Tube Bending: The 33% Wall Loss in a 1D Bend, the Mandrel That Fixes Ovality and Thins the Wall, and Why 304 Lands 2.3° Open
A 25.4 mm OD × 1.5 mm wall 304 stainless line leaves the bender with two 90° bends at 1D. It is a good-looking part: no wrinkles, no visible marks, the bend planes line up. Then four numbers come back from inspection. The wall at the extrados of each bend measures 1.02 mm against a 1.5 mm nominal. The section has flattened to 9% ovality. The angles are 88.7° instead of 90°. And both legs are 25 mm short.
None of those were manufacturing failures. The bender did exactly what it was told. Each number was decided before the tube was loaded: the thinning by the radius-to-diameter ratio, the ovality by the tooling that ratio demands, the springback by the material's flow stress over its modulus, and the leg length by a drawing dimensioned to the intersection of centre lines while the cut list added the arc but not the setback.
Tube bending has this in common with press brake work: the machine is precise and the arithmetic is where the part is lost. But tube adds a second axis of difficulty that sheet does not have — the cross-section is a closed hollow section, so every bend is a competition between the outside wall stretching, the inside wall thickening, and the section trying to collapse into an oval. The tooling exists to referee that competition, and the ratios that decide the tooling are two divisions.
This guide works through them in the order a part meets them: the geometry that sets the tooling, the wall-thinning law and the counter-intuitive role of the mandrel, flattening and the code limits that govern it, springback and seam orientation, the cut length arithmetic, the codes for pressure service, and the cost structure that makes a 1D bend expensive.
Five tools, two ratios
Rotary draw bending is the job-shop standard because it is the only common process that can hold a tight radius in thin wall without collapsing the section. The tube is clamped to a circular bend die of fixed centreline radius (CLR), and the die rotates, dragging the tube around it. Five tools do the work:
- Bend die. Sets the radius. Its groove diameter is the tube OD plus clearance, and its working diameter is 2R_c where R_c is the centreline radius. One die per OD and per CLR — which is why shops keep 1.5D, 2D and 3D sets on the shelf and quote tooling for anything else.
- Clamp die. Grips the tube against the bend die so the tube cannot slip while the die rotates. Clamp length and clamp force are what stop the tube pulling through; the minimum straight length for the clamp is roughly 1.5–2 × OD.
- Pressure die. Backs the straight section and pushes the tube forward into the bend, resisting the axial tension the draw creates. Its pressure setting is the most-abused control on the machine: too little and the inside radius wrinkles continuously, too much and the outside flattens between the pressure die and the clamp die.
- Mandrel. Sits inside the tube at the tangent point and supports the bore so the section cannot collapse. Plug, single-ball, multi-ball, or cable-linked ball mandrel depending on how much support and how much articulation the radius needs.
- Wiper die. Fills the gap between the bend die groove and the tube on the inside of the bend, with a feathered tip that reaches to the tangent point. It removes the space a wrinkle would occupy. A wiper is not a mandrel substitute; it is what you add when the mandrel has stopped the collapse but the intrados still humps.
Two ratios decide which of those five a job needs, and both are dimensionless:
F_w is the wall factor — outside diameter over wall thickness. F_d is the D of bend — centreline radius over outside diameter. A 50.8 mm tube on a 76.2 mm radius is a 1.5D bend regardless of whether the drawing is in millimetres or inches. A 50.8 mm tube with a 1.65 mm wall has a wall factor of 30.8.
The two ratios are independent, and that independence is the whole design space. A thick-wall 1D bend and a thin-wall 3D bend can both be easy or both be miserable. Combined, they place a job in one of four bands:
Difficulty · D of bend · Wall factor · Tooling normally required
Easy · 3D and above · Under 20 · Bend die, clamp die, pressure die
Moderate · 2D to 3D · 20 to 30 · Add a plug or single-ball mandrel
Hard · 1.5D to 2D · 30 to 40 · Multi-ball mandrel, wiper die usually needed
Severe · Under 1.5D · Over 40 · Multi-ball mandrel, wiper die, often a following pressure die
Two published rules of thumb compress the chart into a sentence each. A wall factor of 15 or more and a mandrel becomes likely; a D of bend of 1.5 or less and a mandrel is almost certain. For the wiper, the usual trigger is a wall factor of 20 or more with a D of bend of 3 or less. These are starting points for an enquiry, not a process specification — the ovality limit on the drawing is what actually decides, because ovality is the one defect a mandrel is unambiguously good at fixing.
