Computed tomography (CT) has a parallel life beyond its service in diagnostic medicine. From aerospace and automotive to defense and electronics, it’s one of manufacturing's most powerful quality assurance technologies.

By creating detailed three-dimensional views of both external and internal part features without destructive testing, CT allows manufacturers to detect defects, verify dimensions, and improve product performance. Now, artificial intelligence and machine learning are accelerating those capabilities even further, making inspections faster, more accurate, and increasingly practical for production environments.

In this episode of the Great Question Podcast, Hexagon Manufacturing Intelligence Technical Sales Engineer Roger Wende details how AI-enhanced CT scanning is reshaping non-destructive testing, metrology, defect analysis, and the future of digital manufacturing.

Below is an excerpt from the podcast:

Robert Brooks: Hello, and welcome to a new installment of the Great Question Podcast, brought to you by Endeavor Business Media's Manufacturing Group. I'm Robert Brooks with American Machinist and Foundry Management Technology. And both of those markets are keenly attuned to the salience of quality assurance and quality control in their operations and in their finished products. 

One of the most advanced and reliable quality assurance tools, or more accurately, quality assurance technologies, is computed tomography (CT). And like every other aspect of manufacturing, quality assurance is significantly influenced by the expansion of artificial intelligence, digitalization, and machine learning. And I am no expert on computed tomography. My guest today is that expert. Roger Wende is technical sales engineer for non-destructive testing with Hexagon Manufacturing Intelligence. Welcome, Mr. Wende. Thank you for being with us today.

Roger Wende: Thank you for inviting me.

RB: Now, many of our listeners will be familiar with CT because it's a widely applied medical diagnostic tool, but it has a significant record in manufacturing. And I know you will bring all of this into clear focus for me and for them. So will you begin by sketching out the basics of this technology, computing tomography?

RW: Yes, I'm happy to. So CT or computer tomography, as you just stated, really fascinating technology that really started in the 1800s here in the UK. What we saw over the years and kind of understanding what CT stands for, computer tomography is actually coming from the Greek words tomo, meaning slice, and then tomography is to draw or to write. So essentially what you have is drawing the slice. 

And what you're doing with the CT scans, you normally have an X-ray tube, you have the object you want to scan, and then you also have a digital detector. With the X-ray tube, when this was originally discovered, you'd actually have these electrons that would shoot down in a vacuum, hit a piece of metal. When that collision occurred, photon X-rays would radiate out through the object. Now, in order to capture this information, as these photons fly through an object, which usually has a multi-material, different absorption rates of these photons, they'll be collected in today a digital detector. 

Now, these digital detectors, when these photons go in, go into something called a scintillator. It's essentially these crystals that turn the photons into visible light. And this visible light gives us a gray value information about the part. So what's really fascinating when you see an X-ray, as you stated earlier with the doctor's office, you'll see the different gray value information, and that's the absorption that you've got going through the material, and then that visible light, that different gray value information that we'll see in our display. 

With a CT scan, instead of doing a traditional X-ray, you're actually rotating the part normally. In the human case, you're using standing still in the detector and X-ray source is rotating around you. But you're essentially taking pictures, lots and lots of pictures at different angles. And then you go through what's known as a reconstruction. This is a Feldkamp CT reconstruction method. And when you take all those sliced images, you'll get this volumetric data set. So now I'm able to slice through this information that we've collected and just actually do a lot of different types of engineering applications. I know most people are used to it in the medical side. For us, we're actually really focusing on the industrial side, working a lot of different engineering applications.

RB: That's an excellent explanation. That's right up my alley. I love that, I love this. So tell us then, what is the customary role for CT in manufacturing operations?

RW: So it's fascinating to see because when you look at classical NDT or non-destructive testing, you always have the core package. It's like a toolbox. I got all these tools to do a lot of different engineering tasks. So I have ultrasonic testing, visual testing, radiography testing. 

