The fusion zone has no datasheet

A weld between two different metals is not a seam between two known materials. It is a third material made while welding, and no certificate that comes with either parent describes it. The carbon side arrives with its own conditions: the A106 pipe page offers pipe suitable for bending, flanging and similar forming operations and for welding, then presupposes that a welding procedure suitable to the grade and intended service will be used. The A53 building tube page puts it more plainly, that the grades denote certain chemical and mechanical properties which should be noted in selection. Both are written for the case where the far side of the joint is the same family of steel. At a dissimilar joint they cover only part of what is in the pool, because the carbon parent melts into it as well. The filler therefore has to be chosen knowing that dilution arrives from both sides, and nothing in the paperwork covers the mixture that results.

Two properties pull against each other

The mismatch starts with expansion. 304 quotes a thermal expansion coefficient of 17.2 x 10⁻⁶ ⁄K, conductivity of 16.2 W/m.K and a modulus of 193 GPa. Carbon structural steel sits well below both, and the difference becomes movement: across one metre and a hundred degrees of service an austenitic member wants to grow roughly four tenths of a millimetre more than its carbon neighbour, and something in the assembly absorbs it. Duplex narrows the gap rather than closing it. Alloy 2205 publishes linear expansion of 7.5 to 8.1 x 10⁻⁶ ⁄°F across its elevated-temperature rows, which converts to roughly 13.5 to 14.6 x 10⁻⁶ ⁄K, closer to the ferritic and carbon side than to 304. The 430 coil page makes the same comparison in words, noting better conductivity than austenite and a smaller expansion coefficient. Conductivity then complicates the arc itself: the 904L page states that austenitic steel carries roughly thirty per cent of the conductivity of carbon steel and gives twelve W/m.K at twenty degrees, while Alloy 310S quotes thirteen point eight at one hundred degrees rising to eighteen point seven at five hundred. Heat leaves the joint asymmetrically, so heat input, travel speed and where the arc sits in the groove all change when one parent is stainless.

The filler belongs to the joint

Every alloy page that takes welding seriously publishes a filler designation, which is a hint about how much the choice matters. 2507 lists weld wire AWS ER2594 with electrodes E2594-16 and E2553T-1. Alloy 718 names Filler Metal 718 under AWS 5.14 as ERNiFeCr-2. Monel 400 gives both routes, Filler Metal 60 as ERNiCu-7 and Electrode 190 as ENiCu-7. Alloy 625 is described as stabilised by niobium against sensitisation during welding. None of these is met by reaching for whatever suits the cheaper parent. The principle is stated outright on 904L: the corrosion stress has to be regarded when choosing the filler, and a higher alloyed filler can be necessary because of the cast structure of the weld metal. A cast structure has no rolling direction, no ductility to spare and less alloy than either parent once dilution is done with it. The answer is a filler richer in chromium, nickel and molybdenum than the lower-alloyed side needs, named by AWS classification on the drawing rather than left to the shop floor.

After-weld heat treatment divides the two halves

Carbon practice often wants a stress relief at twelve hundred degrees Fahrenheit and above, which is the same band that causes trouble on the corrosion-resistant side. 2205 should be annealed at nineteen hundred degrees Fahrenheit minimum followed by rapid cooling, ideally water quenching, and the treatment applies to both solution annealing and stress relieving, because stress relief at any lower temperature risks precipitating detrimental phases. There is no safe low-temperature stop for duplex. Hastelloy C-276 reaches the same place from the other direction, with low carbon minimising grain-boundary carbide precipitation so that the heat-affected zone keeps its corrosion resistance without any treatment afterwards. Where a cycle cannot be avoided, the stabilised austenitics exist for exactly that reason. The 347 page explains that columbium and tantalum have a stronger affinity for carbon than chromium, so they take the carbon and leave chromium in solution; 321 is the titanium equivalent, weldable by most resistance and fusion methods. If the joint will be heat treated with the carbon half, one of those two belongs in the specification from the start.

Choose the stainless half by service

Once the joint exists, the environment usually decides the alloy, and the ladder is written on the pages. Cyclic high-temperature duty points at the stabilised grades first: 347H pipe quotes five hundred and fifteen MPa tensile, two hundred and five MPa proof and thirty-five per cent elongation, while 310 and 310S pipe carries oxidation resistance in continuous service to two thousand degrees Fahrenheit. Temperature-dominated work lands on 409, supplied in stabilised forms listed as S40930, S40920 and S40910 with niobium, titanium or both. Chloride duty moves onto molybdenum: 316L, then 317 at three to four per cent, then 904L and 254 SMO. Where strength and expansion both argue against austenitic, 2205 pipe offers roughly twice the yield and sits closer to carbon steel thermally, with the caveat that 2507 is limited to service below three hundred and sixteen degrees Celsius.

Sometimes the better joint is not a weld

Three alternatives are worth costing before a procedure is qualified. Absorb the mismatch mechanically: Alloy 625 is listed for bellows and expansion joints because it tolerates movement that a rigid weld has to resist. Or butter the carbon face first, building it up with the intended filler, then treat it as a carbon steel item before the final stainless run, so no cycle ever touches the corrosion-resistant metal. Or remove the problem by running one material through, which is what the structural pages assume: frames built from welded H sections in Q235B and Q345B, or the steel structure programme carrying the whole frame in one family. Where the stainless run is short, a transition piece is cheaper than a qualified procedure.

Five lines for the procedure

A dissimilar joint is described completely in a short specification, and every line prevents a known failure. Name both parents with their standard: Q235 plate, Q345 and A36 are all listed for weldability, and none of them is stainless. Name the filler by AWS classification, since dilution is not kind to either parent. State whether post-weld heat treatment applies, to which half, and whether buttering replaces it. Give the service temperature range, which drives both the expansion mismatch and the choice between 321 and 347. Then add the separation rule the 316 and 316L page already hints at when it warns that copper and zinc must not contaminate the site of welds: stainless keeps its own wheels, brushes and handling gear, and where preparation is needed afterwards the surface treatment, cutting and welding programmes run as separate line items alongside custom processing.