The bend difficulty rating
Tooling suppliers also publish a single number that combines the ratios with the bend angle. In the form Bend Tooling uses:
with K_r \approx 2 for most applications and B the bend angle in degrees. A rating of 7 or below is a straightforward rotary-draw bend; above 7 the setup has to be precise and the process watched. Evaluate the same 50.8 mm tube three ways:
Case · OD × wall · CLR · F_w · F_d · Angle · F_b
A · 50.8 × 2.0 · 101.6 (2D) · 25.4 · 2.0 · 90° · 4.8
B · 50.8 × 1.65 · 76.2 (1.5D) · 30.8 · 1.5 · 90° · 7.1
C · 25.4 × 1.2 · 25.4 (1D) · 21.2 · 1.0 · 180° · 9.2
Case A is a two-minute setup. Case B sits exactly on the threshold, which is why 1.5D in 16-gauge is the bend that separates shops that know their tooling from shops that quote and hope. Case C is a hairpin return in thin wall: three or more balls, a wiper, low pressure, and a first-article check on every setup.
The wall-thinning law: one divided by two D plus one
The outside of a bend stretches. Take the outer fibre at radius R_o = R_c + D/2; a bend through angle \theta takes it from an arc of length \theta R_c to \theta R_o. The elongation at the arc is:
The wall cannot stretch in the hoop direction, so constant volume forces the thickness to absorb that strain. The fraction of the original wall lost is not P_e but P_e/(1+P_e), because the final thickness is the original divided by the stretch ratio:
That single expression is the most useful number in tube bending. It says thinning depends only on the D of bend, not on the alloy, not on the wall thickness, not on the machine:
D of bend · Ideal extrados thinning · 2.0 mm wall becomes · 1.5 mm wall becomes
1D · 33.3% · 1.33 mm · 1.00 mm
1.5D · 25.0% · 1.50 mm · 1.13 mm
2D · 20.0% · 1.60 mm · 1.20 mm
2.5D · 16.7% · 1.67 mm · 1.25 mm
3D · 14.3% · 1.71 mm · 1.29 mm
4D · 11.1% · 1.78 mm · 1.33 mm
5D · 9.1% · 1.82 mm · 1.36 mm
The law is geometric, so it is an upper bound rather than a prediction. Three effects pull real thinning below it: the neutral axis shifts toward the intrados as the section yields, the intrados thickens and feeds material around the bend, and pressure-die assist pushes metal into the extrados. Measured values on a 90° bend with a three-ball mandrel land near 25.8% thinning against 15.3% thickening at the intrados — the asymmetry that tells you the neutral axis moved. On a 1.5D bend in 1.65 mm wall, expect 20–25% at the extrados in production: the geometric bound is 25%, and tooling and boost claw back a few points.
The design consequence is blunt. If the pressure design or the fatigue calculation needs 1.2 mm of wall in the bend, you cannot buy 1.2 mm tube and bend it. You buy:
At 1.5D that is 1.2 \times 1.333 = 1.60 mm. At 1D it is 1.80 mm. The same formula in TEMA's notation, t_0 = t_1[1 + D/(C R_c)], is the code version of this arithmetic, and the constants are worth reading closely: TEMA uses C = 4 for carbon steel, low-alloy, ferritic and austenitic stainless and copper alloys, and C = 2 for martensitic, duplex, super-austenitic stainless, titanium and high-nickel alloys. C = 2 reproduces the pure-geometry allowance exactly; C = 4 halves it. The code is not being generous to carbon steel — it is assuming that ductile, non-work-hardening material redistributes strain around the arc better than a duplex or a nickel alloy does, and history says it does.
The mandrel paradox
Here is the part that surprises people who have only read the tooling charts. The mandrel supports the bore, kills ovality, and lets you bend a 1D radius in thin wall. It also increases wall thinning.
A finite-element and experimental study of 2169 stainless tube bent to a 19.05 mm radius found maximum extrados thinning of 6.94% with no mandrel, 12.1% with a cylindrical mandrel, and 13.6% with a bulb mandrel. The mechanism is friction. The mandrel is in sliding contact with the bore through the whole bend zone; that contact drags against the material trying to flow into the bend, raises the tangential tensile stress at the extrados, and the wall pays for it. More contact area, more drag: the bulb mandrel thins more than the cylinder, and a larger mandrel diameter thins more than a smaller one — 10.1%, 11.6% and 13.6% for diameters of 5.23, 5.33 and 5.43 mm on the same bend.
