Automating nut tightening using Machine Learning
Paper in proceedings, 2020

At the Volvo Truck assembly plant the repetitive task of nut tightening is not ideal regarding
quality and ergonomic. The solution to both these issues would be to significantly increase the level of
automation. However, automating this specific station requires solutions to two specific problems. The
first problem is to find and identify what nuts that need to be tightened, since they are not always on the
same position for this highly customized product. The second problem is that the automated solution
needs to accommodate the working space which is a moving assembly line with human operators. This
paper investigates how these two problems ban be solved using machine learning and collaborative
robots. A realistic mockup of the assembly station has been created at Stena Industry Innovation
Laboratory (SII-Lab) where all the testing has been done.
The problem to identify the nuts to tighten is further complicated by the fact that some nuts are placed
backwards for future further assembly which must be avoided. Therefore, the selected solution is to use
supervised machine learning for object recognition. This way, the system can be trained to recognize both
nuts that need to be tightened and those mounted backwards, and possible other objects needed. Tests
have been conducted with different types of CNN (Convolutional Neural Network) algorithms. Results
have been very successful, and the test setup has successfully managed to connect the whole task of
identifying the correct nuts and move the collaborative robot to that specific position.

assembly

Machine learning

CNN

collaborative robot applications

Author

Kevin Wedin

Chalmers

Christoffer Johnsson

Chalmers

Magnus Åkerman

Chalmers, Industrial and Materials Science, Production Systems

Åsa Fasth Berglund

Chalmers, Industrial and Materials Science, Production Systems

Per-Anders Alveflo

IFAC Proceedings Volumes (IFAC-PapersOnline)

14746670 (ISSN)

virtual IFAC World Congress
Berlin (virtual), Germany,

Demonstrating and testing smart digitalisation for sustainable human-centred automation in production

VINNOVA, 2017-05-15 -- 2020-03-09.

Subject Categories

Mechanical Engineering

Computer and Information Science

Areas of Advance

Production

More information

Created

8/4/2020 7