Here is a theory I have based on what I have read http://www.weldinginspectionsvcs.com/COncaveWeld.htm It may be some help.

Have a nice day
Gerald Austin

If you are melting through to the round backing, in a V-groove weld joint, it seems like you would have a very high depth to width ratio, which promotes cracking. Have you been able to cross-section the weld to measure this? If this is the case, you may want to switch to a U-groove so that the backing would be closer to the root of the weld and not penetrate through so much, or switch to the triangular backing. Also, the diameter of the ceramic seems very large. It is unclear to me exaclty your joint is configured to use such large ceramic backin.

Just a few thoughts... You mentioned a double vee groove. To me that means a 2 sided weld. If that is what you meant, then I don't understand the reason for the ceramic backing, unless there is insufficient root face and too much root opening. If that is the case, I would think you would have better results in leaving a 6mm (minimum)root face with zero or very little root opening. I would bevel the second side to 40 degrees because after backgouging, 30 degrees will usually be too narrow for good fusion unless the groove is opened up more. For that reason, I like to bevel the first side to about 60% of the depth and the second side the remainder less the face dimension. That way backgouging doesn't have to go further than to the halfway of the thickness.

Another thought is that we usually run 675 to 750 amps with that wire size, with the same voltages, and between 35 - 50 cm/min. (I am assuming carbon steel) Your parameters seem low in overall heat input but I'm not sure of your application.

I too am thinking that your depth to width ratio is not correct.
Especially if you are welding a double vee groove with round backing and are melting to the backing. 23mm diameter backing would leave a large void at the other side of the root. If you must use that configuration, you might want to consider using flux held in place with a bar for a backing instead of the ceramic, or at least somehow using flux to exclude air at the root. Combined with adequate root opening and face, you should be able to be good results.

More preheat did not solve the problem for us when we had problems with centerline cracking.

A U groove was also suggested. Those take more time and money to prepare that simple vee grooves but good results can be obtained with them. Time spent on the preparation could prevent spending a greater amount of time on rework.

Hope this helps,

Chet Guilford

http://www.gullco.com/products/katbak/katbak.htm

Are you using the 1G20-RD backing? On the site it says...
"These tiles are for similar uses as the 1G6-RD but for thicker plates up to 15mm thick",
and you stated you have joints up to 100mm thick and the 38mm plate that you tried has centerline cracked.

Maybe you need to try the triangular backing as suggested earlier and adjust your groove angles to allow for the correct width to depth ratio.
Here is what the site had to say about the triangular backing.....

"All 1G33 tiles are suitable for K or X preps where the bevel angles are 45, 60 or 90 degrees. Also suitable for welding with M.I.G./M.A.G. solid wire, metal cored or flux cored wires. They are a substitute for round tiles where the benefit of the 1G33 range of tiles is a "full-face" contact with the joint preparation... reducing the risk of burn-through as may occur with the round tiles where you have a single point contact of the tile with the work piece. 1G33-90 also used when reverse side of a single bevel, single V butt joint where the reverse angle will always be 90°."

John Wright

Some of my thoughts on the issue:
a) I am not sure what polarity you are using, DCEN could provide improved results.

b) You may quick check the material test reports for any excessive sulphur content.

c) Increasing Preheat / Interpass temperature might improve the results which reduces the weld metal cooling rate.

Muthu