The same trade appears in mandrel extension length. Pushing the mandrel nose further past the tangent point improves roundness and reduces cross-section distortion, and simultaneously raises thinning — 19.3% at zero extension, rising as the nose advances, in a large-diameter study where three balls was the optimum for section quality (22% distortion with one core, 4.2% with three, 3% with four) while thinning went the other way.
The practical reading is not "avoid the mandrel". It is that the mandrel is a section-quality tool, not a thinning tool, and the two objectives have to be traded explicitly. If wall thinning governs — pressure service, fatigue, a hydro test — the levers are radius, starting wall and axial assist, in that order. If ovality governs — hose barbs, ferrules, bearing fits, tube-sheet holes — the mandrel and wiper are the levers, and the wall allowance must be bought up front.
Boost and assist, and the springback bill
Modern CNC benders offer two axial aids that are easy to confuse:
- Pressure die assist feeds material into the extrados only, by driving the pressure die slightly faster than the bend die's linear velocity. Its effect is reduced flattening and reduced thinning on the outside radius.
- Boost applies axial compression to the whole circumference through clamps behind the pressure die, pushing material through the tangent line into both the intrados and the extrados.
Boost is the stronger tool and the more expensive one. Published work on boosting shows the gain comes from two mechanisms at once: the friction from the boost creates a compressive stress that partly cancels the bending tension, and the axial feed physically transports material into the bend zone. The condition for it to work is that the boost velocity be slightly greater than the bending linear velocity — studies sweep 0.9× to 1.3× and find the useful range just above 1.0×. Below that, the boost does nothing; far above it, the tube upsets ahead of the bend and the process destabilises.
The bill comes due at springback. Pressure-die boosting reduces thinning and axial elongation, and increases springback angle. That is not a defect — it is the same physics from the other side. Any process that reduces the plastic strain in the outer fibre leaves a larger elastic core to recover on unloading, and the elastic core is what springs the part open.
Flattening: the number the codes actually limit
A bend does not stay round. The section ovalises because the resultant of the tensile stress at the extrados and the compressive stress at the intrados both point toward the centre of the section, and nothing resists that flattening except the wall's own bending stiffness and whatever tooling is inside the tube. Ovality is measured as the difference between the largest and smallest diameter in one cross-section:
Note the denominator. Some inspection procedures divide by the mean of the two readings instead of nominal; on a tube inside its mill tolerance the two differ by a fraction of a percent, but on a rejected bend the difference is enough to argue about. Name the definition in the inspection plan.
The limits in pressure service are not shop folklore:
Specification · Limit · Applies to
ASME B31.3 para. 332.2.1 · 8% of nominal OD · Bends under internal pressure
ASME B31.3 para. 332.2.1 · 3% of nominal OD · Bends at weld ends under external pressure
TEMA RCB-2.3.1 · 10% of nominal OD · Heat-exchanger U-bends
API 660 · 10% · U-tubes; minimum mean radius 1.5D (2.0D for duplex, super-austenitic, Ti, high-Ni)
Common shop guideline · 5% · Where no code limit is stated
Mandrel bending, careful · 3–5% · Achievable, not guaranteed
B31.3 also forbids the fix that every shop has tried once: removal of metal shall not be used to achieve the flattening requirement. Grinding the high side of an out-of-round bend to make a caliper reading pass is a code violation, not a repair, and it attacks the wall thickness the same paragraph requires.
What the numbers mean in practice: without a mandrel, a 1.5D bend in thin wall comes out at 8–12% ovality and a 1D bend can exceed 20%. With a multi-ball mandrel and a wiper, 3–5% is routine and 2% is achievable with a good setup. That is the entire commercial argument for the tooling: not cosmetics, but the difference between passing and failing a specification that has a number in it.
