Advantech improves DIP solder joint inspection efficiency by upgrading the AI.AOI PCBA solution with WISE-PaaS/AIFS
Printed Circuit Boards Assembly (PCBA) manufacturers understand that visual inspection for Dual In-line Package (DIP) solder joints requires the use of complicated procedures, eyesight, and extensive training. Similarly, glare caused by highly reflective solder joints and eye fatigue create circumstances in which defective PCBA are missed — and then passed onto subsequent workstations — creating quality control issues. To address these ongoing issues, PCBA manufacturers are seeking an "AI-assisted machine vision" technology solution.
To meet these needs, Advantech spent one year developing the Vision AI.AOI/PCBA solution at a trial site at the Linkou Manufacturing Center. The AIFS/AI.AOI PCBA, an industrial App (I.App), successfully integrated AI Automated Optical Inspection (AOI) functions into the PCBA visual inspection procedure. This shortened the solder joint defect inspection time to seven seconds and increased the production efficiency of electronics factories.
Detect PCBA solder joint defects in 7 seconds
Despite long-term use in the electronic product manufacturing industry, DIP AOI technology has many disadvantages. For example, AOI equipment uses rule-based measurement and does not support newer measurements or detailed inspection. Similarly, AOI measurement programming procedures are exceedingly complicated. The combination of requirements for near 100% fault detection and high misjudgment rates caused by restrictive rule-based measurements create further difficulties. Even if the AOI procedure is completed, considerable human resources are required for re-inspection procedures.
The improvement in efficiency derived from the integration of AI and AOI equipment has transformed operation scenarios for PCBA solder joint defect detection solutions. Take Advantech's AIFS/AI.AOI PCBA solution, which consists of two industrial apps, as an example. The first industrial app is for edge computing and features built-in pre-trained AI models aimed at helping factories quickly establish a standard PCBA DIP AI AOI operation procedure. Factories simply feed images of PCBA rear side solder joints into the inference engine. Following this, AI inspection can be completed in 7 seconds. This inspection identifies short circuits, solder skip, excessive solder, insufficient solder, insufficient wetting, appearance, solder beads, and other defects. In sum, this system increases detection efficiency by more than 30%. Similarly, fault detection rates reached 99%, and the misjudgment rates dropped to lower than 10%.
The above-mentioned pre-trained AI models are the result of long-term image labeling and procedure importation from Advantech’s team. The model training process used thousands of image files and data on tens of thousands of DIP defects. Using pre-trained models reduces time spent adjusting the AI.AOI PCBA architecture. It only requires 1 to 2 months to complete all the deployment of all procedures and launch it online.
Realize lean production while improving inspection efficiency and solder joint quality
The second AIFS/AI.AOI PCBA solution example is a retraining industrial app. In applications with large scale or high-mix-low-volume production; false positives or overkills are generally unavoidable during AI interpretation processes. Addressing the situation requires a retraining mechanism capable of continuously optimizing AI models as needed. Using this industrial App, production line operators can conduct the entire retraining process independently. This reduces the need for engineering department assistance. Users simply upload labeled images to the AI model training server, and then launch Advantech’s WISE-PaaS/AIFS (AI Framework Service) platform for AI model retraining and redeployment.The WISE-PaaS/AIFS platform provides a friendly and intuitive web user interface through which users can set deployment conditions. These deployment conditions regard when specific times and rules are met, or when fault detection rates pass a predetermined percentage. These conditions can be extended to include when and where the deployment of an AI model on production equipment is necessary. Using these preset deployment conditions, users can deliver notification alerts to system maintenance personnel. Additionally, using this convenient and smooth retraining process gradually improves fault detection rates.
In sum, effective PCBA solder joints defect detection remains a time-consuming and labor-intensive necessity in electronics manufacturing. Identifying flaws and foreign objects requires visual inspection on individual PCBA solder joints in well-illuminated environments. Even so, ensuring that no abnormal product passes to the next workstation is insufficient. As the benchmark for visual inspectors is high, the position requires 6 months training to obtain the required skillset — creating ongoing and troublesome human resource demands.
Advantech’s Vison AI.AOI/PCBA solution helps inspectors’ compensate for vision loss and prolong their occupational life. Similarly, newcomers only need 3-months on-the-job training to be competent for inspection tasks. Advantech’s solution can help factories improve inspection efficiency and solder joint quality while simultaneously achieving lean production and implementing smart manufacturing. Vision AI.AOI/PCBA solution delivers a comprehensive package of solutions.
The WISE-PaaS data application platform end-to-cloud architecture provides a comprehensive development environment for data collection, analysis, management, and applications, while promoting interoperability between I.Apps via the WISE-Marketplace and accelerating the development of industrial solutions.
Advantech's AI-driven AOI solution for PCBA DIP solder inspection aims to facilitate the inspection of soldering quality, thereby improving visual inspection accuracy, worker efficiency, and production processes.