by Joseph Booth
•
1 November 2024
Artificial Intelligence (AI) has become a major buzzword across industries, yet its true potential and applications are often misunderstood. While AI technologies differ in capabilities, their impact on various sectors is undeniable. The ongoing advancements in AI will likely continue to reshape industries, but how is AI affecting electronics manufacturing today? And what does the future hold? Artificial Intelligence (AI) is increasingly integral to electronics manufacturing, particularly in the transition toward autonomous smart factories within the framework of Industry 4.0. AI technologies are driving significant advancements, transforming the industry and bringing it closer to the vision of smart manufacturing. While certain aspects of this transformation are realised, achieving a fully autonomous "lights-out" factory may still be a distant goal. Nevertheless, AI continues to expand the boundaries of what is possible in this field. The Vision of the Smart Factory The concept of the smart factory has been evolving over the years. While we are seeing tremendous progress, achieving 100% automation remains challenging. In my view, factories will always require some level of human involvement, particularly in the final stages of automation. The first 90% of automation delivers the most value, but the last 10%—the phase that seeks to eliminate all human intervention—will be the hardest to achieve. Despite these challenges, the next 10-15 years will likely see a dramatic transformation in manufacturing, driven by AI tools. These tools already address long-standing pain poi nts and enable process improvements that once seemed impossible. The Smart Manufacturing Vision Pyramid We can look at the 2022 Smart Manufacturing Vision Pyramid to better understand the progress toward the smart factory. This framework highlights the different stages of smart factory development, from basic automation to full autonomy. Many companies have achieved aspects of Smart 2.0, while the technologies associated with Smart 1.0 have become standard practice. The pyramid’s stages are built on key technological advancements, each of which plays a critical role in moving toward full autonomy: 1. Data Acquisition: This foundational stage involves collecting high-quality, accurate, and repeatable data from various sources across the manufacturing environment. Robust data acquisition lays the groundwork for all subsequent stages of smart manufacturing. 2. Insights: Once data is collected, it must be analysed. Statistical analysis and AI-driven insights allow manufacturers to understand their operations better and identify areas for improvement. 3. Digital Twin and Simulation: The next stage involves creating a digital twin—a virtual model of the physical production environment. This simulation can be used to experiment with different variables, such as staffing levels, machine placement, production volumes, and process flow. By optimising these factors in the digital realm, manufacturers can make more informed decisions before implementing changes in the real world. 4. Prediction: AI can predict future outcomes by leveraging both real and simulated data. For example, AI can forecast maintenance needs, enabling predictive maintenance that reduces downtime and minimises the risk of equipment failure. This predictive capability is essential for optimising production and reducing costs. 5. Autonomy: The final stage of the pyramid is autonomy, where machines communicate with each other, make real-time adjustments, and correct errors without human intervention. Examples include autonomous process tuning, automatic calibration, and machine-to-machine communication for seamless production flow. While full autonomy remains a goal, many of these capabilities are already in use today, thanks to pioneers in AI for Surface Mount Technology (SMT) manufacturing. Real-World AI Applications in SMT Manufacturing A leading example of AI in SMT manufacturing comes from Koh Young, a company at the forefront of measurement-based inspection processes. Koh Young’s solutions generate vast amounts of data on products built in SMT lines, which can be leveraged to enhance production quality and efficiency. While valuable on its own, data becomes exponentially more powerful when combined with AI tools like machine learning. For instance, inspection systems rely on test programs with set tolerances to ensure product quality. However, creating these programs can be challenging, especially when limited data is available, as is often the case during New Product Introductions (NPI). Insufficient data can result in weak inspection programs that either fail to catch defects or generate too many false alarms. To overcome this, AI can generate additional data using technologies like Generative Adversarial Networks (GANs) and Convolutional Neural Networks (CNNs). These AI-driven techniques can simulate thousands of products, effectively "training" the inspection system to perform better, even with limited initial data. The result is a more robust inspection process that reduces the likelihood of defects and minimises the burden of false positives on engineers. Another example of AI in action is the optimisation of the interaction between screen printers and Solder Paste Inspection (SPI) machines. Traditionally, SPI systems would inspect solder paste deposits and flag issues for engineers to address manually. However, this process was time-consuming and relied heavily on the engineer’s experience. Today, AI allows SPI systems and printers to communicate directly. The SPI system identifies issues and makes real-time adjustments to printer parameters, such as printing speed and pressure, ensuring optimal performance. This automated feedback loop reduces the need for manual intervention and enhances the overall efficiency of the production line. The Future of AI in Electronics Manufacturing Looking ahead, one AI-driven capability that has the potential to revolutionise production is mounter diagnosis. However, this technology may initially be limited to manufacturers with pre-reflow Automated Optical Inspection (AOI) systems, which require a significant investment. Despite the cost, the benefits are substantial. Mounter diagnosis uses AI to analyse real-time production data and pinpoint the root causes of issues within the mounter, such as problems with the nozzle, head, feeder, reel, or component. By identifying the exact source of a problem, AI enables engineers to address issues quickly, reducing downtime and improving production quality. This capability is particularly valuable in high-volume manufacturing environments, where even minor disruptions can significantly impact overall efficiency. These examples demonstrate that AI is not just a buzzword in electronics manufacturing—it is already delivering tangible benefits. From improving inspection accuracy to optimising machine interactions and diagnosing equipment issues, AI is helping manufacturers move closer to the vision of a smart factory. The Future AI in Manufacturing While the concept of a fully autonomous, "lights-out" factory may remain a distant goal, the advancements we are seeing in AI today are transforming the manufacturing landscape. AI is not something to be feared; rather, it should be embraced for its unprecedented opportunities. Much like we wouldn’t revert to using an abacus instead of a calculator or drawing by hand instead of CAD software, AI is poised to replace many manual, time-consuming tasks. By alleviating the burden of repetitive, labour-intensive work, AI allows skilled workers to focus on higher-level challenges, ultimately driving innovation and efficiency in manufacturing. As AI continues to evolve, the possibilities for its application in electronics manufacturing will only grow. The smart factory of the future may not be completely autonomous, but with the help of AI, it will be smarter, more efficient, and more capable than ever before.