A gloved hand holds a tool inside the outlet of a stainless merge collector. The tool is a segmented expansion die: the splits visible around its circumference let it grow in diameter when something is driven into its centre, and the bronze-coloured piece at the core is what does the driving.
The collector above it is finished and polished. The bore it has to fit is at the bottom of the picture, and that is the surface being worked.
This is not forming. The shape already exists. What is happening is sizing — a correction applied to a welded assembly so that its outlet is round and at the diameter the next component needs. The caption on the image says precision sizing and perfect fit, and the order of those two words is the whole story: sizing first, fit as the consequence.
The outlet dimension on the drawing describes a part before anyone welded it.
Welding moves metal. A circumferential weld around a cone-to-tube joint contracts as it cools, and it contracts unevenly, because the heat was not even and the restraint was not even. The bore pulls in slightly. The section that was a circle becomes very slightly not a circle. The end of the tube, which was cut square, picks up a small amount of out-of-plane movement depending on where the last weld ran and in which direction.
None of that is a defect. It is what welding does, to every part, every time, and a shop that pretends otherwise is either not measuring or not telling you.
The consequence is simple: the outlet you measured before welding is not the outlet you have after welding, and the difference is small enough to look acceptable and large enough to make a part not fit.
Follow the chain downstream. A collector outlet usually meets one of three things: a turbo inlet flange, a midpipe slip joint, or the next section of the system.
All three are fits, and every fit has a window. Too large and the joint relies on the clamp to crush the tube down, which distorts it, which changes the bore the exhaust is flowing through, and which on a V-band leaves a gap that will not seal no matter how tight the bolt goes. Too small and the part will not go on at all — and it will not go on more stubbornly when it is hot, or after a season of thermal cycling has grown a layer of oxide on both surfaces.
The window is not generous. It is a few tenths of a millimetre on a nominal diameter, and inside that window the part either slides together by hand or it does not.
That is why a merge collector is judged at the outlet. The merge point determines what the exhaust does. The outlet determines whether the exhaust is on the car.
Here is the failure that hides from a caliper.
Take a tube that has been pulled slightly out of round at a weld and measure it across one diameter. It comes out near nominal. Rotate the caliper ninety degrees and measure again, and it also comes out near nominal. The part passes two checks and is still wrong, because a section can be oval in one direction and pinched in another, and two diameters do not describe a shape with three lobes.
A three-lobed bore will not seal on a round clamp. It will slide onto a round pipe and rock. It will read correct on a two-point measurement and leak in service, and the person who assembles it will conclude that the clamp is bad or the flange is warped.
This is the reason the tool in the picture is segmented rather than a single wedge. A die made of discrete petals that all move outward together pushes the bore out in every direction at once, so the shape it produces is round. A single-wedge expander pushes hardest where it touches, and turns a two-lobed oval into a three-lobed one while the number on the caliper improves.
A shop can be doing an expansion step and still be making ovals. The difference is the tool and how many points around the circumference it acts on.
There are two directions to correct a bore, and they are not interchangeable.
Squeezing from the outside reduces diameter. It also flattens: a reducing die pushes the section inward and the material has to go somewhere, so it thickens at the contact points and the section loses roundness. This is a legitimate operation for making a reducer, and the wrong one for restoring a weld-distorted outlet, because the correction you need is outward.
Expanding from inside grows the bore back to size and, done with a segmented die, grows it evenly. The wall thins slightly as the circumference increases, which is the cost, and it is why the amount of expansion is a decision rather than a habit. A part that needs to grow a lot has a problem upstream — a cone cut short, a joint fitted with the wrong gap — and expansion will hide that problem rather than solve it.
There is a second effect worth knowing: expanding work-hardens the material at the outlet. That is mostly helpful at a slip joint, because the surface that has to survive being clamped repeatedly is now harder. It also means the outlet will not take a second correction as easily as it took the first, so the first one has to be right.
The same physics that makes a bent tube give back a degree of angle when the tool releases applies to an expanded bore. Push the die to the target and the tube will spring back below it.
Which means the number the die is set to is not the number the part finishes at, unless someone measured the difference on that material and that wall thickness and built it into the setting. A shop that expands to nominal and stops will produce outlets that are consistently a fraction under, and consistently under is the failure mode that gets blamed on the clamp.
Ask what the die is set to relative to the finished diameter. An answer in terms of a measured allowance means someone checked. An answer in terms of the nominal means someone assumed.
It is worth being clear about the limits, because the image suggests a fix-all and it is not one.