Springback: why 304 lands 2.3° open
Release the load and the bend opens. The bend zone is not uniformly yielded — an elastic core survives near the neutral axis — and on unloading that core recovers, taking the whole section with it. For a thin-walled tube treated as fully plastic, the plastic moment is M_p = 4\sigma_f r_m^2 t and the elastic unloading curvature is M_p/(EI) with I = \pi r_m^3 t, which gives the scaling:
Read it as a shape law, not a calculator. Springback grows linearly with bend angle, grows with the D of bend (a gentle 3D bend springs much more than a tight 1D bend), and grows with the ratio of flow stress to modulus. It falls as the strain-hardening exponent rises, because a strongly hardening material spreads plastic strain through the section and leaves a thinner elastic core — the same reason an annealed 304 is easier to predict than a cold-worked one even though both have the same modulus.
The scaling formula under-predicts the real thing by roughly a factor of two, because the section is only partly plastic, the neutral axis shifts, and mandrel drag raises the axial stress the bend zone sees. A bending house's published reference table for 90° bends at 2D, with overbend programmed to land at 90°:
Material · Program · Springback · Character
Mild steel 1018/1020 · 91.8° · 1.8° · Baseline
DOM steel · 91.7° · 1.7° · Cleaner wall, same order
HREW steel · 91.9° · 1.9° · Hot-rolled, slightly more
Stainless 304 · 92.3° · 2.3° · Work-hardens in the bend
Stainless 316 · 92.5° · 2.5° · The worst of the common stainless
Aluminium 6061-T6 · 91.4° · 1.4° · Low modulus, but low flow stress
Aluminium 5052 · 91.3° · 1.3° · Softer again
Titanium Grade 2 · 93.6° · 3.6° · High \sigma/E
Titanium Grade 5 · 94.5° · 4.5° · Hard to bend cold at all
Inconel 625 · 94.5° · 4.5° · Heavy springback, heavy work hardening
Inconel 718 · 95.4° · 5.4° · The top of the table
Two warnings about that table. First, springback grows with the D of bend: the same 304 at 4D will spring 4–5°, and at 1D it will spring less than 2°. Second, the numbers are for one material lot at one wall. Lot-to-lot yield variation in austenitic stainless moves the angle by 0.3–0.5°, which is exactly the band a ±0.5° drawing tolerance lives in.
Three compensation strategies, in increasing order of cost:
- Program an overbend. Measure one test bend, add the difference to every bend of that material and radius. Cheap, universal, and wrong the moment the lot changes.
- Closed-loop angle measurement. A laser or camera reads the formed angle on the part and the control corrects the next stroke. Modern CNC benders hold ±0.2–0.3° this way. This is the only method that tracks material variation in real time.
- Elastic compensation in the model. FE springback compensation writes the die surface slightly over-bent so the part lands on nominal. It works, it costs a simulation cycle, and it still needs a first-article check.
There is a fourth variable most drawings never mention: the weld seam. ERW tube has a longitudinal seam whose properties differ from the parent metal. If the seam sits on the extrados it is in tension and can crack on a tight bend; if it sits on the intrados it is in compression and can wrinkle. Rotate the tube so the seam lies on the neutral axis — the side of the tube, in the plane of the bend — and mark the orientation so every length is loaded the same way. A shop that reloads tubes with the seam in a random clock position gets a spread of springback that looks like machine drift and is not.
Cut length: the 25 mm that disappears at the apex
Drawings dimension tube two ways: to the tangent points, or to the intersection of the centre lines (the apex). Shops cut from a bend table, and the bend table works in tangent lengths. The conversion is one tangent function per bend:
At 90° that is R_c \tan 45° = R_c — a full radius vanishes from each leg. On a 25.4 mm CLR that is 25.4 mm per bend, which is exactly the kind of error that produces a part that is "25 mm short at both ends" and a customer convinced the shop cannot read a drawing.
The developed length — the length of tube to cut — is the sum of the straights plus the arc lengths along the centre line:
Worked: a 50.8 × 1.65 mm tube, one 90° bend at 76.2 mm CLR, legs dimensioned 300 mm each to the apex.
- Tangent length per leg: 300 - 76.2 = 223.8 mm
- Arc: \pi/2 \times 76.2 = 119.7 mm
- Cut length: 223.8 + 119.7 + 223.8 = 567.3 mm
If the drawing had been dimensioned to tangents, the cut length would be 300 + 119.7 + 300 = 719.7 mm. Same part, 152 mm of tube difference, and only one of the two is right. This is why a bend table — feed length, rotation, bend angle (LRA) — belongs with any bent-tube drawing, and why "dimension to the intersection of CLRs and end points" is standard advice in bending design guides.
