[OT] ideas for detecting cracks in welds?

[Ben]

I have been asked to detect some cracks in welds as they come out of the welder.

So what kind of approach would you suggest I chase down?

Physical contact with the widget is probaly not an option.

I am just attempting to do some brainstorming before I take on this challenge.

All ideas are welcome!

Ben

[TiTou]

Hi Ben,

Are you thinking of doing it with vision ? Do you have any samples of "common" default you have to detect ?

In my company we do a lot of vision application and sometimes we first make a small system that will run for a few days just to take a lot of pictures and during this we ask the customer to alternate "good production" and "typical defects" so that we can build up a script.

NB : In this type of application the choice of the lightening system is very important 😉

Hope this can help you in your brainstorming

[smercurio_fc]

Wow, this takes me back to my first co-op job with the Philadelphia Electric Company. I remember one of the engineers there working on something like this.

I suspect the approach will be highly dependent on the type of material and the type of weld, as well as the kind of crack you're looking for. I've heard of the eddy current method being used. Perhaps the following might help:

http://www.gepower.com/prod_serv/serv/pipeline/en/insp_srvcs/crack_detection/index.htm
http://www.tfhrc.gov/hnr20/pubs/0018.pdf#search=%22detecting%20cracks%20in%20welds%22

[Greycat]

Hey Ben:

I've seen some sonic sensors do wonders .. .they can let you see directly into a metal and detect many things ... just a thought - sonics might be the way to go but you usually have to either touch the material or usually use pure water or other such sound wave conducting material between sensor and unit under test.  Just a thought.  Here are a couple of links.  The first one is the one I think you would be more interested in.  I haven't done much work with these so that would be about the extent of help I could give you.

Here and Here

Greycat

[Dennis_Knutson]

Years ago I did some work for an oil pipeline company. Ultrasonics were used to analyze every weld. Pretty expensive technology back then and it still might be for all I know.

[Ben]

Excellent suggestions!

One of the engineers I work with is following up on the links you have posted.

When it comes to brainstorming, all ideas are good ones.

Feel free to add to the above if you have other suggestions.

Thank you,

Ben

[Conseils]

In one application I worked on, it was necessary that the weld was air tight. So we tested the system with pressure if it didn't leak we considered it good enough.

The suggestion here is that you might look for a secondary effect or required property. A bit like testing a bell to see if it's cracked by exciting it. i.e it might be cheaper to test actual functionality as the equipment may already exist.

I guess you may already have though of this, but you did say any suggestions.

[mischi_miyagi]

If you have to detect very small defects...x-ray comes to my mind

Michi

[Ben]

"...but you did say any suggestions. "

Yes I did!

Your mention of the bell is great!

If I beat on the widget I could listen for funky sounds!

Did you ever wonder why steam railraod locamotive engineers used to carry a hammer in the loop of their cover-all?

They used to beat on the boiler to check for bad rivets.

"There is nothing new under the sun." (Eclesiastes ? (sp?))

Ben

[gwd]

There is a very extensive knowledge base on this and I suggest you look into the techniques first and start thinking about Labview much later.

A good place to start is http://www.ndt.org/.