In our increasingly competitive global environment, increasing efficiencies in the manufacturing sector has been a key strategic objective. On this week’s episode of The Playbook, host Mark Collier, business consultant for the UGA Small Business Development Center, is joined by Dr. Jack Hu, Senior Vice President for Academic Affairs and Provost at the University of Georgia, and the UGA Foundation Distinguished Professor of Engineering.
Transcription:
Mark Collier:
I am very pleased today to welcome a true intellectual giant in the smart manufacturing space. Dr. Jack Hu. Dr. Hu’s own research in smart manufacturing, assembly, technologies, and systems has been supported by more than $46 million in external funding from federal agencies, such as the U.S. Department of Energy and the National Science Foundation, as well as Fortune 100 companies, such as General Motors. Dr. Hu has authored or co-authored nearly 200 peer-reviewed journal articles, as well as several book chapters, government, and industrial reports. In addition, Dr. Hu holds eight patents, is Co-Founder of a startup company based on his research, and worked closely with several key industry partners to enhance manufacturing quality and productivity among his numerous professional accomplishments. Welcome into The Playbook, Jack.
Dr. Jack Hu:
Thank you, Mark. Great to see you.
Mark Collier:
This past year’s pandemic has put manufacturing efficiencies at the forefront of most companies’ strategic goals. But before we launch into that, share with me and our ASBN audience, the vital role that you fulfill at the University of Georgia.
Dr. Jack Hu:
So the Provost and Senior Vice President for Academic Affairs is the Chief Academic Officer. I have oversight for instruction, research, public service and outreach, and information technology. In addition, for the schools and colleges, through the Dean, they report to me. So you can see the provost has a lot of meetings, a lot of responsibilities to ensure that we have outstanding programs for research and education, and also engagement with the state. As a professor, I teach and do research, but due to my role as Provost, you can see my time of doing research and also teaching has been limited, but I do work with students, supervise them, and mentor them on their projects.
Mark Collier:
The Society of Manufacturing Engineers recently named you one of the 20 Most Influential Academics in Smart Manufacturing.
Dr. Jack Hu:
Thank you for mentioning that.
Mark Collier:
All right. So let’s talk about smart manufacturing. Kind of give me an overview of what it is and why it’s important for small businesses to consider it.
Dr. Jack Hu:
So the definition of smart manufacturing has evolved over time. I would say in the seventies and eighties, it’s about computer aided design, computer aided manufacturing so the integration of CAD and CAM, so computer aided-design, computer aided-manufacturing. So Smart manufacturing today is more about integration of big data with artificial intelligence and smart machinery. When you make full use of the data that is available, and then try to extract useful knowledge from the big data and then make decisions on your machines and system and also enterprise so that you can improve manufacturing efficiency, improve manufacturing quality.
Mark Collier:
Well, you hit on two of the future trends, I mean, artificial intelligence and big data are buzzwords in all industries today and they’re very important for businesses to consider, because as you said, when you’re able to actually incorporate all the data and the artificial intelligence, you’re able to enhance your bottom line and that’s the goal of all companies.
Dr. Jack Hu:
That’s correct. So data is always part of smart manufacturing early on, I think the data may be not as rich, not as autonomous, not as fast. So you can manually collect data and then derive information from that to make decisions, right?
Mark Collier:
Yes.
Dr. Jack Hu:
So I think everything about smart manufacturing is really to make manufacturing organizations and learning organizations define what you want to measure and then from the data, you get a sense of what’s going on with your manufacturing system and with your enterprise, and then you can make improvement based on data.
Mark Collier:
Now that makes perfect sense. One of my recurring mantras that I impart to my clients is, “What cannot be measured cannot be improved.”
Dr. Jack Hu:
That’s exactly right. You can’t improve a process that you cannot measure.
Mark Collier:
Absolutely. All right.
Mark Collier:
So let’s kind of peel back a few more layers and kind of detail for me, kind of the similarities and differences between smart manufacturing and 3D printing.
Dr. Jack Hu:
So we talk a little bit about smart manufacturing using data, make decisions, and leveraging artificial intelligence or smart machines. 3D printing is a very innovative process, which is now very popular among manufacturers. So 3D printing is also called additive manufacturing.
Mark Collier:
Additive manufacturing
Dr. Jack Hu:
Additive, not deductive.
Mark Collier:
Okay.
Dr. Jack Hu:
That is additive.
Mark Collier:
All right.
