A Study on Defect Monitoring in Stainless Steel Laser Welding Processes Using a Multi-Sensor System

In the world of modern electric vehicles, the battery is the heart, but the cooling system is the circulatory system that keeps it from overheating. These systems rely on thin metal panels that must be joined together perfectly to prevent leaks and ensure safety. For years, engineers have used aluminum for these parts because it is light, but aluminum can struggle when temperatures get extreme, such as during a thermal runaway event where a battery overheats dangerously. To solve this, manufacturers are turning to a specific type of stainless steel, known as SUS304L, which can withstand higher heat and offers greater structural strength. However, joining this steel requires a precise, high-energy process called laser welding. Because the laser moves so fast and carries so much energy, it can easily create hidden flaws like tiny air pockets or gaps along the seam if the conditions are not perfect. These flaws are dangerous because they can lead to leaks or structural failure, yet they are often invisible to the naked eye once the welding is done.

To catch these flaws before they cause trouble, a team of researchers from the Korea Institute of Industrial Technology and industry partners set out to build a smarter way to watch the welding process in real time. They focused on a newer type of laser called an Adjustable Ring Mode laser, which can shape its energy beam to create a more stable pool of molten metal. Instead of relying on a single tool to watch the weld, they created a multi-sensor system that acts like a team of observers, each looking at a different physical clue. One sensor watched for flashes of light in the ultraviolet and infrared spectrums, which reveal how the superheated plasma and molten metal are behaving. Another sensor, a high-speed thermal camera, mapped the heat distribution around the weld. By combining these different views, the team aimed to distinguish between a perfect weld and one that was developing a defect, all while the laser was still moving.

The researchers tested their system by welding stainless steel sheets under various conditions, deliberately creating three common types of defects: porosity, which are tiny holes trapped inside the metal; undercut, where the edge of the weld melts away too much; and non-penetration, where the weld fails to go deep enough to join the bottom layer. They compared the sensor data against X-ray images and physical cross-sections of the metal to see exactly what the sensors were seeing. Under normal, perfect welding conditions, the sensors showed a calm and steady rhythm. The infrared light signal stayed within a predictable range, and the ultraviolet signal remained flat and stable. This steady state served as the baseline, the "all clear" signal that told the researchers the molten pool was behaving as expected.

When the researchers introduced defects, the sensors began to scream in different ways, depending on the problem. When porosity occurred, the ultraviolet sensor was the first to react. The signal, which usually sat steady, began to drop and spike erratically. This happened because the formation of air pockets destabilizes the plasma, the ionized gas that forms above the molten metal, causing the ultraviolet light to flicker. In contrast, when an undercut defect appeared, the infrared sensor took the lead. The heat signal showed irregular, sharp peaks that indicated the molten metal was flowing unevenly at the edges of the weld. The researchers found that while the ultraviolet light was excellent for spotting air pockets, the infrared heat signal was the best indicator for edge defects.

The team also discovered that how they looked at the heat mattered just as much as what they were looking for. They tested two ways of using the thermal camera. One method, called total area analysis, watched the entire length of the weld bead to see the overall heat history. While useful for a broad view, this method was easily confused by splashes of hot metal, known as spatter, which created false peaks in the data. The second method, local area analysis, kept the camera focused tightly on the laser spot and the immediate molten pool. This approach provided a much more stable and reliable reading of the heat right where the welding was happening. When the researchers tested a condition where the laser power was too low to fully penetrate the metal, the local area analysis clearly showed a change in the thermal response, while the total area view was less distinct. This confirmed that keeping a close, steady eye on the interaction zone was the most effective way to catch deep flaws like insufficient penetration.

By combining these insights, the researchers established a new set of rules for monitoring weld quality. They found that no single sensor could tell the whole story; instead, a perfect weld required a stable infrared signal, a steady ultraviolet signal, and a consistent local heat pattern. If the ultraviolet signal went wild, it likely meant air pockets were forming. If the infrared signal spiked irregularly, the edges were melting too much. If the infrared signal fluctuated widely while the local heat pattern changed, the weld was likely not deep enough. This multi-sensor approach proved that by listening to the different "voices" of the welding process—light and heat, seen from different angles—engineers can detect defects the moment they happen. The study suggests that this method could be a vital step toward safer, more reliable electric vehicle batteries, ensuring that the stainless steel components holding them together are free of hidden flaws.