Beyond Inspection: How APEX-X100 Optimizes PCB Solder Inspection with AI-Driven AOI

Innodisk treats its modular product lines as intelligent building blocks that seamlessly integrate hardware, computing platforms, software, and firmware into ready-to-deploy Edge AI Systems. Designed with flexibility in mind, Innodisk’s edge AI systems offer configurable GPU options, memory and storage capacity, and expansion modules, enabling customers to achieve the optimal balance of performance and cost. 

The APEX-X100 is a compact edge AI system powered by up to an Intel® Core™ i9 processor and an NVIDIA Blackwell Max-Q GPU. With its industrial-grade design and versatile I/O support, it serves as a reliable foundation for smart factory AOI applications.

Enhanced Computing for Solder Inspection 

Soldering is a critical step in nearly all PCB assembly processes, serving both as a mechanical bond and an electrical connection for components. However, insufficient solder can lead to poor contact, while excessive solder can flow to adjacent joints, causing short circuits between pins. In addition, factors such as moisture in the PCB material, contamination, or process inconsistencies can result in defects such as voids or solder splatter.

With increasingly complex PCB layouts and denser solder joints driven by AI systems, the customer upgraded its vision-based AOI station. In addition to the existing 2D cameras used for locating solder joints and detecting defects, a 3D camera was introduced to enable precise measurement and reduce the impact of reflective surfaces. However, the resulting surge in image data required a more powerful edge AI system.

The customer selected the Innodisk APEX-X100 with an Intel Core i9-14900 processor. The CPU enables real-time fusion of 2D image data and 3D depth information from multiple cameras. With support for up to 192GB of DDR5 memory, the system significantly reduces processing latency, providing a strong foundation for accurate and reliable defect inspection.

AI-Powered Decision Making

After visual data is collected, the APEX-X100 enables real-time, AI-driven decision making within the Manufacturing Execution System (MES). In this deployment, the customer selected the NVIDIA RTX PRO 6000 Blackwell Max-Q GPU to accelerate inspection and analysis. 

Unlike traditional AOI systems that rely on post-process review, where defects may only be identified after an entire batch has already been produced incorrectly, the AI-powered MES acts as an execution-driven system that responds immediately to inspection results.

For individual defects, the system can automatically flag and remove defective items. When anomalies occur at scale, it can initiate production control actions, including stopping the line and alerting engineers, effectively minimizing yield loss in real time.

Leveraging the GPU’s parallel processing capabilities, the system also enables cross-line data synchronization, helping manufacturers maintain consistent quality across production lines.

On-Device Continuous Learning

Powered by the Tensor Core acceleration of the APEX-X100 GPU, the AOI system goes beyond inference to enable continuous on-device learning. Through periodic background training, the system can adapt to new conditions and progressively improve its inspection accuracy over time.

Subtle changes in the production environment, such as lighting degradation or slight camera shifts caused by vibration or other unnoticed factors, are automatically accounted for, preventing false defect detection. 

As the system becomes more familiar with inspection targets, it can further identify emerging patterns and anticipate potential quality issues in PCB production. This shifts the role of inspection from simply stopping defective products to preventing defects from occurring in the first place, transforming reactive correction into proactive quality control.

The deployment, following the customer’s AOI station upgrade, resulted in measurable improvements in production efficiency and quality. First Pass Yield (FPY) and False Reject Rates both showed significant improvement, outperforming industry averages and helping minimize unnecessary rework while improving overall production stability.

As the core of the AOI station, the Innodisk APEX-X100 integrates the most suitable CPU and GPU, along with industrial-grade memory, storage, and rich I/O, providing the foundation required for AI-driven smart manufacturing and enabling a truly intelligent production line.

Disclaimer: The factory layout and content presented in this article are for demonstration purposes only. Due to confidentiality considerations, certain details may be incomplete or differ from actual conditions.