Google's AI Vision Rethink of Package Testing for Shock & Vibes: Packaging World

How are new tools, in conjunction with AI, helping engineers at Google?

Packaging World Editor-in-Chief Matt Reynolds sits down with Ken Leung, 
Senior Mechanical Engineer at Google, to discuss high-speed imaging and AI-driven analysis for packaged data center hardware. Leung explains why traditional testing methods may be missing the point because of changes to the products they're meant to protect.

Read the full featured article on Packaging World.

Transcript

SPEAKER_00 0:03

Welcome to Through the Line, a podcast exploring innovations and information across the packaging and processing landscape. From topics impacting consumer packaged goods and healthcare packaging to the latest technologies and food processing operations.

SPEAKER_02 0:22

Hi, I'm Matt Reynolds, editor of Packaging World magazine, back with another edition of the Through the Line podcast. Have a listen as I speak with Google's Ken Loom from the International Safe Transit Association TransPEC Forum, where we explored the critical role of shock and vibration testing in protecting today's increasingly complex hardware. Whether that's data center racks to consumer electronics that you and I buy on Amazon, Ken shares how traditional testing methods may be falling short, and how new approaches using advanced sensors, high-speed imaging, computer vision, and open source tools are now reshaping how engineers understand what's really happening inside a package during transit. As you'll see, physics, data, next gen testing methods, and course AI all play a central role in discussion. Welcome, Ken. Why don't you introduce yourself?

SPEAKER_01 1:06

Yeah, my name is Ken. I've been doing socket vibration testing at Google for 15 years now. And that means preparing all the Google's data center hardware, the racks, the servers, the line cars, the switches, all ready for the transportation and handling around the world.

SPEAKER_02 1:22

Because they're not building these racks at these data center sites, right? So they build them at one place and then you've got a complete rack that can basically be plugged in.

SPEAKER_01 1:31

The supply chain is completely complicated. It's really complicated. You may have servers that are built in Asia, you may have racks that are built in America, and then they have to integrate together somewhere else in the world, wherever it is, maybe closer to where the data center is going to be. Integrated there as a fully populated rack full of machines. And this goes on an air ride truck or your normal typical trucks, and it goes to a data center, and you want to be able to wheel this off the truck, go directly to inside and plug it in and turn it on. Okay.

SPEAKER_02 2:05

That's the goal. Okay. So I think it stands to reason, but why is vibration and movement? You see, yeah, it's vibration and what was your specialty? Vibration and shock and vibration. Shock and vibration. Shock and awe, shock and vibration. Why is it such a big issue for today's hardware? And this isn't just the biggest and heaviest pieces of equipment. This is down to you could imagine this being uh an issue for, let's say, cell phones or or computer equipment that's on the consumer side. Why is uh shock and vibration such a big issue?

SPEAKER_01 2:34

Anytime you have a physical hardware where you're building it somewhere else where the place is going to be used at, you're gonna be introducing these movements to your product. Depending on how you package it and depending on what kind of truck you use, that vibration can go inside the product and cause damages that you might even never imagine. So fractures down to the material layer and start cracking. And once things start cracking, it's gonna fall apart and you don't wanna you don't want damaged product getting to your customers. You don't want damaged packaging actually to get to your customers. So they're important in both sides.

SPEAKER_02 3:13

Okay. So this would translate to the consumer side to anything that's got a lot of componentry in it. Anything that's let's say a solid state is probably uh less uh susceptible to this. But when you've got components that are joined together, soldering, soldered, for instance, together or wired together, all of this vibration, any one of these uh points where the two pieces interact can be a problem. But it's also the physical solid structures too, that bend, break, tension, and compression are issues. So these are all things that you're trying to solve for. Good. You said that failures can show up later, and it's almost better when you roll this uh large uh rack off the truck and plug it in that it doesn't work. Explain that because uh you want it to work. But why is it more important more important to know right away?