Dr. Jack Hu:
So what that means is from computer aided-design, from your CAD file, you can generate an algorithm and send that to a printer which is a 3D printing machine. So you can make a part, dot by dot, or point by point, or drop by drop. For example, like a welding machine, an arc welding machine that has droplet so you can think of that as an additive process.
Mark Collier:
Yes.
Dr. Jack Hu:
Because you are welding two piece of metal drop by drop. Right?
Mark Collier:
Okay.
Dr. Jack Hu:
So that’s an additive process. You can also do layered manufacturing, so one layer at a time.
Mark Collier:
Okay.
Dr. Jack Hu:
For example, you have papers and then you somehow glue them together and then finally cut that into shade.
Mark Collier:
Okay.
Dr. Jack Hu:
So that’s additive manufacturing, it’s called 3D printing.
Mark Collier:
Okay. All right.
Dr. Jack Hu:
So basically this is a process you can make 3D printing very smart.
Mark Collier:
Yes.
Dr. Jack Hu:
Because it has all the essential elements, computer aided-design, programming, you also have a lot of signals that you can collect. So 3D printing is a process. It’s not smart manufacturing per se, but you can make 3D printing smart.
Mark Collier:
Okay. That makes sense. So now there may be a little bit of overlap, although they’re technically different, two different.
Dr. Jack Hu:
Technically they cover different aspects-
Mark Collier:
Okay.
Dr. Jack Hu:
– Of manufacturing.
Mark Collier:
Fantastic.
Mark Collier:
So if you can please share with our ASBN audience, some of the most impactful, smart manufacturing projects that you’ve worked on, the ones that you enjoy the most or deliver the most impact.
Dr. Jack Hu:
So I can group some of my work into two types, one is around larger scale systems. I can give you an example that is automotive assembly. I work at the University of Michigan for a long time and I have close collaborations with General Motors and Chrysler and so on. So just think and imagine when you put a car body together.
Mark Collier:
Yes.
Dr. Jack Hu:
That’s basically sheet metal from small pieces, assembling to sub assembly, finally, you have that entire frame so let’s think about that measurement scenario. When you put the parts together, we can put computer vision sensors where they check the dimensions or the dimensional accuracy of the bodies that were assembled.
Mark Collier:
Okay.
Dr. Jack Hu:
Usually there are a hundred to 200 critical dimensions on a car body. Why those dimensions are important, right? So you need to put a windshield glass on.
Mark Collier:
That’s right. That’s right.
Dr. Jack Hu:
You need to put the doors on so if the dimensions are having too much variability, then the glass cannot fit too well, the door cannot fit too well.
Mark Collier:
Got it.
Dr. Jack Hu:
So the result of such bad fit it would be wind noise when you drive the car on highways or crack windshield because of stress.
Mark Collier:
Right.
Dr. Jack Hu:
So dimensional accuracy of car bodies are very, very important.
Mark Collier:
Okay.
Dr. Jack Hu:
But let’s say we have a two shift production so 500 cars for eight hour shift, so two shift production will yield a thousand cars a day. Right?
Mark Collier:
Okay.
Dr. Jack Hu:
So that’s a thousand car bodies over a two shift production and then 200 data points per car. So that’s 200,000 pieces of data each day.
Mark Collier:
That’s a lot of data going into–
Dr. Jack Hu:
That’s a lot of data. Right? So then how do you convert such volume of data into useful information?
Mark Collier:
Right.
Dr. Jack Hu:
When I started this work, I would say AI and big data actually were not as fashionable as today.
Mark Collier:
Yeah.
Dr. Jack Hu:
So we use traditional statistical method to try to extract relationships among the measurements.
Mark Collier:
Right.
Dr. Jack Hu:
So if you think the different piece are assembled together, and then if you have a good understanding of the processes of assembly, then you can look for relationships among the various measurement points.
Mark Collier:
Makes sense.
Dr. Jack Hu:
Right? So due to their connection for being on the same part or at the same location in the assembly and so on. So by doing that systematically due to the statistical relationship among measurement, and then correlate that to the process of fixing and welding, and the part holding and then you can identify where the root causes of variation were introduced.
Mark Collier:
Okay.
Dr. Jack Hu:
And then you can remove such root causes by fixing the process, fixed, welding, or location, then you reduce the variability. So we have a lot of success with this particular application. So this was down in the middle nineties.
Mark Collier:
Okay.
Dr. Jack Hu:
But I think today with machine learning, I think the method can be much more efficient.
