A bend is usually specified as a centreline radius and an angle. Neither of those describes what happens to the tube. Before the axis turns, the cross-section stops being round, and that change creates most of the problems people later blame on the bend.
Follow a tube through a bend and the geometry is easy to see. The outside of the curve has further to travel than the inside. Since the material at both extremes started as the same length of wall, something has to give. The outside wall stretches and thins. The inside wall compresses and thickens, and if it cannot thicken fast enough it wrinkles instead. Between those extremes the section flattens, so the diameter measured across the plane of the bend shrinks.
That flattening has a name, ovality, and it is read as the difference between the largest and smallest diameter at the same station. A formed bend always has some. The question is how much, where it sits, and whether anything downstream cares.
The outside wall is doing the work of the bend. It is stretched along the tube axis, which reduces wall thickness in proportion, and on a tight radius in thin wall the loss is large enough to change how the part behaves in service. A thinned outside radius is the first place a fatigue crack appears on a vibrating exhaust, and it is also the first place a weld will burn through if someone joins a hanger or a boss there.
The inside wall is doing the opposite. It is being pushed together, so the metal upsets and thickens. Up to a point that is harmless. Beyond it the wall buckles into a series of ripples along the intrados, and a wrinkled bend cannot be recovered by any downstream operation.
Between them, the neutral line stays close to its original length, which is why the developed length of a bent tube is measured on the centreline rather than on either wall.
The tooling inside the tube during the bend decides how much of this happens. With no internal support the wall is free to collapse inward and the section flattens early. A mandrel — a shaped plug on a rod, positioned where the bend begins — holds the section open from the inside. A ball mandrel goes further and supports the intrados through the turn, which is what allows a tight radius in thin wall without wrinkling.
The cost is setup and drag. A mandrel has to be sized to the bore, positioned accurately and lubricated, and it leaves the inside of the tube needing a clean before anything else happens. It also has to come out again. On a long tube with several bends in different planes that is a real constraint, and it is one reason a fabricated turn made from cut segments sometimes beats a formed one.
Flow area falls with the diameter, so a section that has flattened in one plane loses area even though its perimeter is unchanged. In an exhaust, a small loss of area in a single bend usually matters less than the change in shape, because the flattened section presents a different profile to the gas and disturbs the boundary layer around the whole turn.
Where it matters more is a bend close to a component sensitive to velocity profile. A flattened section immediately upstream of a catalyst, a sensor boss or a merge collector changes the distribution entering it. Whether that is measurable in a given system is a question for testing rather than assertion. What a fabricator can say with confidence is that the smoothest path is the one that keeps section shape as well as direction.
The cost that shows up first in a workshop is fit. Every slip joint, every clamped hose connection and every spigot relies on two round sections being close enough in diameter for the clamp to do its job. Put a bend close to a joint and the section at the joint is no longer round, so the clamp has to close an oval onto a round and the seal is made at two points rather than around the circumference.
The same applies to a flange welded onto the end of a bend. A flange wants a square, round end to sit against. An oval end leaves a gap that varies around the joint, and the welder closes it by adding filler where the gap is largest, which puts more heat and more shrinkage into one side of the part than the other. That is how a bend ends up with a flange that is not square to the axis.
Metal does not stay where the tool leaves it. The elastic recovery after the load comes off is springback, and it is larger on thin wall, on high-strength material and on generous radii. A bend set to a nominal angle comes out slightly open, so shops compensate by overbending or by pressing the angle afterwards.
Springback also varies along the bend, because the material at the start of the turn has been work hardened by the tooling and behaves differently from the material in the middle. If the finished angle matters, it has to be checked on the part rather than assumed from the machine setting, and on a multi-bend part the error accumulates from one bend to the next.
A bend next to a weld concentrates three problems in one place. There is a heat-affected zone on one side, a cold-worked and thinned wall on the other, and a change of section between them. If the assembly is later stress relieved or heat straightened, the two zones respond differently and the joint can move.
Where the design allows it, a straight length between the end of the bend and the weld gives the welder a round section to work with, gives the inspector a datum, and keeps the arc away from the thinnest part of the wall. That straight length costs packaging space, and packaging space is usually the thing under pressure.
The most useful thing a drawing can do is say where the bend starts and stops. A centreline radius and an angle do not tell the shop where the tangent points are, and without them the joint may be placed anywhere the fitter finds convenient. Marking the tangent points, and adding a minimum straight at each end, turns a preference into an instruction.
The second most useful thing is to allow ovality where it does nothing. A bend in the middle of a run with no joint and no sensor nearby can be allowed to flatten; a bend feeding a clamped hose cannot. That distinction belongs on the drawing rather than in the shop's judgement.
Our exhaust pipe work covers both formed bends and fabricated turns, and the difference is obvious when a mandrel bend is set next to a pie-cut one.
Ovality is not a defect to be eliminated. It is a predictable result of a process, and the useful conversation is about where it is acceptable.
Bends that are later joined into a system are shown among our exhaust pipe sections and in our exhaust pipe range, and the welding that follows is described under TIG stainless steel welding. Where a formed bend is replaced by a fabricated turn, the assembly side sits under titanium fabrication components, and the route from drawing or sample is set out on our customization page.
