Prediction and Optimization of Load-Bearing Capacity in Resistance Spot Welded Titanium Joints Using Neural Networks...

Area of Science:

• Materials Science
• Manufacturing Engineering
• Computational Modeling

Background:

• Titanium alloys are crucial in aerospace and medical industries.
• Resistance spot welding is a common joining method for titanium.
• Optimizing welding parameters for dissimilar titanium joints (Grade 1 and Grade 5) is challenging.

Purpose of the Study:

• To investigate the mechanical performance of resistance spot-welded titanium lap joints.
• To develop predictive models for joint load-bearing capacity using artificial neural networks (ANN).
• To optimize welding parameters for dissimilar titanium joints through evolutionary algorithms.

Main Methods:

• Experimental testing of Grade 1 and Grade 5 titanium lap joints.
• Development of three ANN models for different joint configurations (Grade 1/1, 1/5, 5/5).
• Optimization of the most accurate ANN model using a genetic algorithm.

Main Results:

• The mixed Grade 1/Grade 5 joint model showed the highest predictive accuracy (R² = 0.9289).
• Models demonstrated high reliability with 4-6% mean relative errors.
• Genetic algorithm optimization identified optimal parameters (2.89 kA, five pulses) yielding a 3.2 kN predicted capacity.

Conclusions:

• Combining ANN modeling with genetic algorithm optimization is effective for designing efficient welding processes.
• Optimal parameters were found near the boundary of the high-strength region, balancing quality and efficiency.
• This approach successfully meets the required load-bearing capacity threshold for dissimilar titanium joints.