back to top