Frankly, I don't know exactly; I can only give you some suggestions

• Use PAW with shielding gas (NOT plasma Gas)with temperature and velocity so that the weld pool is solidified before contact with water. If PAW plasma nozzle is not jammed, try some volatile silicohalides mixed with carrier gas that would decompose to a glassy silica layer before coating. Find out oxides most effective against hydrogen diffusion, and add some volatile species to PAW carrier
• use SAW coated with hydrophobic materials that can retain hydrophobicity up to high temperature. Water-contact curing cement-like material maybe included in SAW flux, but organic ones might introduce too much C into the flux.

One of the methods with great potential for applications in underwater repairs is local cavity welding. In local cavity method, cooling conditions and diffusible hydrogen amount in weld metal are nearly the same as those existed during welding in the air. This paper presents the results of literature survey and preliminary tests of the effect of local cavity welding conditions on diffusible hydrogen amount in a deposited metal. A Plackett–Burman design was applied to screen seven parameters: thickness of elastic cover, stick-out of electrode wire, welding current, voltage, travel speed of welding, salinity of water, and flow rate of shielding gas.