Mark Collier:
Oh, absolutely. I mean, I think that the advances in technology today would make a great use of the initial work that you did or that early work and to drive even greater efficiencies for the manufacturers in automobiles.
Dr. Jack Hu:
That’s correct. The efficiency for the data analysis, information extraction.
Mark Collier:
All right.
Dr. Jack Hu:
I can give you a second example if you have time.
Mark Collier:
Please.
Dr. Jack Hu:
So I did quite some smart manufacturing in welding and joining.
Dr. Jack Hu:
Okay.
Mark Collier:
Including our quality, resistance welding, and also ultrasonic welding for Lithium ion batteries. Okay. So when you have battery cells that are welded together for electrical vehicles, first application was around ultrasonic welding.
Dr. Jack Hu:
Okay.
Mark Collier:
Later on, I think laser welding. So with each and every welding method, you really cannot do destructive testing.
Dr. Jack Hu:
Right.
Mark Collier:
In terms of evaluating weld quality. Right? But once you separate the well joint, then the cells will not work together anymore. That’s why it’s quite destructive. So we have to use nondestructive method to evaluate the weld quality. So what I had done here is to use the process signatures, the signals from the welding machine, as well as the song and vibration from the weld process to identify whether weld is good or not good. So that involve signal processing feature extraction, and then machine learning in classifying the quality of the wealth so that’s another example of smart manufacturing.
Mark Collier:
Very good. Very good. So let’s talk about small businesses for a moment and the wealth of data and information that’s out there now, what initial steps should small businesses take to begin implementing or practicing smart manufacturing in their current manufacturing processes?
Dr. Jack Hu:
So, Mark, as you said earlier, you can’t really improve a process that you cannot measure. So I would say for small businesses to start implementing smart manufacturing, the key is to understand the fundamentals. You know, what are the key metrics that you want to measure in the machine or in the process. Define those metrics and then measure the processes so the regular data collection, whether that is through sampling or continuous measurement, so we have to have sensors.
Mark Collier:
So that sets your baseline more or less. Okay.
Dr. Jack Hu:
Correct. So define what you want to measure, what you want to improve, and then measure those and then from the data, try to identify what is going on. Is the process stable is the process out of control? So the traditional ideas of statistical control can come into play and then from there you learn about the behaviors of your machine and system and then you can improve. So I think, even though today’s technologies, data science, big data, artificial intelligence, are much more sophisticated, but the principle are still the same. You need to marry your processes, understand the data, understand what the data is telling you, and then improve the processes.
Mark Collier:
All right, that makes sense for first initial steps for small businesses. So now lets kind of take a look forward, kind of share with me some of your insights on potential smart manufacturing initiatives in the future, as we progress through our digital and technological revolution.
Dr. Jack Hu:
So machines are become smarter and also modeling capabilities are much better today. One particular trend in smart manufacturing is called the digital and physical trends. So T.W.I.N., so you have a physical system, which is your manufacturing machine, or other machinery, and then you can create a virtual, a digital version of that same machine. So the two fit into each other. So the sensors that you have on the physical machine will collect data and the data will fit into the virtual machine and then the virtual machine, which is a model, very sophisticated model, you can diagnose with causes, you can fine tune control, and then you will feed back to the physical machine, adjust the performance.
Dr. Jack Hu:
So really two very similar models, the physical model, which is the real machinery, and then the virtual model, which is a computer creation, the data that you collect from the physical machine got fed into the virtual machine and then the adjustment and the fine tuning and the feedback control from the virtual machine, you can feed into the physical machine so the two connect together, so that’s the digital trends of the physical machine.
Mark Collier:
And those are future trends for small businesses.
Dr. Jack Hu:
That is very much so, a lot of interest for manufacturers.
Mark Collier:
All right. Dr. Jack Hu, Senior Vice President for Academic Affairs and Provost at the University of Georgia. I want to thank you for taking time out of your very busy day to come in and discuss smart manufacturing, which I feel is a very important development for small businesses and one that they need to look at very closely because all businesses have the same goal and that’s to enhance the bottom line and smart manufacturing can certainly put them on that road to do that.
Dr. Jack Hu:
Well summarized, Mark. It’s a pleasure to see you and thank you for inviting me.
Mark Collier:
All right, very good. Have a great day, Dr. Hu.
The Atlanta Small Business Network, from start-up to success, we are your go-to resource for small business news, expert advice, information, and event coverage.
While you’re here, don’t forget to subscribe to our email newsletter for all the latest business news know-how from Atlanta Small Business Network.
