Issue link: https://iconnect007.uberflip.com/i/1536435
JUNE 2025 I PCB007 MAGAZINE 41 What does the data flow look like in closed- loop optimization and real-time corrections? ere are two basic scenarios. First, you have an external, in-line optical inspection tool between the various steps of the plating process. is tool takes pictures of the panel and understands them. en, if a defect is detected or predicted, the tool can send a signal (a command, actually) to the control soware of the tool, telling the equipment, for example, to stop or to change the process to reduce the detected dri. at's a closed loop. But the challenge in this sce- nario is that, typically, the sys- tem integration is difficult because you have an external tool that must be able to talk to the control system of the equip- ment and might need addi- tional devices (loaders, unload- ers, etc.). Both the equipment and the external tool may each have different real-time protocols and con- straints/specifications, and the footprint would be impacted, for example. In the second scenario, the in-line optical inspection could be embedded in the equip- ment, either at the beginning or at the end of the step, based on the use cases. Out of the box, the in-line AOI tool indicates a dri in the process, and an adjustment is needed. e dri can be adjusted because we can see a trend and change the recipe, for example. at is also a closed loop. is is a more straightforward scenario, but the PCB manufacturers are "forced" to use the in-line quality system available inte- grated in the tool itself. Another area to compare the two scenar- ios could be on the energy consumption: It is possible to have a real-time, closed-loop system that can reduce the energy consump- tion during a process without impacting the quality. e plating equip- ment has many components that consume energy and pro- duce heat. If you find the opti- mal sequence, you will save energy. ere's an intelligent closed-loop engine that can do that (second scenario). If you require an external system to do it (first scenario), it will take significantly more time because it has to integrate, understand, and then commu- nicate. If it's embedded in the machine, it is much faster and brings value immediately because the feed- back loop is designed for that exact process and equipment. How does a machine notify you when defects are found? Can you customize that output? Yes, you can customize it. e AI models can be trained to identify not only the defect but also the level of acceptance or the sever- ity of the defect. Tolerance based on thresh- olds can be set, and several behaviors can be examined. e maintenance engineer, " We can get to that defect identification, prediction, and reduction before so much cost is built into the panel. "