Issue link: https://iconnect007.uberflip.com/i/1305670
30 SMT007 MAGAZINE I NOVEMBER 2020 based on the respective experience of the engi- neer. Deep learning methodology is designed to assist with this daunting task of process optimization by predicting optimal param- eters based on historical data. The goal is to achieve the highest production quality regard- less of the capability or experience of the engi- neer. The performance of the line must then be monitored in real-time through continuous col- lection and analysis of production data. The concept is as follows. Let's say we have collected historical big data for squee- gee speed, squeegee pressure, etc., across multiple screen print- ers and across various PCB models. At the same time that we are collecting this process data, we are also collecting inspection results data from the SPI machines. By ana- lyzing and correlat- ing the inspection results with the var- ious squeegee speed and squeegee pressure settings, we can proac- tively determine the opti- mal parameters for a given application, which may then be applied at the inception of the SMT manufacturing process to ensure optimal performance. Keep in mind that this is just one of the machines within the line. The same methodology is applied to the entire SMT manufacturing line. Johnson: That process optimization, real-time adjustment is happening in the database with the AI engine, and that AI expertise is some- one else's core competency. Is there a lot of collaboration with the companies specializing in that work? D'Amico: Absolutely. AI is a very large and comprehensive term. The AI that I am refer- ring to is, of course, specialized toward the SMT manufacturing process. But AI and big data are used worldwide for many different applications. As an example, many of us have visited websites online only to be greeted with a test to see if we are robots. These tests typi- cally direct you to "click on all the images that show a bus, bridge, bicycle, pedestrian, etc." Few people realize, however, that this test is part of an AI engine collecting big data that is then provided to assist in the programming of autonomous vehicles. The bottom line is that an AI engine is highly specialized for any given application. Therefore, there must be a great deal of collaboration between the AI team and SMT equipment suppliers, as what we're doing in our industry is very specialized versus other industries. Johnson: How does M I RT E C c o n n e c t with the AI engine? Do you leverage a third-party tool or a group of approved A I s u p p l i e r s , f o r example? D'Amico: Our AI engine is specifically designed for our products. However, the system must be flexible enough to be able to process data from multiple pieces of produc- tion equipment, including competitive inspec- tion systems. With this in mind, MIRTEC has made a strategic decision to partner with companies like Cogiscan that specialize in machine-to- machine communication, allowing us to con- nect to virtually any machine within the man- ufacturing line. Machine data is then collected, formatted, and entered into a repository from which it is made available to other systems within the line. This also overcomes the hurdle of work- ing with some competitive systems. Together, MIRTEC and Cogiscan have collaborated on a fully integrated Industry 4.0 solution, which is