Mild Steel GMA Welds Microstructural Analysis and Estimation Using Sensor Fusion and Neural Network Modeling

Area of Science:

• Materials Science
• Manufacturing Engineering
• Computational Intelligence

Background:

• Gas Metal Arc Welding (GMAW) is a critical industrial process.
• Understanding weld microstructure is essential for structural integrity.
• Predicting the heat-affected zone (HAZ) remains a challenge.

Purpose of the Study:

• To evaluate sensor fusion efficiency using neural networks for microstructure estimation in GMAW.
• To develop models for predicting weld bead and base material microstructure.
• To assess the capability of artificial neural networks (ANNs) in estimating the HAZ extension.

Main Methods:

• Deposited AWS ER70S-6 wire on SAE 1020 steel with varied welding parameters.
• Analyzed thermal behavior using infrared thermography.
• Characterized microstructure via optical microscopy, SEM, and XRD.
• Developed microstructure estimation models using neural network sensor fusion.

Main Results:

• Achieved a high R-value of 0.99472 for microstructure modeling using ANNs with Bayesian Regularization.
• The developed ANN models accurately estimated microstructure across all zones.
• Demonstrated the capability to predict the heat-affected zone (HAZ) extension.

Conclusions:

• Sensor fusion with neural networks is highly efficient for estimating weld microstructure.
• ANNs provide a reliable method for predicting the heat-affected zone (HAZ).
• This technology can assist in the design and quality control of welded joints.