Researchers Optimize Friction Welding for AA2024-T351 Aluminum Alloy

A scientific paper published on May 10, 2026, details efforts to improve welding outcomes for AA2024-T351 aluminum alloy, a material valued for its strength-to-weight ratio in aerospace and automotive applications. Conventional welding methods have struggled to produce reliable joints in this alloy.

The study examined direct drive friction welding and tested a novel drilled-to-bossed geometry modeled on traditional mortise-and-tenon construction to increase mechanical interlocking. Experiments followed an L18 orthogonal array design to assess effects on tensile and torsional strength.

Parameters included friction pressure, forging pressure, spindle speed, faying surface geometry, and friction time. All trials were performed on a customized engine lathe. The optimal settings identified were 30 MPa friction pressure, 70 MPa forging pressure, 2200 rpm spindle speed, the drilled-to-bossed geometry, and 4 minutes of friction time.

This combination produced a tensile strength of 538 MPa and a torsional strength of 325 MPa. The drilled-to-bossed configuration delivered higher joint strength than conventional flat-to-flat joints.

Three deep learning architectures were trained to predict ultimate tensile strength from the welding parameters: an ensemble residual network, a compact attention network, and an adaptive multiscale network. The ensemble residual network recorded the highest accuracy, reaching an R² value of approximately 0.81.

The models demonstrated an ability to capture nonlinear relationships between process variables and mechanical performance. Such predictive tools could support process optimization without exhaustive physical testing. The research was carried out by teams from institutions including the National Taiwan University of Science and Technology, Debre Berhan University, Politecnico di Milano, Bahir Dar University, and Manipal University Jaipur.