A maker of architectural products approached Boulder Imaging, based in Louisville, Colo., with a problem: intermittent fires on a plant floor caused by the production process. As part of manufacturing, materials passed through an oven at temperatures ranging between 150 and 300 ºF. This treatment removed moisture and was essential to ensuring the material could then be further processed into a high-quality final product. There was, however, an issue. “Material exiting the oven occasionally caught on fire for unknown reasons, posing serious safety risks and extended process downtime,” said Joey Nesbitt, Boulder Imaging’s director of hardware engineering. “For those reasons, these fire events represented serious cost and safety implications.” To keep throughput at an acceptable level, the oven temperature couldn’t be lowered. For quality product, the oven had to be used. Carlos Jorquera, Boulder Imaging CEO and chief technology officer (left), and Andrew Bour, the company’s field application engineer, on the factory floor testing the system. Courtesy of Boulder Imaging. As a fix, the company had technicians watch for burning material. But this solution, which relied on the visible spectrum and human eyes, could spot only smoke and flames. Thus, this approach couldn’t identify hot spots in the material that posed a risk of starting a fire later. The architectural product maker turned to Boulder Imaging to devise a solution. Boulder Imaging specializes in developing innovative machine vision solutions that involve artificial intelligence and proprietary software and hardware to solve high-value inspection problems. A typical approach using cameras and illumination operating in the visible might improve upon spotting smoke and fire. Boulder Imaging often uses high-intensity lighting from Chromasens, Nesbitt said. But for the product-drying project, it opted for a different approach, one that didn’t use lighting — at least in the visible sense — at all. Boulder Imaging’s engineers decided to go long — as in they chose to use a longwave infrared-based solution. The longwave IR part of the spectrum runs from 8 to 15 µm and is also known as thermal infrared. Objects produce emissions in this wavelength range that sensors can pick up, and they do so without the need for external illumination. Because longwave IR is far removed from the 0.4 to 0.7 µm range of the visible spectrum, it requires different sensors. For this application, Boulder Imaging evaluated a number of cameras and selected one from Xenics. In the technology assessment process, Nesbitt said it quickly became apparent that there were far fewer candidate cameras in the longwave IR than there are in the visible. He said the company’s engineers had to implement software solutions to make up for shortcomings in the longwave IR hardware. Boulder Imaging’s technical staff also had to deal with environmental conditions. “This production process generates significant dust and debris that coats everything in sight, including the camera,” Nesbitt said. The solution involved sealing the camera and capturing images through a viewport. Because the viewport had to be transparent to thermal IR, Boulder Imaging made it of germanium glass. To keep dust from collecting on it, engineers designed the system with a blow-off nozzle. Periodic use of the nozzle would then keep the viewport clear and the camera unobstructed. Nesbitt said Boulder Imaging’s Vision Inspector software, which was part of the solution, had not been used with longwave IR before. The software, nonetheless, proved as capable of adapting to IR camera sensors as it had in other applications that used visible monochrome and color cameras. Architectural product dried by traveling through an oven reaches an elevated temperature, as shown by these thermal IR images. A vision system developed by Boulder Imaging would set off an alert in response to the image on the right, based on the white dot that indicates a hot spot, seen about a quarter of the way from the top. Courtesy of Boulder Imaging. Today, the thermal IR camera-based solution scans all material exiting the oven. The system detects hot spots and sends alarms to alert operators of fire risks. It also provides data about average temperature, which can be used for process improvement. The customer can quarantine material that is at risk, thereby reducing or preventing fires. This improves workplace safety, avoids extended downtime, and results in substantial cost savings, Nesbitt said. What’s more, with the temperature data, the customer’s technicians and engineers can make changes to enhance process quality and increase overall yield. And so, while lighting is key to successful machine vision, it needs to be the right illumination. The illumination may be provided by the material being inspected, which can bring benefits such as enabling the detection of a hot spot before it catches fire. Nesbitt said of the Boulder Imaging solution, “Thermal infrared allows us to take a more preventive approach and detect risk areas before they become a problem.”