Sizing corrects diameter and roundness. It does not correct:
Angle. If the outlet axis points three degrees off where the drawing says, expanding it produces a perfectly round outlet pointing the wrong way. The joint will assemble and the system will sit wrong.
Flatness. If the collector butts to a flange, the face has to be flat as well as round, and an expansion die works on the bore, not the face.
A joint that was mis-fitted. If the cone and the outlet tube were joined with the wrong gap, the resulting step is inside the part and the die cannot reach it. Expansion makes the outside correct around an internal defect.
A weld that already cracked. Expanding a cracked weld opens it.
Four things that a sizing step does not touch, all of which live upstream in the fit and the weld. This is why the operation appears late in the sequence rather than early: it is the last chance to correct the geometry, and it can only correct the geometry it is pointed at.
The order matters more than the operation.
Cut the primaries and the cone. Form the merge. Fit and tack. Weld out. Then size the outlet, because everything before it moves the metal and everything after it depends on the outlet being right. Then trim the end square, and then check the face if there is one.
Sizing before welding would be measuring a part that is about to change. Sizing after trimming risks the trim being referenced to a bore that is still oval. The step belongs between the weld and the final trim, and in a shop that has this operation as a station, it is there for exactly that reason.
The photograph supports this reading: the collector is completed and polished, and the tool is in it afterwards.
It is gloved, and it is holding the tool. This is not a machine cycle with a guard around it.
That is worth saying plainly rather than dressing up. Sizing a collector outlet is a discrete, checked operation performed per part: locate the die, expand, release, measure. The judgement is in the setting and in the measurement, not in the leverage.
The value of that is adaptability. A die set that can be adjusted to a target covers the range of diameters the shop builds without a dedicated machine for each. The risk is that a manual step is only as good as its measurement discipline, which is why the interesting question is not whether the shop expands the outlet — many do — but what they measure afterwards and how many points they measure it at.
Four questions, and none of them requires a shop visit.
What is the outlet diameter relative to the mating part? Not the nominal, the clearance. A number, in the unit you work in, with the direction stated.
How is roundness checked, and at how many points? Two diameters at ninety degrees is the minimum and does not catch a three-lobed bore. Three or six points does.
Is sizing done after welding? If it is done before, the answer to the first question describes a part that no longer exists.
What is the springback allowance on the die? Because expansion springs back the same way bending does, and an allowance that was measured is different from one that was assumed.
A supplier who can answer all four is running a process. A supplier who answers two is running an operation and hoping.
There is a reason to photograph this step rather than the merge.
The merge is where the engineering is interesting and where every supplier's marketing goes, because it is the part with the theory attached. The outlet is where the consequences are, and it is much harder to make impressive: a round hole at the right size is not a photograph anyone shares.
But it is the surface that decides whether a beautifully merged collector becomes a working exhaust or a part that has to be persuaded on with a soft mallet and then leaks at the clamp. The merge point is the part you buy for; the outlet is the part you live with.
Precision sizing and perfect fit, in that order. The first is a measurement, the second is what the measurement buys. Between them sits a die, a springback allowance, and somebody deciding to check the bore after the last weld instead of trusting the drawing.
🎯 Outlet sizing after welding — the bore is corrected once the assembly has stopped moving, not before.
🔵 Segmented expansion die — discrete petals grow outward together, so the section comes out round rather than lobed.
📐 Diameter and roundness, both — a two-point caliper check misses a three-lobed bore; sizing works the whole circumference.
🧱 Expanded from the inside — correcting outward instead of squeezing inward, which is what a weld-distorted outlet actually needs.
⚖️ Wall thinning traded knowingly — expansion costs a little wall, so the amount is a decision, not a habit.
🌡️ Springback allowance built into the die — the setting is not the finished size, and someone measured the difference.
🛠️ A station in the sequence, not an afterthought — cut, form, fit, weld, size, trim: the order is the process.
🧐 Measured at several clock positions — because a section can read correct across two diameters and still leak.
🚫 Clear about what it cannot fix — axis angle, face flatness, an internal step and a cracked weld all survive sizing.
🔁 Work-hardened at the joint — the surface that takes repeated clamping comes out harder than it went in.
🤲 Per-part, checked, adjustable — a hand operation with a measurement behind it, which is what the photograph shows.
🏭 Four facilities, five material families — exhaust, silicone, carbon fibre and pre-filter production, working in titanium, stainless steel, aluminium, carbon fibre and silicone.