Rapid advancements in information technology are driving the miniaturization of electronics using smaller and more sophisticated electronic components and printed circuit boards. In the past, PCBAs were checked through manual visual inspection. However, as miniaturization makes parts smaller, defects are becoming harder for inspectors to detect with the naked eye, making accurate and consistent checks difficult. Now, more and more manufacturers are looking for machine vision solutions to meet their business needs.
Automated optical inspection (AOI) based on machine vision technology, combined with AI and deep learning, can significantly improve the accuracy and efficiency of inspection. Electronics manufacturing plants are switching from manual visual inspection of PCBAs to AOI to cope with increasingly sophisticated inspection standards and ultimately increase production capacity.
An automated equipment manufacturer in southeast Asia used the Cincoze GP-3000 with the GEB-3301 GPU expansion box and NVIDIA GeForce RTX 3070 GPU card for a PCBA defect detection system. The company utilizes AOI and deep learning software to accurately detect and flag PCBA defects, such as surface defects, soldering defects, and missing or misplaced components. The GP-3000's superior performance enables the analysis of highly complex images and can automatically separate working and defective products.
An AOI system must be equipped with a high-performance embedded computer to quickly and accurately perform image recognition and model operations on high-resolution images. The core embedded computer also needs to connect with many peripheral high-speed devices to complete the AOI system, so a high-performance CPU is required to support multitasking and intensive workloads.
GPU Accelerated Deep Learning
A system incorporating deep learning must continuously accumulate training data. Thousands of images are used to improve the calculations and inference needed to detect different defect types, enabling the software to learn to better distinguish between defective object images. Therefore, in addition to the CPU, a 220W NVIDIA GeForce RTX 3070 GPU card needs to be deployed to speed up the graphics calculation process.
PCI Expansion and Rich I/O
The AOI system integrates with an infrared sensor, industrial camera, image capture card, light source, motion controller, and other components. The conveyor belt transports the PCBA to the shooting platform, where an infrared sensor senses the PCBA, triggering the cameras to take pictures, which are sent to the computer through the capture card for analysis. If the software detects a defect, it will trigger a mechanical arm to move the PCBA aside. Therefore, the computer needs abundant I/O interfaces such as USB, COM, LAN, DIO, and PCI to connect to peripheral devices.
Xeon® Workstation CPU
The GP-3000 has a 9th/8th gen Intel Xeon®/Core™ workstation-class high-performance CPU with up to eight cores and supports two sets of DDR4-2666 ECC/non-ECC SO-DIMM memory up to 64GB. As the core of the machine vision system, the GP-3000 supports a large power consumption of 720W to meet the requirements of dual GPU cards and full-load work.
GPU Card Expansion
The GP-3000 can expand to two 250W high-end full-length GPU cards through the exclusive GPU expansion box (GEB), with the GEB-3301 supporting a single GPU card and the GEB-3601 supporting two. The Adjustable 3D GPU Card Mounting Bracket (Patent No.: I763318) can provide solid locking of the GPU card. The GP-3000’s Dustproof Cooling Mechanism (Patent No.: I778522) provides effective heat dissipation to ensure that the CPU and GPU will not overheat under harsh conditions. The GP-3000 meets the demands of emerging GPU edge computing applications such as autonomous driving, visual inspection, and security monitoring.
Rich I/O and Modular Expansion
The GP-3000 has multiple built-in high-speed I/O ports, including 5x GbE LAN and 6x USB 3.2. Additional high-speed I/O cards, such as image capture cards, can be added through the GEB's multiple PCI/PCIe slots. In addition, Cincoze's exclusive CMI/CFM modules can expand I/O or other functions according to specific needs, such as 10 GbE LAN, USB, PoE, and IGN modules.
The powerful GP-3000 also has many robust features to operate in harsh environments, such as 9–48VDC wide voltage input and -40–70°C wide operating temperature. It has passed MIL-STD-810G military standard, E-mark, and EN 50155 (EN 50121-3-2 only) certifications.