Please use this identifier to cite or link to this item:
Title: A machine learning approach to detect surface features for automatic robot taping
Authors: Chong, Bing Sheng
Keywords: Engineering::Mechanical engineering::Robots
Issue Date: 2019
Abstract: Taping is a common process in part manufacturing, usually performed before surface treatment operations are done. Previous studies have come up with an automatic robot taping method for parts with general and simple surfaces, able to generate a 3D model on the fly and plan out the optimum path for taping operations with force feedback to ensure adequate adhesion of the tape to the surface. However, more work needed to be done for the automatic robot taping system to be able to work with parts with complex geometries, such as gaps and holes. This paper aims to create a machine learning model that can analyze force feedback data from the load cell and potentiometer to identify the surface features in contact with the taping tool head, as well as identify the incidence angle of taping tool head to the surface. The sensors were first assembled onto the taping tool head and calibrated properly. Then, a regression model was designed and trained to identify the incidence angle. The regression model had an error range of 15%. On the other hand, a classification model was designed and trained to identify the contact surface between three categories of surfaces: top-hole, no-holes and bottom-hole surface. An accuracy of 95.3% was achieved for the classification model. Using these two models we were able to identify the orientation of the tool head and the surface features on the model, providing valuable insight for the tool head to make adjustment such that optimum adhesion of tape can be achieved.
Rights: Nanyang Technological University
Fulltext Permission: restricted
Fulltext Availability: With Fulltext
Appears in Collections:MAE Student Reports (FYP/IA/PA/PI)

Files in This Item:
File Description SizeFormat 
  Restricted Access
1.05 MBAdobe PDFView/Open

Page view(s)

Updated on Oct 23, 2021

Download(s) 50

Updated on Oct 23, 2021

Google ScholarTM


Items in DR-NTU are protected by copyright, with all rights reserved, unless otherwise indicated.