This content is not included in
your SAE MOBILUS subscription, or you are not logged in.
Machine Learning Models for Weld Quality Monitoring in Shielded Metal Arc Welding Process Using Arc Signature Features
- K. Rameshkumar - Amrita Vishwa Vidyapeetham, Department of Mechanical Engineering, Amrita School of Engineering, India ,
- A. Vignesh - Amrita Vishwa Vidyapeetham, Department of Mechanical Engineering, Amrita School of Engineering, India ,
- P. Gokula Chandran - Amrita Vishwa Vidyapeetham, Department of Mechanical Engineering, Amrita School of Engineering, India ,
- V. Kirubakaran - Amrita Vishwa Vidyapeetham, Department of Mechanical Engineering, Amrita School of Engineering, India ,
- J. Sankaran - Amrita Vishwa Vidyapeetham, Department of Mechanical Engineering, Amrita School of Engineering, India ,
- A. Sumesh - Amrita Vishwa Vidyapeetham, Department of Mechanical Engineering, Amrita School of Engineering, India
Journal Article
05-15-04-0023
ISSN: 1946-3979, e-ISSN: 1946-3987
Sector:
Topic:
Citation:
Rameshkumar, K., Vignesh, A., Gokula Chandran, P., Kirubakaran, V. et al., "Machine Learning Models for Weld Quality Monitoring in Shielded Metal Arc Welding Process Using Arc Signature Features," SAE Int. J. Mater. Manf. 15(4):347-365, 2022, https://doi.org/10.4271/05-15-04-0023.
Language:
English
Abstract:
Welding is a dominant joining process employed in fabrication industries,
especially in critical areas such as boiler, pressure vessels, and marine
structure manufacturing. Online monitoring of welding processes using sensors
and intelligent models is increasingly used in industries for predicting weld
conditions. Studies are conducted in a Shielded Metal Arc Welding (SMAW) process
using sound, current, and voltage sensors to predict the weld conditions. Sensor
signatures are acquired from the good weld and defective weld conditions
established in this study. Signal processing is carried out, and time-domain
statistical features are extracted. Statistical features are also extracted from
the power waveform derived from the current and voltage data for all the weld
conditions. Classification And Regression Tree (CART) and Support Vector Machine
(SVM) algorithms are used to build the statistical models to predict the weld
conditions. SVM algorithm with Quadratic Kernel function trained using power
signature features predicts weld conditions considered in this study with an
accuracy of 99%.