In this case, the CT is essentially a new tool in the box. It's really come about in the last 25, 30 years. That's how new it is. But it's an exciting tool, and I'm happy to say that I was really part of it from the ground up. And we saw the roots of it heaviest in Europe, in Germany, because of the automotive industry. And the big companies would actually buy this expensive equipment, CT scanning parts, to see the failures that you may have. As time's gone through and time's gone by, because one of the unique devices collects all my surface information, as well as my material information. Over the years, what we're able to do is capture that data to allow us to go beyond quality control, but actually look at it from a metrology standpoint, when you do critical measurements, complete with GD&T. Maybe I need to do a material analysis because I am interested in finding voids, cracks, inclusions. 

The other thing that's fascinating is working with composites and understanding the orientation of the fibers. because it's these fibers that give the parts their strength. And these fibers can consist of plastic parts with fibers, the actual composites you see, or even concrete that's got steel fibers on them, that can give the structural strength that you need. The other areas that sometimes people will actually do with CT scanning is reverse engineering, working with CAD files, assisting in additive manufacturing, helping with the design team, so now we're not just looking at metrology and quality engineering, But also we're working with the CAD designers. 

And not to leave out the finite element analysis team, the ones that do simulation, because unlike a FEA model is what they call it, the finite element analysis, when I have this perfectly CAD drawn, I want to do physical testing on it today. Maybe I'm doing what's known as computational fluid dynamics or Von Mis stresses. I can actually capture the actual part with its known defects to see how the part can actually perform under those different strains.

RB: Wow, you're all across the manufacturing sector. I mean, everything. How does CT scanning fit into a manufacturing workflow?

RW: So in originally it was used for quality control. That's literally how it's first adopted. I'm interested in looking for cracks. I'm looking for porosity. Then what was interesting is in the plastic injected molding business, really in the automotive side, it's like, can we use a CT to actually convert the CT scan data in a metrology world, similar to a coordinate measurement machine or some of these laser scanners? So a lot of development, understanding about the CMM world, essentially, these coordinate measurement machines. So that's the golden standard, see how I can take some of these standards and understand it and apply it to the CT scans. 

So the evolution really sped up in looking at it from a metrology standpoint, doing first article inspection work. So that was a really big area. But what was nice is I look at for defects too. Then we evolved a little bit further and said, hey, we have all this information. Maybe I do need to look at with other design teams. But normally in the very beginning, it was really this workflow of working at the defects. looking for missing components or damaged components. But then the metrology really kind of took over and we saw with certain products, especially with injected plastic molding business, it's actually faster to actually use a CT scanner than some of the traditional CMM machines or laser scanners.

RB: So I mean, the obvious off ramp for me here is to discuss all of this information, all of this data gathering. How does artificial intelligence changing the CT inspection process or the approach to CT inspection?

RW: So the whole idea of AI is fairly new in the CT world. When you take a look at all this gray value information, and this gray value information that you see, when you look at your part, you'll see these, it's actually a part's made of tiny little gray cubes known as a voxel. And these voxels come from 2 words, come from meaning from the volume, the VO from volume and the last few layers pixel. So you get the word voxel that has this gray value information. 

So taking that voxel information and with this gray value information, you can always see that gray value information, use software to actually work with the different gray value information. But with CT scans, you also have something known as artifacts. and they can be different types of artifacts. Some call it ring artifacts, streaking, beam hardening. These are unique smearing of the gray value information that can make it challenging when you're looking at something as defects. Or if I'm doing battery analysis and I have to do a really fast scan and I want to be able to look at my anodes and cathodes and make sure that the angles are correct or I got the correct amount, there's not any particles. When you have a very noisy data set, and you're just working with traditional gray value information, you can actually have a lot of false accepts and false rejects. 

Now this is where AI comes interesting because you can actually collect data sets and create trained data sets. Because what we're working on now is not just gray value information anymore, but what AI does, like our eyes, will help us define shapes. So we're literally teaching using AI, looking at not just gray value information, but shape. And that's where it becomes a game changer because it really lowers our false accepts and false rejects. 

And the second thing you can actually do is speed, performance. So if you have very large CT scans and maybe 100 gigabytes, that's a large data set. The time to do traditional defect analysis may take quite a bit of time, but using this trained model, you can actually do that analysis much, much faster. So now we're talking about doing inline production work, and that's what we're seeing today is CT evolving with AI. when it comes to automation on the line or doing batch sampling. And that's becoming very, very popular.