Inspecting hot steel as it rolls off a manufacturing line can present a difficult problem for machine vision systems. For one thing, the steel glows, making surface imperfections hard to see. For another, steel is manufactured in environments with dirt, high temperatures and humidity. Now, the private nonprofit Robotiker-Tecnalia foundation has come up with a solution that enables defects to be spotted while the metal is hot. All that was needed was to look at things in the right way and with the right light, said project manager José Angel Gutierrez. “The idea is to irradiate the object with [a] wavelength far enough from the emitted spectrum of hot steel. In this case, blue, but it could be green in some cases, or violet.” Hot steel glows visibly as it emerges from a push bench, making automated machine inspection of the surface difficult. Images courtesy of José Angel Gutierrez, Robotiker-Tecnalia. A few years ago, steelmaker Tubos Reunidos SA of Amurrio approached Robotiker-Tecnalia with a problem. The company makes hot rolled seamless pipes ranging in diameter from 140 to 220 mm and in length from 7 to 20.5 m. The very last step in the manufacturing process involves press rolling at a push bench, with tube temperatures of about 1200 °C. Defects introduced as a result of the rolling stands repeat down a tube and from tube to tube, until the stands are changed. Consequently, catching and correcting defects as early as possible is important. However, the heat of the steel made it glow an orange-red, bright enough to be visible in sunlight. That glow made the steel itself a light source for any machine vision camera trying to inspect its surface. For a solution, the team at Robotiker-Tecnalia used a laser, a high-speed linear camera, and thermal and passband filters. It packaged these up in three protective enclosures equally spaced around the steel, cooling everything with an air-based refrigeration system, and pressurized the enclosures to keep out dust. The optical filters, Gutierrez explained, eliminated everything from the camera’s view but the reflected light from the laser. He noted that the monochromatic source helped with this, but that it was just as important that the laser was a bright source that could be focused as needed. Enclosures protect cameras and light sources (top) that illuminate the processed steel using blue light and scan it for defects (below). The three cameras captured 1024 pixels per line and 14,000 lines per second. All of this data was fed over a fiber link to a processing unit, which looked for defects on the tubes as they traveled past the inspection point at 5.5 m/s. The system has been operational for some time. Some defects are picked up well, but capturing others must be improved. Better algorithms for defect classification or other enhancements could help with that. There also have been problems with laser nonuniformities and with aligning the laser and camera. Nonetheless, the system has attracted the attention of other metal-processing companies. It isn’t just steel or metal that could make use of the technology, Gutierrez said. “It can be used in any application where hot elements must be inspected.”