Detecting Corrosion to Prevent Cracks in MLCCs with AI

Real-Time Defect Detection in Electronic Components during Assembly through Deep Learning

“…Additionally, we documented several use cases of corrosion detected using this method, subsequently verified by lab analysis [16,34]. Moreover, we devised a method to mitigate the risk of cracks in MLCCs (multilayer ceramic capacitors) by the early detection of the evidence of corrosion in the soldering terminals as we demonstrated that corrosion serves as a precursor for crack development [12].…”

“…In each section, we highlight the standard sections that are relevant to the presented method and that the method can detect automatically on all inspected components. The detection algorithm is only mentioned in this work and is elaborated in [12,16,30,32,34]. In this work, the novelty is in adapting the methods to real-time operation.…”

“…The visible defects are scratches, dents, distortions, deformations, peeling, shorts, etc. this work and is elaborated in [12,16,30,32,34]. In this work, the novelty is in adapting the methods to real-time operation.…”

“…Figure 4 presents examples of bent leads and coplanarity issues automatically detected by the algorithm that exceed 10% of the lead's width. this work and is elaborated in [12,16,30,32,34]. In this work, the novelty is in adapting the methods to real-time operation.…”

“…Defects encompass a wide range of issues, spanning from cracks [11][12][13][14][15], fractures, and the peeling of metallization to deformations, discoloration, mold, corrosion [4,[16][17][18][19][20][21][22][23], bent leads, deformed leads, and misshapen BGA balls, among others (see examples in Figures 1 and 2). To provide clarity and establish a baseline for defect identification, leading standards such as IPC-A-610H [3] and IPC-STD-J-001 [24] offer comprehensive definitions of defects on assembled PCBs.…”