A Method for Real-Time Rejection of Defective Components During the Pick and Place Process Using AI Technology
This paper presents a real-time, AI-driven defect detection system designed to enhance quality control in electronic assembly, providing manufacturers with precise, customizable screening parameters. The system utilizes a two-stage approach: an initial high-sensitivity detection phase that aligns with IPC-A-610 standards, followed by a secondary defect classification stage, activated only on approximately 200 defective parts per million (ppm). This configuration enables high-speed processing with minimal impact on computational resources, making it suitable for high-throughput environments.
The approach includes a “quality control knob,” allowing manufacturers to adjust defect thresholds, selectively target specific defect types, and apply tailored inspection parameters to different component classifications. This flexibility allows a balance between high-quality standards and production efficiency, reducing attrition rates and minimizing waste. For example, setting the quality threshold to meet IPC-A-610 standards results in a defect rejection rate of 207 DPM at an attrition cost of $0.62 per million components, while more lenient settings can reduce attrition to 59 DPM, costing only $0.18 per million components. This enables manufacturers to make informed trade-offs between quality and cost.
In addition to improving product reliability, the system reduces rework, scrap, and the environmental impact associated with production inefficiencies. Test results demonstrate the system's effectiveness in detecting diverse defects such as corrosion, foreign object debris (FOD), bent leads, and structural damage, allowing manufacturers to uphold stringent quality standards. This flexible, adaptive inspection method not only enhances quality control but also advances sustainability in electronic manufacturing by significantly lowering defect-related issues.