Two more cut-length facts that change the number:
- Springback lengthens the part. The arc is formed to an overbent angle and springs back; the developed length is computed on the nominal radius, which drifts outward on unloading. The error is small — a fraction of a millimetre on a 90° bend — but it accumulates on multi-bend parts.
- Boost shortens it. Axial feeding pushes material into the bend zone, so the part grows or the straight sections shorten depending on how the machine controls the feed. Studies of pressure-die boosting treat axial elongation as a first-order effect, alongside thinning and springback, not as a rounding error.
Material behaviour that changes the numbers
Mild steel, DOM and ERW (300–450 MPa). The baseline. Wall factor under 20 bends without a mandrel at 3D; 1.5D needs a mandrel and usually a wiper. Springback 1.7–1.9° at 2D. ERW is cheaper and has a seam to orient; DOM (drawn-over-mandrel) costs more, holds the wall and the roundness better, and is the usual choice for tight bends and hydraulic service. Cold-drawn 1020 and 4130 chromoly bend well, 4130 with noticeably more springback and, in aircraft practice, a stress relief after forming.
304 / 316 stainless (515–620 MPa UTS). Two effects compound: higher flow stress means more springback (2.3–2.5° at 2D, 3–5° on tighter radii and harder tempers) and steep work hardening means the bend zone gets harder as it forms. Stainless galls against tool steel, so polished tooling and a proper bending lubricant are not optional on visible or thin-wall parts; a dry setup will pick up and tear. Austenitic stainless is also the grade where lot-to-lot yield spread is most visible in the angle.
Aluminium. The alloy and temper decide everything. 6061-T6 at 3D or tighter cracks; the standard remedies are to bend in T4 and age afterwards, or to anneal the bend zone, or to change to 5052-H32. The published minimum-radius arithmetic for aluminium tube, for fully tooled rotary draw bending, is:
with F = \pi/(4S) where S is the minimum elongation in percent, and tabulated factors for low-elongation alloys. For 38.1 mm OD × 1.5 mm 6061-T6, F = 0.0785:
That is 2.78D, which is why "3D minimum for 6061-T6" is the shorthand every shop uses. 5052-H32 has the same 10–11% elongation and therefore the same factor, so the temper — not the alloy family — is what the formula is reading. Drop to annealed 5052-O at F = 0.0491 and the same tube gives R_{min} = 47.5 + 29.9 = 77.4 mm, or 2.03D. Temper buys radius; alloy designation alone does not.
Titanium and nickel alloys. Grade 2 titanium bends cold with generous radii and 3.6° of springback at 2D; Grade 5 usually needs heat or a very large radius. Inconel 625 and 718 work-harden hard, spring back 4.5–5.4° at 2D, and need heavy tooling, slow draw speeds and often interpass annealing. Duplex and super-duplex stainless sit in the same family: high strength, high springback, and a TEMA thinning constant of 2 rather than 4 because the strain does not redistribute.
Wall condition. A mill tolerance on wall of ±10% is normal for seamless tube. A 1.5 mm nominal wall can arrive at 1.35 mm, and the bend then thins it to 1.01 mm at 1.5D. If the design has no margin for that, specify minimum wall, not nominal.
The setup that makes it repeatable
Bend Tooling's four-step procedure — forward mandrel, low pressure — is worth knowing because it inverts what most operators do. The four steps, in order, are: mandrel nose placement, direct pressure-die setting, wiper tip rake, and pressure-die assist. The order matters because each step assumes the previous one is correct; turning up the assist to hide a misplaced mandrel is how a shop ends up with high pressures, short tool life and inconsistent parts.
Three formulas from that procedure are genuinely useful at the quotation stage, because they tell you whether the tooling can be made at all:
Mandrel nose diameter for single-wall tube:
For 50.8 × 1.65 mm: D_n = 50.8 - 3.65 = 47.15 mm. The 2.21 constant is clearance, not a fit: too large a nose and the mandrel binds and marks the bore, too small and the section collapses ahead of the nose.
Mandrel nose radius: 0.1 D_n when the wall factor is under 50, 0.02 D_n above it. For our tube, 4.72 mm.
