Issue link: https://iconnect007.uberflip.com/i/1305670
14 SMT007 MAGAZINE I NOVEMBER 2020 improvement, especially when leveraging AI solutions. Experts link the effectivity of deep learning to the quality and quantity of the data used to address many different requirements from many fields. Most advanced inspection system providers want to use AI to deliver smarter inspection systems. However, it has been difficult to realize due to the limitations of 2D and 2.5D imaging. Not only is it difficult for 2D/2.5D AOI systems to identify solder and dimensional-related defects, but these systems cannot also gener- ate reliable measurement-based data. Every aspect of 2D inspection relies on 2D features like contrast, binarization, and reflection; thus, it is extremely challenging to correlate to the quantitative measurement of 3D objects. Suppliers must measure the true 3D infor- mation of components and solder joints. If properly accomplished, the AOI can offer very valuable data, thus becoming the most reli- able "sensor" on the line (Figure 3). Compa- nies can only guarantee the validity of 3D data if the system uses true 3D technology to extract the exact body dimensions for all components types. "Blob detection" is simply not reliable and compromises the data; it also contributes significantly to false calls. The combination of multipoint measure- ment and process data collected from SPI, pre-reflow AOI, and post-reflow AOI systems, combined with data from printers, mounters, and reflow ovens will allow manufacturers to deliver an AI-powered, zero defect, self-heal- ing line. Indeed, the quality of data is more important than the quantity of data to create effective and reliable solutions with high value proposition. Yet, it is the combination of reli- able data, along with a statistically relevant quantity of data, that delivers the next level of relatable results. Connect the Data A single inspection system has limits and cannot manage and optimize a complete line while in isolation. Knowing this limitation, advanced companies have been working with the industry to connect its inspection systems with other process equipment like printers and mounters to streamline total communica- tion and realize a zero-defect future. Looking inward, companies must design and deliver suitable process controls for seamless com- munication across their own core set of equip- ment, consolidating the individual machines to deliver a synergistic effect. Realizing a smart factory means taking a practical approach to processes and systems while examining areas to improve productiv- Figure 3: The inspection machines can be the sensors. Accurate full 3D AOI data-based inspection.