EN 
ES 
FR 
AR 
PT 
RU 
DE 
IT 
In recent years, the electronics manufacturing industry has undergone profound changes, with artificial intelligence (AI) and machine learning rapidly emerging. These cutting-edge technologies have driven automation, optimized quality control, and provided valuable data and insights in the surface mount technology (SMT) manufacturing of printed circuit board assembly (PCBA). Integrating AI into electronics manufacturing marks a shift from traditional methods to smarter, more adaptive approaches. On the other hand, the fusion of AI and AOI presents a transformative opportunity. Overcoming challenges, the SMT industry has adopted AI solutions driven by advancements in hardware, machine vision, and AI algorithms. These developments have strengthened mechanical structures and reduced false positives. In industries with high reliability requirements, such as automotive and aerospace, AI addresses the shortcomings of automated inspection while meeting production demands and continuously improving processes. Coincidentally, Nectec’s pick and place machines and x-ray machines all have implemented the AI real-time computing functions to better assist the precision and speed of chip component mounting. Interestingly, what makes AI an indispensable part of the SMT manufacturing process. There are five points that are worth mentioning.
First point, AI’s automation assistance. Early on, AOI programming involves engineers manually configuring hundreds of inspection parameters based on PCB CAD data. This complex and tedious process can take up to 10 hours for each new design. AI programming solutions have transformed this process by automatically generating complete AOI programs in minutes without human intervention. These automated programming tools work by analyzing PCB design files, bill of materials, component shapes/sizes, and automatically proposing optimal inspection conditions. Machine vision and deep learning algorithms can quickly extract key information from design files to recommend inspection programming suitable for the PCB. This automation simplifies switching between circuit boards.
Second point, AI’s reliable inspection system. One of the most useful advantages of AI inspection is its reliable detection system for visual inspection of common defects and complex surface components. When inspecting damaged parts of SMT components, such as chips, integrated circuits, connectors. And actually, it is difficult to predict the appearance of the damaged parts. By training on examples based on human learning, AI can learn how to identify defects. Currently, component types such as SMD chips, LEDs, OSC, MLD, SOD, SOT23, RNET, CNET, ICs, and connectors achieve high accuracy rates. It is recommended to consult the AI model owner to determine which types are available for enabling this feature, thereby enhancing verification accuracy and reducing the operator’s workload.
Third point, AI inspection algorithm implementation. Traditional OCV/OCR algorithms require separate training and consume significant time and manpower to configure. OCV/OCR is easily disrupted by font differences and missing characters, resulting in a high false positive rate, sometimes can reach over 10%-20%. AI OCV/OCR has built and fine-tuned a font library optimized for character accuracy. AI OCV/OCR can easily detect low-contrast characters, which is nearly impossible for traditional methods. The presence of low-contrast defects and noise in imaging poses a challenge in optical inspection, such as voids in X-ray inspection and adhesive on surfaces in optical inspection.
Fourth point, Smart sorting function implemented within AI. AI can not only detect defects, but also intelligently classify them by type, severity, and source of origin. This classification allows for targeted root cause analysis, reducing recurrence and contributing to a more robust quality control system. One example of classification is for new components with varying shapes and sizes, which typically require reprogramming of the inspection system. AI addresses this challenge by enabling the inspection system to quickly train on new components without the need for reprogramming. Another AI inspection application is foreign object detection.
Last and which is fifth point, Smartness and flexibility of AI in AOI. The flexibility of AI enables AOI systems to handle various component types, colors, and circuit board materials without reprogramming. By training on representative images that include expected variations, AI algorithms learn to distinguish between acceptable process variations and true defects. This flexibility is particularly valuable in high-mix production environments where product variations are frequent. As a result, post-optical inspection repair stations are becoming smarter through AI; these stations now leverage AI’s potential to make human-like decisions, reducing the need for manual reinspection, lowering operational costs, and providing real-time inspection status data analysis.
To conclude, AI’s transformative power in inspection and smart factories promises unparalleled efficiency, adaptability and quality assurance. It reshapes the future of innovation and automation-driven electronics manufacturing. The impact of AI extends beyond inspection to the entire electronics manufacturing ecosystem, and we look forward to this new era.