Brazed cold plates usually scrap for three reasons, and only one of them is the braze joint itself. A vacuum or controlled-atmosphere braze cycle takes the whole part above roughly 570 C, which anneals 6061 to the O condition and releases the residual stress left by machining, so the plate moves and flatness is lost. A single incomplete joint out of hundreds fails the leak test and condemns the whole assembly. And a brazed part that has already been through the furnace generally cannot go back in, so there is no repair route and the loss is the full machined value of the part, not the value of the joint.
Where does the money actually go on a scrapped cold plate?
By the time a cold plate reaches the braze furnace, nearly all of its cost is already committed: billet, machining of the channel pocket, cleaning, and fixture time. The braze cycle itself is a small share of unit cost and a large share of unit risk. That asymmetry is what makes brazing scrap expensive rather than merely annoying.
| Stage | What is committed | Recoverable if the part fails later? |
|---|---|---|
| Material | Plate or billet, often 6061 or 6082 | Scrap value only |
| Machining | Channel pocket, ports, mounting features, flatness prep | No |
| Pre-braze cleaning and fixturing | Chemical clean, assembly, furnace fixture time | No |
| Braze cycle | Whole-part thermal cycle above roughly 570 C | No |
| Leak and pressure test | Pass or fail on the complete assembly | Fail condemns everything above |
| Post-braze skimming | Machining back to flatness after distortion | Consumes wall thickness |
Why does a brazed plate lose flatness even when the joint is sound?
Two mechanisms act at once, and both are properties of the furnace cycle rather than of the joint.
- Temper loss. A vacuum or CAB cycle holds the entire part above roughly 570 C. For 6061-T6 that is well past solution temperature, so the plate comes out effectively annealed. Yield strength falls, and the plate is now easier to distort during handling, test and assembly.
- Stress relaxation. Machining a deep channel pocket into rolled plate leaves an unbalanced residual stress field. The furnace cycle relaxes it. The plate then moves to a new equilibrium shape, which is not the shape the drawing asked for. On long, thin plates the movement scales with length and with the depth of material removed.
The usual answer is to add post-braze skimming, which costs a second machining operation and eats into the wall thickness above the coolant channel. That wall is also the thermal path, so there is a limit to how often you can buy flatness this way.
Why does one bad joint out of hundreds scrap the whole part?
A brazed cold plate or heat exchanger is joined all at once. Hundreds of individual joints — dimples, fillets, tube-to-header interfaces — are formed in a single cycle, and the assembly is then leak tested as one object. Yield at the assembly level is the product of the yield at every joint, so an individually excellent joint yield still produces assembly-level losses once the joint count gets large. Worse, the failure is discovered after the whole part has been built.
This is why brazing scrap tends to arrive in batches rather than as a steady trickle: a drift in furnace atmosphere, clad thickness, flux coverage or fixture loading shifts joint yield slightly, and the assembly-level yield falls sharply.
Why is there no repair route?
A brazed assembly that failed a leak test cannot usually go back into the furnace. Re-running the cycle re-melts the filler everywhere, not only at the defect, risks flowing filler out of sound joints, and puts a second full thermal cycle into a part that is already annealed. In practice the part is condemned even though the defect might be a single millimetre-scale feature that is visible, accessible from the outside, and mechanically trivial.
That gap — an accessible defect on an otherwise good, fully machined part with no sanctioned repair — is worth attacking directly. See repairing a failed brazed heat exchanger.
What did other industries do about it?
Industries that hit the same wall — committed cost, whole-part thermal cycles, no repair route — largely moved the closure joint out of the furnace and made it a local, solid-state joint instead. Friction stir welding is the process that made that practical for aluminium. A rotating tool stirs and forges the material below its melting point, so the joint is formed by local plastic deformation rather than by melting and re-solidification.
The consequences matter more than the mechanism:
- Heat is applied locally along the weld line, not to the whole part, so the bulk of the plate keeps its temper and its residual stress state is disturbed far less.
- There is no filler, no flux, no shielding gas and no furnace, so there is no clad layer to consume and no controlled substance to administer.
- Because there is no melting, the weld has no solidification porosity and no hot cracking — the two defect families that dominate fusion welding of aluminium.
- Joints are made sequentially and can be inspected and, if necessary, repaired individually rather than gambling the whole assembly on one cycle.
- Aluminium traverse speeds are typically in the range 750 to 1500 mm/min, against roughly 130 mm/min for TIG, so a closure weld is not the bottleneck.
Liquid cold plates and EV battery cooling assemblies are now among the most established commercial applications of friction stir welding, precisely because the closure weld on a machined pocket is the geometry the process suits best.
Does that mean brazing is the wrong process?
No, and this is worth being blunt about. Brazing remains the right answer when a part has hundreds of joints that must all be made at once at furnace throughput, or when joints are internal and no tool can physically reach them. A friction stir welded design has to be a design with reachable joint lines. If your part is a 700-joint furnace part, the honest recommendation is to fix the furnace process, not to change joining technology.
Where friction stir welding displaces brazing is the opposite case: a modest number of long, reachable closure or perimeter welds on a part whose value is already in the machining, and where flatness, temper and leak-tightness are the things that decide yield.
What to do next with a scrap problem you already have
- Split your scrap by mechanism: flatness or distortion, leak or joint defect, handling damage. The three have different fixes and are usually reported as one number.
- Check whether the value is in the machining. If it is, the cost of a failed joint is the cost of the part, which changes the economics of any alternative.
- Check reachability of the closure joint line. That single geometric question determines whether friction stir welding is even a candidate.
- Trial before you commit. A structured feasibility trial on representative material answers the flatness and leak-tightness question with measurements rather than argument.
Frequently asked questions
Can a cold plate be friction stir welded without losing the 6061 temper?
Away from the weld, largely yes. Friction stir welding is a local process, so the bulk of the plate never approaches solution temperature. There is still a heat affected zone adjacent to the weld line where properties change, and it has to be accounted for in the design. What does not happen is the whole-part anneal that a furnace cycle imposes.
Is a friction stir welded cold plate leak tight?
Leak-tight closure welds are the normal application. Because the process does not melt the material, the weld does not contain solidification porosity, which is the usual leak path in fusion-welded aluminium. Acceptance is proven by test on the customer specification.
How much distortion should I expect compared with brazing?
Friction stir welding has the lowest heat input of the routes typically assessed for this component family, and stationary-shoulder variants lower still, so distortion is substantially reduced rather than eliminated.
Why does one bad joint out of hundreds scrap the whole part?
A brazed assembly is joined in a single cycle and leak tested as one object, so assembly yield is the product of the yield at every joint. With a large joint count, even a high per-joint yield produces assembly-level losses, and the failure is only discovered after the part is complete.
Why can a failed brazed part not be repaired in the furnace?
Re-running the cycle re-melts filler at every joint rather than only at the defect, can flow filler out of sound joints, and puts a second whole-part thermal cycle into a part that is already annealed.
Next step A structured feasibility trial answers the question on your geometry, with measurements rather than argument.