Issue link: https://iconnect007.uberflip.com/i/1509873
74 PCB007 MAGAZINE I OCTOBER 2023 may reject their suppliers' cost-cutting efforts through the use of new technology if they have the potential to negatively impact product quality, which is their top priority. is is why they may be reluctant to accept any ML imple- mentation by their suppliers that might jeopar- dize that goal. Harnessing Machine Learning and AI for Smart Factories PCB manufacturers, on the other hand, are in the business of making money, and cost sav- ings with such technology directly translate to higher profit margins, though they may face resistance from their customers, particularly those with significant bargaining power. Finally, there are practical and technical challenges associated with AI development and implementation, the most significant of which is the collection and processing of massive amounts of high-quality pre-classi- fied (labeled) data required to develop a via- ble model. e data provided by PCB manu- facturers can be messy, particularly in terms of classification accuracy. As a result, AI develop- ers would have to expend significant effort to clean and reclassify the data in order to make it suitable for AI model training purposes. is is a time-consuming process that frequently results in the discarding of some amounts of low-quality data. In conclusion, the success or failure of AI technology adoption for the AOI process is primarily determined by the optimal balance between the rate of escapes and the overall filter rate, resolving issues around accurate defect specs while periodically retraining AI models with new and relevant data, and navi- gating the agency dilemma between PCB man- ufacturers and their customers. Machine Learning for Optimizing Production Processes ML can significantly enhance the optimiza- tion of PCB production processes, including production planning and scheduling, as well as inventory management. For example, ML models can analyze historical sales data, mar- ket trends, and other relevant factors to fore- cast future demand more accurately. is infor- mation helps in setting production targets. ML algorithms can optimize resource allo- cation by considering factors like machine availability, labor capacity, and job specif- ics (such as complexity and required process flow, as well as raw material availability) to create efficient production process routes and schedules. For example, the AI models that we have developed allow users to identify the optimal route through production processes down to a specific equipment, which max- imizes efficiency and yield while minimizing production time and the costs. Such a model learns from the vast database of historical data on the performance of each indi- vidual equipment with different types of jobs and their effect on the overall defect types and rates. One of the biggest practical chal- lenges in Smart factory adoption and implementation is the diffi- culty in collecting reliable and con-