In the actual operation of our production line, we found that once bridging defects are cured at the high temperature of Reflow Soldering, metallurgical bonding has already formed between the solder joint metals. At this point, forcibly separating them with a soldering iron or hot air gun not only easily damages the pads and PCB substrate, but also causes adjacent component displacement or cold soldering under the BGA. Therefore, the inspection node must be moved forward to before reflow, and the JiajinYuan AOI system is precisely a real-time interception tool designed based on this rigid demand.

Bridging already lays hidden dangers during the Solder Paste printing stage. For example, burrs on the edge of the stencil aperture will cause solder paste trailing, uneven squeegee pressure will lead to solder paste accumulation, and component offset during placement may squeeze solder paste between adjacent pads. None of these factors alone is sufficient to immediately determine as a defect, but when superimposed, they appear as abnormal continuous grayscale bands between pads on the AOI image. JiajinYuan's algorithm precisely triggers early warning by capturing this subtle grayscale gradient change.

We have debugged three types of 0201 resistor placement boards with different thicknesses. When the stencil thickness increased from 0.1mm to 0.12mm, the AOI's bridging detection rate increased from 89% to 97%, but the false alarm rate also rose by 2.3 percentage points synchronously. This shows that parameter setting cannot only pursue high detection rate, but also must dynamically balance with the current production line's changeover frequency and manual re-inspection capability, otherwise it will seriously slow down the UPH of the entire line.

The light source angle of JiajinYuan AOI supports ±15° mechanical adjustment. When testing 0.4mm pitch QFN, we found that adjusting the incident angle to 12° can significantly enhance the reflection contrast of the pad edge, making the solder paste bridging area present a clearer bright band feature in the image. If the default 18° is used, it is easily confused with the highlight of the solder paste body, leading to missed judgment. This kind of detailed adjustment must be completed on site by experienced technicians familiar with this model of equipment, and cannot be done by one-click import of remote parameter templates.

A customer once reported that the AOI frequently alarmed but no physical bridging could be found. When we went to the site to disassemble and inspect, we found that there was a 0.08mm height difference in its conveyor track, causing the PCB to slightly warp in the inspection area, which resulted in a deviation between the AOI lens focal plane and the actual position of the pad, and local blurring appeared in the image. The originally continuous bridging feature was misjudged by the algorithm as discrete noise points. After recalibrating the track levelness, the false alarm rate directly dropped to within 0.15%.

Bridging risk does not exist in isolation. It often occurs coupled with multiple process variables such as excessive solder paste volume, excessive placement pressure, and insufficient nitrogen purity in the reflow oven. The defect coordinate data output by JiajinYuan AOI can be directly connected to the MES system, and perform time-axis alignment analysis with SPI solder paste thickness data, placement machine nozzle vacuum logs, and furnace temperature curve collection points, so as to lock the real root cause process, instead of blindly replacing the stencil or adjusting the squeegee angle after the AOI alarms.