When 1 + 1 = 3: The Power of Fusing Data
If “data fusion” sounds like a complicated mathematical concept, you’ll probably be surprised to learn that you do it every day. For example, when you eat a great steak, your brain fuses the outputs of several sensors: taste, smell, texture, shape, and color. But not all data fusion is this delicious.
In manufacturing, data fusion is not as instinctively used. Optical inspection plays an important role in quality control, as do X-ray inspection systems. But for too long these two modes to inspection have operated independently. At Creative Electron, we are combining optical inspection with X-ray analysis for the ultimate in quality inspection. If you are reminded of the old peanut butter cup ads, “two great tastes that taste great together,” we are right there with you.
That’s when 1 + 1 is more than 2. It’s better than that.
Let’s see how data fusion works when inspecting electronic components. One of our customers was looking at ATMEL922 parts. As you can see in the following figure, having both optical and X-ray images allow us to analyze this component in 2 dimensions. Much like taste and smell, they go together well.

In this example, we have in the vertical axis the results of the X-ray inspection, and in the horizontal axis the results of the optical inspection. Since we’re looking for counterfeit components, the fakes are easy to find when they fail both optical and X-ray inspection. You’re right, in these cases, just one of the modalities was needed. Same for the devices that pass both X-ray and optical. However, what should we do when the modalities disagree – pass in one and fail in the other?
To tell real and fake apart we can adopt a manual approach and have an operator look at the parts where the inspection modalities disagree. But we are Creative Electron, and we love automation.
Automating this process depends on the sample we’re looking at. In this example we designed an artificial intelligence (AI) engine that was brilliant at figuring out if the suspects were real or fake.
How is it all done so each modality (optical and X-ray) ends up being better together than individually? This is the point where we need to start talking about training AI engines, ROC curves, and Bayesian equations. If that’s of interest to you, we have the right people for you to talk to. I mean, they’re not totally right, but they know this stuff.
Combining the strengths of optical inspection and X-ray analysis provides outstanding quality results. Fusing data from both using automation and artificial intelligence yields the most thorough validation, preventing defects from passing because of what one of the two technologies might miss. For high reliability products such as medical devices, automotive, and aerospace applications, the combination of machine vision and X-ray inspection offers the most thorough and accurate defect detection available. If you have one of these applications, reach out, we may be able to help.