System-level productivity and energy performance of manual and cobotic welding in high-mix, low-volume manufacturing

Welding in high-mix, low-volume (HMLV) manufacturing is not the same as welding in high-volume production. Parts change frequently, fixturing is rarely stable, and a large portion of each welding cycle is spent on activities unrelated to the arc itself in HMLV manufacturing. Despite this, most studies comparing robotic and manual welding are conducted under controlled laboratory conditions, conditions that do not resemble real job-shop environments. This study takes a different approach. Manual and cobotic GMAW were compared using full-cycle measurements across ten production parts selected to reflect the geometry and handling variation typical of HMLV work. Each part was welded ten times per method at three industrial sites, yielding 200 total cycles. Cycle time was broken down into arc-on, non-arc productive, and idle components. Energy consumption was measured for the full welding cycle, not just during active welding. Weld quality was evaluated through visual inspection and tensile testing of A36 steel butt and lap specimens. On average, cobotic welding reduced cycle time by 21.0% (from 277.18 s to 219.07 s) and energy consumption by 12.8% (from 0.1708 kWh to 0.1489 kWh), both differences were statistically significant (p < 0.05). The gains were not consistent across parts, however. Cycle time reductions ranged from 5.2% to 53.1%, with the largest improvements on parts that had long continuous welds and limited repositioning, and the smallest on parts with high fixturing complexity. Weld quality was equivalent or better under cobotic welding across all parts tested. The practical implication of this study is straightforward. Cobotic welding can improve both throughput and energy efficiency in HMLV settings, but the benefit depends heavily on part structure. Arc-on ratio and repositioning effort are the key factors. Evaluating these before deployment, rather than assuming uniform gains, is essential.