Fraunhofer Institute for Industrial Mathematics ITWM, Munich, Germany
www.fraunhofer.de/en
Quality control of component surfaces can be very complex. Researchers at the Fraunhofer Institute for Industrial Mathematics (ITWM) in Munich, Germany, say that they have engineered a high-precision modular inspection system that can be custom adapted to specific customer needs and into the production process.
In order to check a component, manufacturers can use image processing methods. Multiple cameras take pictures of component surfaces from various angles, which are then analyzed by a software program. “Every material substance has its own unique surface structure. In order to evaluate its quality, the testing procedure has to be gauged precisely to these specific properties,” explains Markus Rauhut of ITWM. The size and shape of a component also play a role, he explained, as well as the desired resolution of the images.
While there are many surface inspection systems on the market already, they are configured for specific materials and dimensions. Standard solutions cannot cover the entire spectrum of possible testing objectives, Rauhut explained. So, the researchers at ITWM have engineered a modular inspection system dubbed “MASC: Modular Algorithms for Surface InspeCtion,” which can be modified to customer-defined specifications.
“Our system is suited for the most diverse materials—like metals, leather, textiles or paper—and covers a size range from tiny components for medical technology through to entire sheets of rawhide or ceiling panels,” Rauhut stated. MASC-STeX for the inspection of cover plates and MASC-Dehnzelle for the inspection of expansion cells are already in practical use in the industry. (See Figure 1)
System Uses More than 300 Algorithms
First, the surface of the workpiece is illuminated and scanned using multiple cameras set at a wide range of angles. “This is important, so that you can also pick up impact points or fractures that are only visible from one side,” explained Rauhut. With free-formed surfaces, regions are also recorded in this manner that are covered by curvatures or corners. The more complex the geometry, the more cameras are needed, as a rule.
For the analysis of the images, the scientists developed mathematical evaluation algorithms and then built a comprehensive software library. “For instance, one algorithm is programmed to find edges or certain color points in the image,” explained Kai Taeubner of ITWM. The basic version of the software alone comprises more than 300 algorithms that could also be combined, depending on the testing task.
One particular challenge, they said, includes inspection procedures requiring very high resolution. No surface is quite homogenous; they feature small scratches or fluctuations in brightness. For projects where resolution reaches the microscopic level, it can become increasingly difficult to differentiate between anomalies in the surface texture and actual defects, which could lead to defect-free components being sorted as defective with current systems.
Once all testing parameters are set, the procedure is integrated into the production process with the customer. The cameras are either installed directly on the assembly line for this purpose, or applied using robots. When a defect is found, the production process automatically stops, and the machine operator is notified. The detected defects are classified and the test object is divided into quality classes.