Nose set-up depth — how far forward of the line of tangency the nose sits. The mandrel is a straight cylinder inside a tube that is about to curve away from it, so the limit is where the mandrel's outside surface would break through the bore at the extrados:
For 50.8 × 1.65 at 1.5D: R_o - t = 101.6 - 1.65 = 99.95 mm and R_c + D_n/2 = 99.78 mm. The square-root term is \sqrt{9990 - 9955} = 5.9 mm, so S = 5.9 + 4.7 = 10.6 mm. The clearance that produces that number is a fraction of a millimetre — 0.17 mm on the radius here — which is why the depth has to be reset whenever the wall, the mandrel or the radius changes, and why it is step one of the setup rather than something the operator eyeballs.
Wiper feather edge thickness: the larger of 0.0025D and 0.15 mm. Too thick and the wiper cannot reach into the gap; too thin and the tip folds on the first bend.
Tooling is consumable. Mandrels and wipers wear in sliding contact, and a worn wiper produces exactly the defect a new one fixes. Shops that run thin-wall stainless keep wiper dies on a replacement schedule rather than waiting for wrinkles to reappear.
Pressure service: what the codes require
Bent tube in pressure service is not judged on appearance. ASME B31.3 requires that the wall thickness after bending be not less than the thickness required by the pressure design rules, and caps flattening at 8% of nominal OD for internal pressure and 3% at weld ends for external pressure. It does not hand you a thinning allowance: the starting wall has to be thick enough that the thinned extrados still passes.
ASME B16.9 long-radius elbows are 1.5D by definition, and short-radius elbows are 1.0D, which is why piping layouts default to those two numbers — the fittings exist, so the bend exists.
Heat-exchanger U-bends are the most tightly specified tube bends in common industrial use, because they are thin-wall, tightly radiused and carry pressure and thermal cycling at once. TEMA RCB-2.3.1 gives the starting-wall requirement directly:
with t_1 the minimum wall required for a straight tube at the same pressure and temperature, C = 4 for carbon and low-alloy steel, ferritic and austenitic stainless and copper alloys, and C = 2 for martensitic and duplex stainless, super-austenitic stainless, titanium and high-nickel alloys. Worked: 19.05 mm OD tube, t_1 = 1.2 mm, R_c = 28.6 mm (1.5D), carbon steel:
So 1.4 mm wall before bending guarantees 1.2 mm after — which is why U-tube bundles are ordered in 1.65 mm (16 BWG) rather than 1.2 mm. TEMA also caps U-bend flattening at 10%, and allows that cap to be exceeded for R < 2D in work-hardening material or when the straight wall is under D/12, subject to agreement. API 660 adds the radius floor: 1.5D for most materials, 2.0D for duplex, super-austenitic, titanium and high-nickel.
The shop-level acceptance numbers that follow from all this are stable across industries: extrados thinning up to 20% of the pre-bend wall, flattening 8% or better, and a documented ultrasonic reading at mid-arc on the extrados for anything in pressure service.
Defects, causes, and the setting that fixes them
Defect · Where · Immediate cause · Fix
Continuous wrinkling · Intrados, whole arc · Too little direct pressure-die pressure · Raise pressure to the chart value
Terminal hump or wrinkle at end of bend · Intrados, last 20° · Wiper tip not reaching tangency, or wrong rake · Reset wiper rake, check feather edge
Flattening / ovality · Whole section · Mandrel missing, too small, or nose too far back · Larger mandrel, move nose forward, add balls
Excess wall thinning · Extrados, mid-arc · Radius too tight for the wall, mandrel drag, no assist · Larger D of bend, thicker wall, raise assist
Tearing / split · Extrados · Material at its elongation limit; seam on extrados · Bend in softer temper, larger radius, rotate seam
Kink · Intrados, sharp · Radius below process limit · Mandrel + wiper, or redesign to 1.5D
Mandrel marks · Bore · Mandrel oversize, poor lubrication, worn balls · Re-size to D - 2.21t, lubricate
Chatter / witness marks · Pressure die side · Unstable pressure, dry tooling, worn die · Pressure and lubrication together
Angle drift between parts · All bends · Lot-to-lot yield, seam clocked randomly · Closed-loop angle control, seam orientation
The DFM checklist, with the numbers attached
- Design to a die the shop already owns. 1.5D, 2D and 3D are stocked; 1D and anything non-standard is a tooling charge and a lead time. Tooling is specific to one OD and one CLR.
- Keep 1.5D as the practical minimum. 1D is achievable with a mandrel and a wiper in thin wall, and it costs — in tooling, setup time and scrap rate.
