Computer Vision Inspection of Cold-Flow Casting Defect with Neural Network

Syaefudin, Adhy and Sofyan, Edi and Setiawan, Widi (2021) Computer Vision Inspection of Cold-Flow Casting Defect with Neural Network. Masters thesis, Swiss German University.

[img]
Preview
Text
Adhy Syaefudin 21952051 TOC.pdf

Download (2MB) | Preview
[img] Text
Adhy Syaefudin 21952051 1.pdf
Restricted to Registered users only

Download (689kB)
[img] Text
Adhy Syaefudin 21952051 2.pdf
Restricted to Registered users only

Download (3MB)
[img] Text
Adhy Syaefudin 21952051 3.pdf
Restricted to Registered users only

Download (2MB)
[img] Text
Adhy Syaefudin 21952051 4.pdf
Restricted to Registered users only

Download (3MB)
[img] Text
Adhy Syaefudin 21952051 5.pdf
Restricted to Registered users only

Download (580kB)
[img]
Preview
Text
Adhy Syaefudin 21952051 Ref.pdf

Download (763kB) | Preview

Abstract

The quality of the parts in the Aluminum High-Pressure Die Casting (HPDC) injection process is an important concern. If the defects that are not and sent to the next process will cause production cost loss, and can even result in losses in the hands of customers due to engine performance resulted from the part defects. The research thesis presents the implementation of object detection technology based on neural-network to detect cold-flow defect. Computer vision with neural network algorithm will be used to improve the result of visuals human inspections who have been detecting cold-flow defects that will t in aresul more stable and objective in carrying out quality assessments. This thesis will analyze the performance of the YOLOv5s framework in Pyhton. The analysis includes lighting conditions and the characteristics of the dataset. This thesis used cloud computing at Google-Colab during the training processof the Deep Convolutional Neural Network (DCNN), the computer specifications with graphic processing unit known as GPU (Tesla T4, 1.5 GB, 40 processor assigned by Google-Colab) are needed for the training process. The ROBOFLOW was very helpful tools in the dataset preparation phase of the development of this system. An intense discussion with the expert team was also performed, to make adjustments to the neural network object detection result. In conclusion, the developed system has proven to be very successful in assisting the HPDC part visual inspection.

Item Type: Thesis (Masters)
Uncontrolled Keywords: DCNN, Neural-Network, YOLO, ROBOFLOW, Python, Computer-Vision, Object Detection
Subjects: Q Science > QA Mathematics > QA76 Computer software > QA76.87 Neural networks (Computer science)
T Technology > TJ Mechanical engineering and machinery
T Technology > TJ Mechanical engineering and machinery > TJ163.12 Mechatronics
T Technology > TJ Mechanical engineering and machinery > TJ210 Robots (General)
Divisions: Faculty of Engineering and Information Technology > Department of Mechatronics Engineering
Depositing User: Faisal Ifzaldi
Date Deposited: 19 Aug 2021 10:09
Last Modified: 19 Aug 2021 10:09
URI: http://repository.sgu.ac.id/id/eprint/2147

Actions (login required)

View Item View Item