SPEAKER_01 3:54

Because when you turn it on and it doesn't work, it's very easy to diagnose that. You have immediately you know maybe it's the power supply, or there's somewhere in the circuit box. You have a lot of evidence. And the most important evidence is that it doesn't work. That's the first thing. What happens is when these machines go to data center when you have some partial damage or partial connection of materials, over time it's going to wear. The data center hot environment, it's cool, but the machine gets hot when it's used, and then it colds when it's not used. So it's flattening out and then expanding, whatever it is, right? It's really complex. So over time, these are supposed to be rated for five years life, ten years life. Suddenly it's three years' life, right? And that's three years after your product development cycle mode, nobody cares anymore, nobody knows how to look at it. Machine comes back and you say, What happened? How do we fix it? So this becomes a really drawn-out process that costs companies millions of dollars. Sure. Because the failure can be complex, could be any one of those components, costs a lot of money to take it all apart to figure out what's going on. So you just lost millions of dollars.

SPEAKER_02 5:03

Sure. And then if it were to go down during let's your example was let's say Black Friday or something like that, then the company is it's basically it's equivalent to unplanned downtime, but the capital so much goes into it to have it fail at the two-yard line, so to speak, is uh important. So as a packaging engineer, what is something that can we can more broadly discuss for all sorts of different packaging engineers, even on the CPG side, that packaging engineers get wrong about shock and vibration, different ways you think about it.

SPEAKER_01 5:32

Yeah, the traditional methods, like solarometer and shaker tables, is pretty much been unchanged for the past 50 years. Same test profile, still different weights and different flavors of it. It's still a job, which is impact, and a shaker table, which is the vibration. Most people don't really know physically what's happening to the real product. And so they're just making guesses. Most of the time they say put more foam in it or put more materials in it. And actually, sometimes that's make it worse. Because, yeah, because inside for my world anyway, you have a server, you got a giant heat thing, it's like a bubble head, go left and right. You try to put the soft foam around it, it makes the wobbling even worse. And it actually damages things more. So put more foam in it made it worse. But that's not something that's immediately intuitive to people. So we want to spend more time understanding the physics of what actually happens inside the product. So you're not just looking at the box from the outside anymore. You do a test of a cardboard box, you're not going to get any information from it. And this is why we spent 50 years of testing on all these things that were still starting on scratch. So the most important part is to open it up and see what's inside. And that's what I spent a lot of my time on.

SPEAKER_02 6:43

Diagnosing the actual problem as opposed to just trying to put band-aids around it enough so that it's not bleeding. I understand. Are current testing methods missing what really happens down to like the actual stresses and the fractures and stuff like that? Because you're just applying this band-aid as opposed to using an X-ray to see what's inside and using sensors and all the new equipment that we have available. Is that fair to say?

SPEAKER_01 7:04

Yeah, it's uh it's not that easy to see inside, right? A lot of it is steep inside. So in the semiconductor world, what they would do, something called the end price, to soak into the chemical, the chemical, it's red, it's seep inside the cracks. You pull it open, you can see how far a chemical goes tube, and you can see how far the extent of damage. But like I said, that's hard to do for everything. You can cut it open, do a cross section and look at it in a microscope. Same thing, right? It's deep inside it. Okay. So we're trying to figure out how to deal with that. And people don't really have a solution, especially when they don't know how the things move and behave. Okay.

SPEAKER_02 7:41

So that kind of brings us to your project. Why don't you uh tell our audience again, our audience, brand owners, CPGs, what your project is as an open source or will be an open source that you're individually building over time to uh to create tools and methodologies and standards, these sorts of things. Standards might not be the right word, but a way of thinking about the product protection itself in packaging as opposed to just putting more dunnage on it or something like that.

SPEAKER_01 8:02

We're basically open sourcing that research and the tools that we use to investigate other stuff, even well before the standards are developed. Right? The real world is complicated. So one standard doesn't mean you can cover the same vehicle vibration in Asia versus US, North America versus Europe, and other stuff. So you really need that science behind it and other tools to support that work. So a couple of years ago, I just said, okay, the world is getting complex, everything is crazy, there's all these different designs and different structures and all this stuff. What can I do to make it a little bit better? And I just started writing up everything that I don't know about all this, just fundamentally how I think about it, what kind of sensors I use to measure, the brand, the price, how you use it. Um, and then starting to go into from traditional sensors toward high-speed cameras and using computer visions and 3D measurements to really change the way people look at things. Because without these tools, you can't support the development and scientific endeavor to make these things better. Packaging hardware products or whatever make the supply chain better and all that stuff. My thought is to get the tools to the people who need it as quickly as possible, and then to make it as cheap and easily available around the world as it can so that everybody can use it. Students, researchers, packaging engineers, hardware developer, whatever it is.

