Issue link: https://iconnect007.uberflip.com/i/1189040
DECEMBER 2019 I SMT007 MAGAZINE 13 Matties: The material is a hard and fast equa- tion. You could show the numbers. It's quan- titative. Manor: The challenge with material is it requires change management. You have to work differ- ently, as well as the operators; you need to go into a more automated approach, and you might get resistance. We need a strong man- agement team to be on board to say, "We are going to dictate it and make the operators and line managers use the software. If they turn off the software because it gives them a headache, and they stop collecting the traceability data, then this whole thing falls away." Matties: What is keep- ing people from jump- ing into digital factory? You've just authored a great book about what they need to know, but what's really the bar- rier? Manor: Data acquisition is the first barrier. Col- lecting data is not easy. It's not easy on the SMT shop floor, and we have a lot of different data formats and machines. It's a very heterogeneous environment; there is no real standard. CFX is a nice beginning, but it's not deployed yet. It's still evolving, and it's going to take a couple of years until we see a production flow with a lot of CFX data. Then, you have to collect all of the different types of data, and some of these machines, such as SPI machines, are generating thousands of measurements per second. If you have a high capacity, you are making millions of records a day, which is an overload. You have to col- lect the data. Now, you have to worry about storing the data. The biggest challenge later is how to get any insights from the data because you have a massive solution. What do you do with it? Customers want insights; they don't want to pay for the tool and have them do the stuff. They expect us as a digitalization partner to come with the solution, get the data, store it, and give them insights. They expect us to say, "We have looked at different types of correlation. We are going to run this engine on the data now, and try to see if it is applicable for your factory." For exam- ple, consider humidity and quality. Can we save some of the inspection processes? We did a very nice project with our colleagues in one of the Siemens factories in Germany, and we looked at the X-ray, which is a bottleneck machine. We decided to see if the test on the SPI and the AOI are per- fect. Can we skip the X-ray? In 30% of the cases, there is zero chance of a failure found in the X-ray out of 10 million PCBs examined. You can definitely skip that. And since X-ray is the bot- tleneck, you immediately get a 15% improvement of the yield and output. Mak- ing all of these tests and inspections makes the pro- cess more cumbersome. If you can save and test when you need on the part of the board that you need, you can save money, and this can be done with big data. Matties: An argument could be made that data already exists in the X-ray because they know that their jobs are coming out error-free, and they don't need your tool to determine that. Manor: You have to look at millions of records and try to see over because maybe there were three defects because the operator was clumsy. Maybe the air conditioning didn't work, so you want to rule out the environmental stuff. If you look at data for a year, then you rule it out, and you see on average when you can come to decide, "Is my test strategy correct? Can I save some time?" This is a nice example of big data analytics and how I think of using it.