• JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
 
  Bookmark and Share
 
 
Doctoral Thesis
DOI
Document
Author
Full name
Ricardo Luís Balieiro
Institute/School/College
Knowledge Area
Date of Defense
Published
São Carlos, 2019
Supervisor
Committee
Flauzino, Rogério Andrade (President)
Baptista, Fabricio Guimarães
Brandão, Dennis
Leão, Fábio Bertequini
Santo, Silvio Giuseppe Di
Title in Portuguese
Desenvolvimento de abordagem baseada em técnicas de visão computacional e de aprendizado de máquinas para monitoramento e controle de falhas em correias transportadoras
Keywords in Portuguese
Aprendizado de Máquinas
Correias Transportadoras
Reconhecimento de Padrões
Visão Computacional
Abstract in Portuguese
Correias transportadoras têm sido utilizadas como um modo eficiente de transporte de materiais por diversos ramos da indústria. A vida útil desse equipamento é influenciada pelo desgaste natural ocasionado pelo uso excessivo e falhas, tais como rasgos, cortes e perfurações. Nesse trabalho, é proposto um método de identificação de falhas em correia transportadora por meio de processamento digital de imagens, combinadas com técnicas de escaneamento 3D a laser e aprendizado de máquinas (redes neurais e SVM (Support Vector Machine - Máquina de Vetores de Suporte)). A averiguação experimental foi efetuada em duas etapas: a primeira etapa inicia-se com a aquisição, o pré-processamento e a geração do banco de imagens; a segunda, com a extração e a classificação das amostras. O vetor de características foi submetido a dois classificadores: rede neural do tipo Perceptron Múltiplas Camadas (com 15 topologias candidatas) e um classificador SVM, variando sua função kernel: linear, quadrática, polinomial, gaussiano, RBF (Radial-Basis Function - Função de Base Radial) e MLP (Multilayer Perceptron – Perceptron Múltiplas Camadas). Os resultados experimentais mostram que a metodologia proposta obtém uma boa acurácia na estimação das falhas, mostrando-se promissora na tarefa de diagnosticar e classificar falhas em correias transportadoras.
Title in English
Development of approach based on computer vision and machine learning techniques for monitoring and control of conveyor belt failures
Keywords in English
Computer Vision
Conveyor Belts
Machine Learning
Pattern Recognition
Abstract in English
Conveyor belts have been used as an efficient way of transporting materials by various industry branches. This equipment's lifespan is influenced by natural wear caused by overuse and flaws, such as tears, cuts, and perforations. This work proposes a method of identifying flaws in the conveyor belt using digital image processing, combined with 3D laser scanning techniques and machine learning (neural networks and SVM - Support Vector Machine). The experimental investigation has been carried out in two stages: the first stage begins with the acquisition, preprocessing, and generation of an image bank; the second, with the extraction and classification of the samples. The characteristic's vector has been submitted to two classifiers: Multi-Layer Perceptron neural network (with 15 candidate topologies) and an SVM classifier, varying its kernel function: linear, quadratic, polynomial, Gaussian, RBF (Radial-Basis Function) and MLP (Multilayer Perceptron). The experimental results show that the proposed methodology obtains a good accuracy in the failure estimation, being promising in the task of diagnosing and classifying flaws in conveyor belts.
 
WARNING - Viewing this document is conditioned on your acceptance of the following terms of use:
This document is only for private use for research and teaching activities. Reproduction for commercial use is forbidden. This rights cover the whole data about this document as well as its contents. Any uses or copies of this document in whole or in part must include the author's name.
Ricardo.pdf (9.52 Mbytes)
Publishing Date
2019-10-08
 
WARNING: Learn what derived works are clicking here.
All rights of the thesis/dissertation are from the authors
CeTI-SC/STI
Digital Library of Theses and Dissertations of USP. Copyright © 2001-2021. All rights reserved.