- Use 3D or more for 6061-T6, or change the alloy. The R_{min} formula puts 38.1 × 1.5 mm 6061-T6 at 2.78D before any safety factor.
- Leave 1.5–2 × OD of straight between bends and at each end for the clamp and pressure dies. Less than that and the tooling cannot grip or back the tube.
- Dimension to tangent points, or supply a bend table (LRA). If the drawing dimensions to apexes, say so explicitly and expect the shop to subtract R_c \tan(\theta/2).
- Keep holes, welds and bosses out of the bend zone. The bend zone stretches and distorts anything in it; keep features at least 1.5 × OD from the tangent point.
- Specify wall as minimum, not nominal, where thinning matters. A ±10% mill tolerance on 1.5 mm is 0.15 mm that the bend will spend.
- State the ovality limit. 8% is the B31.3 internal-pressure number, 10% is TEMA, 5% is a common shop default, 3% needs a mandrel and care. Unstated means the shop picks.
- State the angle tolerance and whether it is per bend or end-to-end. ±0.5° per bend is standard CNC; ±0.25° needs closed-loop measurement; ±1° is manual.
- Specify the seam orientation for ERW tube, or accept random springback. Neutral axis, marked, every length.
- Avoid bends over 90° where the layout allows. Past 90° the mandrel and wiper have to negotiate the arc already formed, and tooling access degrades.
- Ask for the first-article numbers, not just the part: wall at mid-arc, ovality, angle, and the springback that was programmed. Those four numbers tell you whether the process is in control.
The cost arithmetic behind a bend
Tube bending is setup-dominated in the same way press brake work is, and more so, because the tooling set has five parts and the mandrel has to be threaded into each length.
Assume a CNC tube bender at ₹1,000 per machine hour, a 45-minute setup per program (tooling change, mandrel fit, first-article check on the CMM or a fixture), and 10 seconds per bend including load, index and unload:
Bends per part · Setup amortised · Run time · Cost per bend · Cost per part
2 · ₹750 · ₹6 · ₹378 · ₹756
6 · ₹750 · ₹17 · ₹128 · ₹767
20 · ₹750 · ₹56 · ₹40 · ₹806
100 · ₹750 · ₹278 · ₹10 · ₹1,028
The striking thing in that table is that the cost per part barely moves. A two-bend tube and a twenty-bend tube cost almost the same to bend, because the setup dominates both. This is why a complex bent tube is often cheaper than a designer expects and a simple one is more expensive than they expect — and why batch size, not bend count, is the variable that moves the quote.
Two costs sit outside the table. Tooling is a one-off per OD and CLR, and it is the reason shops steer designs toward their shelf. And scrap is disproportionate: a failed first article on a 1D bend in thin-wall stainless can consume more material than the run itself, which is why tight-radius work is quoted with a higher setup allowance.
Material is the other half of the part cost. A 25.4 × 1.5 mm 304 tube weighs 0.884 kg/m by 0.02466\,t(D-t); at ₹450/kg the material in a 600 mm part is about ₹240 — more than the bending. A 1D bend that forces a thicker wall, 1.8 mm instead of 1.5 mm, adds about 18% to the material bill across the whole part to save a radius on the drawing. Often the right answer is the larger radius.
What to send a fabricator
The package for a bent tube is short:
- A STEP of the formed part, with the centreline sketch.
- OD, wall (minimum, not nominal), alloy and temper — not "stainless".
- CLR per bend, or "shop standard 1.5D" stated explicitly.
- A bend table in LRA if you have one, or dimensions to tangent points.
- Angle tolerance and ovality limit, with the specification they come from (B31.3, TEMA, or your own).
- Seam orientation requirement, if the tube is welded.
- End preparation: cut, deburr, flare, bead, swage, notch, or weld prep.
- Batch size, and whether the first article is a separate approval.
Everything else — mandrel type, ball count, boost, wiper rake, springback compensation — belongs to the shop, and a shop allowed to own those decisions will hit the tolerance more often than one handed a radius and told to match it.
If you have a bent tube in STEP with a bend table, or a drawing you are not sure is dimensioned to tangents or apexes, upload it to FabFlow and compare quotes from fabricators who will tell you which of these ratios your design is fighting. A quote that names its die, its mandrel and the springback it will program is a quote you can hold someone to.