Abstract
In industrial-scale applications of ultrasonic metal welding (UMW), tool condition monitoring (TCM) is one of the most important maintenance tasks because tool health degrades quickly, and tool condition impacts the process physics and joint quality significantly. Moreover, tool replacement constitutes a notable portion of maintenance costs. In UMW, tool health degradation occurs mainly in the form of changes in tool surface geometry. Online TCM, which uses in-situ sensing data to indirectly classify tool conditions, has demonstrated to be effective; however, such indirect methods cannot provide a detailed characterization of tool surface profiles. On the other hand, direct measurements of tool surfaces require expensive and time-consuming high-resolution 3D metrology, which substantially increases the cost of quality and delays maintenance decision-making. To overcome these challenges, this paper develops a fast and cost-effective imaging system for fine-scale TCM in UMW. The imaging system mainly consists of a macro lens and a Raspberry Pi (RPI) high quality camera mounted over a linear rail driven by a stepper motor, and the full system is controlled through RPI GPIO (general-purpose input/output). The imaging system is used in conjunction with coaxial illumination, which enhances tool surface features, to capture a photo of a cast reproducing tool surface geometry. Then, image processing techniques are developed to characterize tool surface profiles and features. We demonstrate the effectiveness of the proposed TCM strategy using tools in three distinct conditions. Results show that the TCM imaging system can effectively reconstruct critical fine-scale geometric features of tools, thus enabling more responsive, interpretable, and reliable TCM for UMW.