SPEAKER_02 9:30

Okay. Good. Now, you're dealing with a very specific set of products that have extremely tight tolerances, let's say, uh compared to uh to that might not exist uh amongst the wider uh consumer package goods, CPG world. But I I have to think that this kind of science, this kind of knowledge and body of knowledge is going to translate in some way. So why don't you expand on that? What are ways that this extremely, and you're talking about terabytes of data for two or three seconds of testing, or I don't know what the translation is, but that's going to impact downstream in all of packaging. So how do you think that translates?

SPEAKER_01 10:05

Materials and stress is universal, and that's why mechanical engineering is such a broad field, is that it's universal for all platforms and all kinds of products. So we're really just looking at how material behave under stress. So whether it's something complicated like a computer server or nuts and bolts for consumer products or whatever it is, it's all going to translate to it in the way that it's the environment creating stress and the durability of structures against it. You want more strength than the stress you can get. And when you're really far above it, then you're not that worried, right? Nuts and bolts, for the most part, are not going to break apart unless you really put thousands of newtons of loads on it. But it's that pipeline of how you look at it and measure and understand the materials to get to and look at your risks and see how far you're above, that's the really important one. It may not matter. And if you have data for that, everyone is going to benefit from that and know how to do the research and testing anymore. They can just say, oh, there's data that says it was good. So we can move on and do other stuff, right? Because there's so much up there. But we don't have that pipeline right now because the accelerometers, you get pages and pages of data and you don't really know what you're looking at. Where it was really matters, right? Is it on the top? Is it on the side? So computer vision and the 3D measurements of it's a video that you can see right away what it looks like, what was actually done. Was it a drop? Was it a vibe? And how did it break? Or if it's boring and nothing happened, then that's even better. You have this record of it. Anybody with overall degree can look at it and almost agree what happened. And then you have your lesson learned and you move on from that.

SPEAKER_02 11:50

You build on top of it. Body of knowledge. You some of your slides today had were uh basic compression intention strength calculations from a 1960s textbook. Once you build the body of knowledge, then the library is there for everybody, and that's what you're working on. So why don't you explain before we'll have one last advice to packaging engineers question? But before I forget, please, if there are any contact information about your project or anything that you'd like to invite our packaging engineers from across the CPG and space to learn more about this project. Yeah. Give us the details. Yeah.

SPEAKER_01 12:20

Go on Google, just imagine Google open source vibration project and GitHub, and because that's where all the papers and the data are released. That almost come up really quickly. And then on LinkedIn, we try to connect to people. I'm not sure I can give out my Google email address to people, because but that's why in the slides you have that OCP address that will eventually get to me. But really look me up on LinkedIn.

SPEAKER_02 12:45

Okay. LinkedIn. I found you on LinkedIn. If I can do it, anybody can do it. All right. Before we go, the audience today was mostly packaging engineers. Everybody was listening in wrapped attention. So clearly there's a translation here between Rodney and SC Johnson and what you're doing. So if you had any advice or what the next 10 years looks like for packaging engineering, how do you see that? What kind of words of wisdom would you impart to packaging engineers in our audience?

SPEAKER_01 13:08

Yeah, so physics is universal, the ultimate limiter to all these stuff, even legal MSS. Physics is the only rules that you can really are bounded by. So understanding that you first of all learn your craft, learn the signs, you go to school and all that. They're really important. Because very quickly you're gonna throw all that out and you got the real world and you got projects stuff. Spend more time understanding the physical part of it. And on the side, we're going to give you all the tools and what you need to help you to do this, right? Because nobody can do this by themselves. There's no silos. These things are incarnate complex, electronics, medical devices, food packaging. Everything is going incredibly complex. So we're learning to work together, connect, and how to come together to build on top of while we're here to go. I think that's what we need to do. Okay.

SPEAKER_02 14:00

I think that really puts a bow in it. I really appreciate your time today. I'm sure our audience appreciates hearing from you. So thanks again to Ken from Google.

SPEAKER_00 14:08

Thank you for listening to Through the Line Packaging and Processing. You can listen to more episodes on all streaming platforms. Be sure to visit us at packworld.com, profoodworld.com, and healthcarepackaging.com for more packaging and processing news. This podcast was edited by Bree Guns.