Artificial neural network prediction of weld geometry performed using GMAW with alternating shielding gases

Stuart Campbell, Alexander Galloway, Norman McPherson

Research output: Contribution to journalArticle

29 Citations (Scopus)

Abstract

An Artificial Neural Network (ANN) model has been applied to the prediction of key weld geometries produced using Gas Metal Arc Welding (GMAW) with alternating shielding gases. This is a recently developed method of supplying shielding gases to the weld area in which the gases are discretely supplied at a given frequency. The model can be used to predict the penetration, leg length and effective throat thickness for a given set of weld parameters and alternating shielding gas frequency. A comparison between the experimental and predicted geometries matched closely and demonstrates the effectiveness of this software approach in predicting weld outputs. The model has shown that the application of alternating shielding gases increases the penetration and effective throat thickness of a fillet weld whilst the leg length is reduced. A sensitivity analysis was performed which showed that the travel speed is the most influential input parameter when predicting weld geometries, this is to be expected for any given welding set-up due to the influence of the travel speed on the heat input. The sensitivity analysis also showed that the shielding gas configuration had the lowest influence on the output of the model. The output from the model has demonstrated that the use of alternating shielding gases during GMAW results in a step change in the weld metal geometry. This suggests that, in the case of alternating shielding gases, an increased travel speed is required to produce a similar weld geometry to that of the conventional Ar/20%CO2 technique.
LanguageEnglish
Pages174S-181S
JournalWelding Journal
Volume91
Issue number6
Publication statusPublished - Jun 2012

Fingerprint

Gas metal arc welding
Welding
Shielding
Artificial Neural Network
Welds
Arc of a curve
Metals
Gases
Neural networks
Geometry
Prediction
Sensitivity analysis
Penetration
Sensitivity Analysis
Output
Gas
Neural Network Model
Model
Lowest
Heat

Keywords

  • artificial neural network (ANN) model
  • key weld geometries
  • gas metal arc welding (GMAW)
  • shielding gases
  • ANN
  • GMAW
  • prediction
  • weld geometry
  • alternating shielding gases
  • plate distortion
  • arc signals
  • parameters
  • strength
  • joint

Cite this

@article{fac5e0092cf14fa4a4a20105e9dbeda2,
title = "Artificial neural network prediction of weld geometry performed using GMAW with alternating shielding gases",
abstract = "An Artificial Neural Network (ANN) model has been applied to the prediction of key weld geometries produced using Gas Metal Arc Welding (GMAW) with alternating shielding gases. This is a recently developed method of supplying shielding gases to the weld area in which the gases are discretely supplied at a given frequency. The model can be used to predict the penetration, leg length and effective throat thickness for a given set of weld parameters and alternating shielding gas frequency. A comparison between the experimental and predicted geometries matched closely and demonstrates the effectiveness of this software approach in predicting weld outputs. The model has shown that the application of alternating shielding gases increases the penetration and effective throat thickness of a fillet weld whilst the leg length is reduced. A sensitivity analysis was performed which showed that the travel speed is the most influential input parameter when predicting weld geometries, this is to be expected for any given welding set-up due to the influence of the travel speed on the heat input. The sensitivity analysis also showed that the shielding gas configuration had the lowest influence on the output of the model. The output from the model has demonstrated that the use of alternating shielding gases during GMAW results in a step change in the weld metal geometry. This suggests that, in the case of alternating shielding gases, an increased travel speed is required to produce a similar weld geometry to that of the conventional Ar/20{\%}CO2 technique.",
keywords = "artificial neural network (ANN) model, key weld geometries , gas metal arc welding (GMAW) , shielding gases , ANN, GMAW, prediction, weld geometry, alternating shielding gases, plate distortion, arc signals, parameters, strength, joint",
author = "Stuart Campbell and Alexander Galloway and Norman McPherson",
year = "2012",
month = "6",
language = "English",
volume = "91",
pages = "174S--181S",
journal = "Welding Journal",
issn = "0043-2296",
publisher = "American Welding Society",
number = "6",

}

Artificial neural network prediction of weld geometry performed using GMAW with alternating shielding gases. / Campbell, Stuart; Galloway, Alexander; McPherson, Norman.

In: Welding Journal, Vol. 91, No. 6, 06.2012, p. 174S-181S.

Research output: Contribution to journalArticle

TY - JOUR

T1 - Artificial neural network prediction of weld geometry performed using GMAW with alternating shielding gases

AU - Campbell, Stuart

AU - Galloway, Alexander

AU - McPherson, Norman

PY - 2012/6

Y1 - 2012/6

N2 - An Artificial Neural Network (ANN) model has been applied to the prediction of key weld geometries produced using Gas Metal Arc Welding (GMAW) with alternating shielding gases. This is a recently developed method of supplying shielding gases to the weld area in which the gases are discretely supplied at a given frequency. The model can be used to predict the penetration, leg length and effective throat thickness for a given set of weld parameters and alternating shielding gas frequency. A comparison between the experimental and predicted geometries matched closely and demonstrates the effectiveness of this software approach in predicting weld outputs. The model has shown that the application of alternating shielding gases increases the penetration and effective throat thickness of a fillet weld whilst the leg length is reduced. A sensitivity analysis was performed which showed that the travel speed is the most influential input parameter when predicting weld geometries, this is to be expected for any given welding set-up due to the influence of the travel speed on the heat input. The sensitivity analysis also showed that the shielding gas configuration had the lowest influence on the output of the model. The output from the model has demonstrated that the use of alternating shielding gases during GMAW results in a step change in the weld metal geometry. This suggests that, in the case of alternating shielding gases, an increased travel speed is required to produce a similar weld geometry to that of the conventional Ar/20%CO2 technique.

AB - An Artificial Neural Network (ANN) model has been applied to the prediction of key weld geometries produced using Gas Metal Arc Welding (GMAW) with alternating shielding gases. This is a recently developed method of supplying shielding gases to the weld area in which the gases are discretely supplied at a given frequency. The model can be used to predict the penetration, leg length and effective throat thickness for a given set of weld parameters and alternating shielding gas frequency. A comparison between the experimental and predicted geometries matched closely and demonstrates the effectiveness of this software approach in predicting weld outputs. The model has shown that the application of alternating shielding gases increases the penetration and effective throat thickness of a fillet weld whilst the leg length is reduced. A sensitivity analysis was performed which showed that the travel speed is the most influential input parameter when predicting weld geometries, this is to be expected for any given welding set-up due to the influence of the travel speed on the heat input. The sensitivity analysis also showed that the shielding gas configuration had the lowest influence on the output of the model. The output from the model has demonstrated that the use of alternating shielding gases during GMAW results in a step change in the weld metal geometry. This suggests that, in the case of alternating shielding gases, an increased travel speed is required to produce a similar weld geometry to that of the conventional Ar/20%CO2 technique.

KW - artificial neural network (ANN) model

KW - key weld geometries

KW - gas metal arc welding (GMAW)

KW - shielding gases

KW - ANN

KW - GMAW

KW - prediction

KW - weld geometry

KW - alternating shielding gases

KW - plate distortion

KW - arc signals

KW - parameters

KW - strength

KW - joint

M3 - Article

VL - 91

SP - 174S-181S

JO - Welding Journal

T2 - Welding Journal

JF - Welding Journal

SN - 0043-2296

IS - 6

ER -