On the fatigue failure of the welds, I'm thinking of two causes:
- Excess Martensite in the weld since the dowel has a high carbon surface. The carbon enriches the molten weld, cools quick, martensite forms. Add high cycles and the brittle martensite fails easily.
- Thermal stresses in the weld, since the cold CRS rod is a heat sink.
Both of these would be helped by a pre-heat. Need to have the surrounding metal just after welding to be above the Martensite Start temp to prevent M formation. This is pretty hard to estimate since the molten pool has unknown % carbon. And it will depend on how much of the dowel is melted in the pool. If the pool is ~1050 steel, M-start is about 600F. I'd talk to a welding expert for a rec on preheat temp. 1/2" dia wouldn't take long to get to temp, and that is good since the exposed pin will lose some carbon and hardness if it is preheated too long. (May need to protect the exposed pin during preheat with a high carbon cover). Perhaps a post-weld stress relief as well, though that is probably more cost than you want.
Make sure both parts and the filler rod have lowest possible alloy content (Cr, Mo, Mn...) for lowest hardenability. Oh!-- if you are currently using an alloy steel dowel, changing to a case-hardened carbon steel dowel would result in much less M w/o any preheat. And have nearly the same surface hardness. And save $.
Yet another idea: Make everything w/ low carbon steel, then carburize the welded ass'y in large batches. No martensite in the weld guaranteed. $ per part depends on how big your parts and oven are.
Good, simple, explanation of Martensite in welds:
David Malicky