Investigation and Optimization of Process Parameters in Ultrasonic Metal Welding for Achieving Quality Welds: A Taguchi–RSM–GA Approach with Thermal-FEA Validation

Abstract

Ultrasonic metal welding (USMW) is a solid-state joining process in which high-frequency, low-  amplitude relative motion applied under moderate normal pressure disperses surface oxides and produces intimate metal-to-metal contact without bulk melting. Because weld quality is governed by a narrow, interacting band of process parameters, this study presents an integrated framework for investigating and optimizing welding pressure, weld time and vibration amplitude for aluminium, copper and copper–brass lap joints. The work is organized in three phases: preliminary trials to bound the usable parameter range; a Design-of-Experiments phase combining the Taguchi orthogonal-array method (with ANOVA-based percentage contribution) and Response Surface Methodology (RSM) using a Face-Centred Central Composite Design, coupled with a Genetic Algorithm (GA) that treats the RSM regression model as its fitness function; and a thermal-field phase in which interface temperatures measured by thermocouple and infrared thermography are compared against Finite Difference Method (FDM) and Finite Element Analysis (FEA) predictions. For Cu–Cu welding, weld pressure, weld time and amplitude were found to contribute approximately 20%, 30% and 9% respectively to the variation in weld strength, with the pressure–time and pressure–amplitude interactions contributing a further 23% and 10%. Maximum weld strength was obtained near 3 bar pressure, 3 s weld time and 57 µm amplitude, while minimum strength scatter occurred near 3 bar, 2.5 s and 42.5 µm. RSM–GA optimization for Al and Cu/Cu–brass joints produced second-order regression models whose GA-optimized predictions agreed closely with experimental weld-bead density and resistance. FEA and FDM temperature-field predictions matched thermocouple and thermal-imaging measurements within acceptable limits, confirming that interface temperature rises with amplitude and weld time up to a threshold before joint degradation sets in. The results provide a validated, statistically grounded parameter-selection methodology that industry can use to replace trial-and-error tuning of USMW processes.