Multi-Objective Optimization of MAG Welding Parameters for Minimizing Distortion and Residual Stress in Industrial-Scale Heavy Machinery Assemblies

When heavy metal parts are joined together by welding, the intense heat creates a hidden problem that can compromise the strength and shape of massive machines. As the metal is heated, it expands, and as it cools, it shrinks. Because this heating and cooling happens unevenly across a large structure, different parts of the metal try to change size at different rates. This internal struggle leaves the finished piece with permanent warping, known as distortion, and invisible internal forces called residual stress. If these issues are not managed, the machine might not fit together correctly, or worse, it could fail under pressure. For the builders of giant forestry equipment, where parts are often over a meter wide and require immense heat to join, solving this puzzle is critical to ensuring safety and precision.

Researchers at Tampere University, working with the heavy machinery manufacturer Ponsse Oyj, set out to find a way to control these forces without slowing down production. They focused on a specific type of welding called Metal Active Gas welding, which uses a continuous stream of gas to protect the molten metal as it fuses two large steel components of a forestry machine frame. The team built a detailed digital twin of the assembly, a computer simulation that mimics the physics of the real world. This virtual model tracked how the steel heated up, how it expanded, and how it cooled down over several hours, accounting for the specific way the metal behaves under extreme temperature changes. The goal was to predict exactly where the metal would bend and where the internal stress would build up before a single spark was ever struck in a factory.

The challenge was that running these simulations is incredibly slow and computationally expensive. A single accurate simulation of the welding process took about five and a half hours to complete on a powerful supercomputer. To find the perfect settings, the researchers would theoretically need to run thousands of these simulations, testing every possible combination of voltage, current, speed, and starting position. Doing this would take years. To overcome this, the team developed a smart, adaptive strategy that acted like a skilled guide. Instead of blindly testing every option, their algorithm learned from a small number of initial tests to predict which settings were promising. It then focused its efforts on the most interesting areas of the search, gradually refining its understanding of the best settings while skipping the ones that would clearly fail. This approach allowed them to find the optimal solution using only sixty simulations, a task that would have been impossible with traditional methods.

The study identified four key variables that the welders could control: the electrical voltage, the current strength, the speed at which the welding torch moves, and the specific angle where the welding begins. The researchers discovered that the starting point of the weld was just as important as the speed or power. In the standard method, welders often start at the same spot for multiple passes, which causes heat to pile up in one area, leading to significant bending. By changing the starting angle for the second pass of welding to a different location on the circle, the heat was distributed more evenly across the structure. This simple adjustment prevented the metal from warping in one direction and significantly reduced the final distortion.

The results of this optimization were substantial. The team found three specific sets of parameters that offered different benefits. One combination reduced the maximum bending of the metal by nearly thirty percent compared to the standard method, while another reduced the internal stress by about four percent. A third option offered a balanced improvement, cutting the bending by roughly twenty-six percent and the stress by four percent. These improvements were achieved while keeping the welding process within the strict safety and quality limits required by the manufacturer. The research suggests that by carefully planning where to start the weld and adjusting the speed and power, factories can produce stronger, straighter machines without needing to change their equipment or add extra cooling steps.

This work highlights that even in established industrial processes, there is still room for significant improvement through careful analysis. The researchers demonstrated that the way heat is managed during the welding process is the primary driver of distortion, more so than the total amount of energy used. By using advanced computer modeling to understand these thermal cycles, they provided a practical roadmap for engineers to reduce waste and improve quality. While the findings were confirmed through simulation rather than physical testing on a full-scale machine, the logic of the model is grounded in the known behavior of steel. The study offers a clear path forward for the heavy machinery industry, showing that a smarter approach to the starting point of a weld can lead to a much straighter, more reliable final product.