Automatic assembly fault detection system using machine vision method and transfer learning

Abd. Halim, Ahmad Bukhari (2022) Automatic assembly fault detection system using machine vision method and transfer learning. Project Report. Universiti Teknikal Malaysia Melaka, Melaka, Malaysia. (Submitted)

	Text (24 Pages) Automatic assembly fault detection system using machine vision method and transfer learning.pdf - Submitted Version Download (754kB)
	Text (Full text) Automatic assembly fault detection system using machine vision method and transfer learning.pdf - Submitted Version Restricted to Repository staff only Download (1MB)

Abstract

Defects and abnormalities in parts that affect the part's quality are a particular issue in additive manufacturing (AM) techniques such as at assembly line. At the moment, destructive and non-destructive testing methods are mostly employed to ensure the quality of additive manufacturing components after production. Machine learning (ML) techniques are increasingly being utilised in this area to allow computer-aided fault detection by automatically classifying production data. Convolutional neural networks (CNNs) based on machine learning techniques are often employed to accomplish this job. In this work, a transfer learning (TL) techniques for automatically classifying the defect in assembly process using relatively little datasets is proposed. The suggested techniques identify excellent faulty picture data and are able to classifiy the parts collected during component production using the CNN models with pretrained weights from the MobileNet dataset as initialization and a modified classifier. These findings demonstrate the efficacy of CNNbased part classification and provide a non-destructive quality assurance and production documentation approach for additively produced components.

Actions (login required)

View Item