AI-Based Inline PCB Defect Detection and Robotic Sorting Using Smart Vision, PLC Logic, Fanuc Vision and SCADA Integration

In the factories that build the world's electronics, a tiny mistake can cost a fortune. A single missing chip on a circuit board, a smudge of solder where it shouldn't be, or a component placed slightly askew can render a device useless. For decades, the solution was to have human eyes scan every board, a slow and exhausting task where fatigue inevitably led to missed errors. Then came machines that could see, using cameras and software to compare what they saw against a perfect template. But these early machines were rigid; if a board shifted slightly or the lighting changed, the machine would get confused, rejecting good parts or missing bad ones. The next step in this evolution was to give the machines a brain that could learn, much like a human does, to recognize patterns and defects without needing a perfect, unchanging template. This shift from rigid rules to flexible learning is the heart of modern industrial automation, promising a future where factories run faster, cleaner, and with fewer mistakes.

A team of researchers at Michigan Technological University has built a working model of this future, creating a complete system that inspects circuit boards and sorts them automatically without human help. They did not just write a computer program to find errors; they built a physical workstation where a camera, a controller, a robot, and a monitoring screen all work together as a single unit. The system uses a smart camera equipped with artificial intelligence to look at circuit boards moving along a conveyor belt. When the camera spots a board, it stops the belt, takes a picture, and instantly decides if the board is good or bad. If the board is defective, a robotic arm reaches down and removes it; if it is good, the belt starts again, and the board moves on. The researchers proved that this entire chain of events—from the camera seeing the board to the robot moving it—can happen reliably, even when multiple boards are on the belt at the same time.

The core of their achievement lies in how they managed the flow of information. In a busy factory, boards do not stop one by one; they often travel in a stream. This creates a tricky problem: if a camera inspects a board and then the belt moves, how does the robot know which board to pick up later? If the system gets confused, it might throw away a good board or leave a bad one behind. The researchers solved this by using a digital queue, a method of holding onto inspection results in the exact order they arrive, like a line of people waiting for a ticket. When a board arrives at the camera, the result is placed at the back of the line. When that same board arrives at the robot station, the system takes the result from the front of the line and uses it to tell the robot what to do. This ensures that every board is judged by its own inspection record, no matter how many other boards are moving around it.

To test this system, the team set up a real-world environment with a conveyor belt, two sensors to detect when a board arrives, a smart camera, and a six-axis robotic arm. They trained the camera's artificial intelligence to recognize missing integrated circuits, which are the small black chips that are the brains of the circuit board. The camera was taught to look for the specific shapes of these chips and to count them. If the camera saw the correct number of chips, it marked the board as passing. If it saw fewer chips than expected, or if the chips were in the wrong place, it marked the board as failing. The system was designed to handle the messy reality of a factory floor, where boards might not stop in the exact same spot every time or where the lighting might vary slightly.

The results showed that the system worked as intended. In a series of tests involving fourteen circuit boards, the system correctly identified which ones were good and which ones were bad. It successfully distinguished between boards that had all their components and those that were missing them. The camera took less than a second to make its decision, and the robotic arm moved with precision to sort the boards. The researchers found that the system was particularly good at handling the physical movement of the boards; the digital queue never lost track of which result belonged to which board, even when the conveyor belt was moving multiple items. The robotic arm, guided by its own vision system, was able to pick up the boards and place them in the correct bins, whether that meant sending a good board down the line or tossing a bad one into a reject bin.

However, the researchers were careful to note that this was a prototype, a working model built to prove the concept rather than a finished product ready for a massive factory. While the system achieved an accuracy of about eighty percent in their specific tests, they acknowledged that it is not yet perfect. The system struggled a bit when the boards were not positioned exactly the same way every time, and the speed of the conveyor belt was limited because the belt had to stop for the camera to take a picture. These are practical hurdles that would need to be solved before such a system could run at the high speeds of a major electronics manufacturer. The team also noted that the artificial intelligence model was trained on a relatively small set of images, which limited its ability to handle every possible type of defect or lighting condition.

Despite these limitations, the project demonstrated that it is possible to combine advanced artificial intelligence with standard factory equipment to create a fully automated inspection line. The system did not rely on a single powerful computer in a distant server room; instead, the intelligence lived inside the camera itself, making decisions locally and instantly. This approach allowed the camera to talk directly to the factory controller, which then told the robot what to do. The entire process was monitored on a screen that showed the status of every part of the system in real time, allowing a human operator to see exactly what was happening at a glance. This integration of smart vision, logical control, and physical sorting represents a significant step forward in how factories can ensure quality. It shows that the gap between high-tech research and the actual machinery on a factory floor is narrowing, bringing us closer to a world where machines can not only see defects but also understand them and fix the problem without human intervention.