C. Jentner GmbH is a specialist providing the full range of electroplating (= electrochemical cutting of metallic coatings) and surface finishing services. The mid-sized company was founded in the 1970s and is today offering coating services for complex workpieces for customers in the medical technology, aerospace, and electrical engineering sectors. The traditional Pforzheim-based company provides a wide range of galvanic coatings, such as nickel, copper, silver, and rhodium plating. Their customers have the highest quality requirements for the refined products, be it functional elements or optical surfaces.
Traditionally, most optical inspections within the plating industry have been done manually. This activity is demanding, time-consuming and resource-intensive, even for trained professionals, as difficult-to-identify minute material defects and repetitive tasks place significant demands on human performance. Moreover, automated analysis of the collected quality metrics is often tricky, and feedbacking the gained knowledge for optimizing the galvanizing process is uneconomical. For these reasons, C. Jentner GmbH turned its attention to implementing automated, AI-supported inspection systems, which was successfully achieved with the AI-supported quality assurance system AI.SEE™ from elunic.
A semi-automated optical inspection system using the AI.SEE™ AI-based quality control system was implemented on a factory inspection table to realize the concept. Part of the visual quality inspection and defect detection system is a self-learning AI system (AI.SEE™ Core) that directly evaluates the received images, assigns defect classes, and controls other down-stream processes. Defects and damage, such as surface elevations or scratches, can thus be detected automatically. The AI system performance was expected to at least match or ideally exceed manual quality inspection. As a next step, the process will be extended to more test tables and fully automated by commissioning cobots. Using artificial neural networks and Deep Learning, the model is further trained to detect all defect classes on all materials and products and to display and analyze them directly for newly uploaded images.
By automating quality control steps, error-prone processes can be minimized. Frequently occurring sources of defects as well as slight flaws, which human inspectors often overlook, can be found and classified quickly and unambiguously. These compelling findings boost quality and operational efficiency.