AI based visual inspection in aerospace manufacturing

[dutchguy]

Dear fellow engineers,

I am looking for information about the use of AI based visual inspection for aerospace manufacturing.

Can anyone point me in the right direction to find information about this or find someone who can ?

Much appreciated and thanks in advance.

 

[rb1957]

I'll rephrase your question so that it means more to me ... "I'm looking for information about how AI is used in the development of visual inspections for airplane maintenance (ICAs = Instructions for Continued Airworthiness)."

In my experience, it isn't.

If you meant something different, please restate.

 

[dutchguy]

Although mine last name is Paulus, this time I meant something different:

Is there someone who can help me find a way to use AI in analyzing data gathered during visual inspection of manufactured aerospace parts by using camera's ie. ? Furthermore, the requirements refer to scratches, dents, irregularities in paint quality, etc. Also, it would be convenient if AI would give me information where those defects are located on the parts by making a plot and the bring defect area would then be colored.

I hope this helps in understanding.

 

[mfgenggear]

OP
I am sure there is a company out there can write a program to accept such criteria. It is not free.
This will be very specialized

 

[rb1957]

OK, so you want "AI" to interpret the basic inspection findings (is this scratch significant or not ?) or possibly to present the findings of an inspection (look ! there's a scratch here !!) ? Neither seem to be applicable to AI ... the OEM defines damage limits on different structures, and how to repair them.

Or do you want to automate the inspection ??

 

[dutchguy]

As AI cannot make decisions on its own, it will have to advise the inspector in finding the errors on the parts/components enabling him to make a decision if the part/component has a non conformity or not and where it is located.

So yes, I want to make the inspection partly automated, but also to help the inspectors. I think that AI could decrease inspection time.

 

[IRstuff]

I don't think AI will "decrease inspection time" per se; one thing that AI undeniably shines at is perserverance, i.e., it won't get tired, it won't get bored, and it won't miss anything at hour 24 that it wouldn't have missed at hour 1.

A well-trained AI will likely more consistent catch things that a human observer would have missed, so if anything, the inspection and rework process will slow down, because more defects will be caught and more defects will require adjudication by a human and more defects will require rework or rejection.

You may find that management will likely ditch the AI inspection idea after a while, because they won't want to make the costly changes needed to minimize the number of defects found by the AI.

 

[SWComposites]

because they won't want to make the costly changes needed to minimize the number of defects found by the AI.

No, they will just beat on Engineering to open up the acceptance criteria.

 

[WKTaylor]

SAE is leading this charge... lots of related info. See SAE website.

 

[mfgenggear]

Well here it is.
The question is can AI do a better job than a human in this case.

 

[Sparweb]

"can AI do a better job than a human in this case?"
<question mark inserted in place of the period to give some sign that a question is being asked>

The answer is "yes" and "no".
Yes of course any form of automation can accomplish well-specified tasks for which the algorithm can be programmed to respond.
No of course a human can recognize and investigate in ways that are very difficult to define or pre-program in advance.

At some point aircraft inspection will take the leap that medicine has done with the analysis of chest x-rays and other diagnostics. An algorithm can be used to speed up and make more accurate the interpretation and diagnosis, but a human still needs to be in the loop to double-check it and validate the result. And while some people convince themselves that the human delivering the computer's result can get dumber and dumber as the technology advances, I believe that the human that accepts and takes action on the diagnosis should always be a professional with the ability to understand and interpret the result themselves. It will probably be the same with aviation if these types of inspections are adopted.

In aircraft, like medicine, the part that computers cannot (and never IMO) do is take responsibility for the inspection. There needs to be a responsible party at the point of validation of any test result, good or bad. Only humans can do that.

 

[dutchguy]

I still think that AI can help us making better (read better quality) products and at the end lower the manufacturing cost, by decreasing non-conforming products.

AI is a tool in improvement of the inspector's task. AI pointing at the defective areas and the inspector making the decisions whether the part is defective or not.

In the beginning it will cost some more in order to improve manufacturing processes and perform extra rework, but it will pay back at the end. The costumer will be happy, less NC's (here at Fokker an NC cost roughly 3000 euro) and higher quality products.

Moreover, if inspection requirements will be lowered by engineering (which in my opinion is not necessary) that's up to them. But why should we go in that direction, when the goal would be higher quality products?

 

[dutchguy]

In addition to the above, who can help us finding a company, which works with AI based visual inspection?

 

[MintJulep]

Machine vision systems have been inspecting a lot of things for a long time.

Machine vision makes pass or fail judgement based on quantifiable measurements.

Are you perhaps conflating AI and machine vision?

 

[rb1957]

AI may help in analyzing statistics of damage from the field and point to areas of recurrent concern. Which may help Engineering, but not the inspector (he does what he's told ... well sometimes !?

 

[BvC58]

In addition to the above, who can help us finding a company, which works with AI based visual inspection?

My work involves using a proprietary volumetric capture system, one that is image based, the post-processing leveraging AI to automate anomaly detection. While our capture tech is agnostic to subject matter, our focus has been to bring inspections of infrastructure, such as substations and bridges, into the modern era of VDI (virtual desktop infrastructure). Image-based ML (machine learning) involves a manual labeling process, someone who knows what what they're looking at, then annotates with various graphics tools the objects or features of interest, in the case of bridges, early detection of hairline cracks in concrete is a game of find Waldo and begs to be automated, both the feature extraction and classification, e.g. good, poor, okay. In your case, scratches, irregularities in paint quality, and dents, these three features vary in a way that points to different approaches to feature extraction, some more straight forward than others.

For dents you don't need AI, unless there's such a volume that motivates deeper understanding how these relate, in which case an engineer would be tasked with classification to then automate that. As is, a precise scan of a part, the 3d model nested inside the ideal ( not dented) or even the CAD model used to make the part, assigning different colors to one and the other allows the finest deviation to jump out.

For scratches, we use various tools to "separate the wheat from the chaff", modify the image in ways to allow subtle features (cracks in concrete down to .01mm, could be fine scratches) to be more discernible to the human eye for annotating. While ML typically requires huge datasets to train a model, one advantage of our capture system is providing uniform lighting, sensor, lens, etc., hence enabling automation using far smaller datasets.

As for detecting irregularities in paint, that's far more uphill for both a human and an AI to take on, but I submit here's yet another advantage to our approach to capture, one involving specialized lighting. Paint has color and it has a specular component, somewhere between glossy and highly matte. Measuring albedo color and albedo "roughness" is best done in component form, our scanning methodology uses polarized lighting to record separately these aspects of texture to make possible any true understanding of material reflectance properties. If an engineer reviewing co-polarized and cross-polarized datasets can more readily discern paint irregularities, if not maybe discover otherwise latent defects, then the same pattern recognition operating within the engineer's learning and assessment can be leveraged to scale up automating with ML.

Perhaps, we can help.

 

[WKTaylor]

Simple question... 'What'...

... Are You inspecting... and 'where'? The Factory? A Laboratory? On-Aircraft structure or systems? Engine? Electronics, electrics and wiring? Landing gear? Accessories?

... Are You 'looking for'? AND what accuracy and level of confidence do you need?

... OTHER [non-destructive] inspections will have-to be done in coordination with the 'visual' inspection?

... for starters...?

 

[Franp37]

Have you check, cad comparison software? I am new in the field, but some of this "Cameras/Scanners" does have CAD Comparison in they tool set. that can give you a start point.