SITE：Bent or Twisted? How Automatic Straightening Handles Torsional Distortion | SHANGDA
A square bar can fail “straightness” in two completely different ways. Its centerline can curve — a bow that a press removes by bending the high point back. Or its cross-sections can rotate around the axis — one end turned relative to the other, the bar wound like a barber pole while its centerline stays perfectly straight. That second defect is twist (torsional distortion), and no straightening press can remove it: you can press a twisted bar dead straight in every plane and it still rocks on three corners. This guide separates the two — how torsion forms in rolled, drawn and heat-treated profiles, how it is measured, why press force cannot touch it, and how a twisting straightening machine grips both ends and applies controlled reverse torque, often with bend correction in one automatic cycle.
By SHANGDA Engineering TeamReading time: 12 minutesFor: bar, profile & rack makers, quality engineers
The one rule behind this article: a press applies force perpendicular to the axis, so it corrects the centerline. Twist is an angular error around the axis and is removed only by torque — gripping the section at two stations and rotating one end back past its elastic limit, then re-measuring. Match the defect to the axis of correction: force for bend, torque for twist.
01Two defects hiding behind one word: “not straight”
Drawings, inspection reports and shop-floor conversation all use the same phrase — “the bar is not straight” — for two geometrically independent errors:
Bend. The line through the centres of successive cross-sections departs from a straight line. The part curves — a bow, a local kink, or an S-shape that wanders between planes.
Twist (torsion). The centerline may be perfectly straight, but each cross-section is rotated around it. Hold a square bar at one end and turn the other: every intermediate square has rotated a little, and the cumulative angle grows along the length.
The two are independent: a bar can be dead straight yet wound end to end, severely bowed yet completely free of torsion, or — the usual case for heat-treated non-round stock — both at once. They share overlapping causes but are measured against different references and removed by physically different processes.
Confusing them is expensive: twist does not respond to press strokes. A workshop that presses a twisted rack and sees the same corner-rocking afterward concludes the press is too weak — when no press on earth can rotate a section by pushing on it. It mirrors the tube-side mix-up of curved centerline vs. out-of-round section (our ovality vs. centerline guide): there the hidden defect was section shape; here it is the section’s angular position. The rest of this article ensures your RFQ names the defect you actually have.
02What a bend is: the centerline leaves its line
Bend is the defect almost every straightening-machine article describes. The centerline is the line through the centroid of each cross-section; bend is the lateral distance between that line and an ideal straight axis, quoted as a total deviation or as a per-metre rate on long stock.
It arrives in familiar shapes: a smooth global bow in one plane, a sharp local kink near a clamp or weld, and a multi-plane S-curve whose high points move around the clock as the part rotates. It is detected by resting the bar on V-blocks or support rollers and reading dial or laser probes at stations along the length while the part turns — the TIR/runout method that gives every high point both a magnitude and a clock position.
And it is corrected by one well-understood mechanism: three-point bending. Two supports define a span, the ram pushes the high point between them slightly past straight, the material yields locally, and after elastic springback the axis lands on target. The bending moment rotates material around an axis perpendicular to the bar — which is exactly why a press is the right tool for this defect and, as Section 6 shows, the wrong tool for the other one.
03What twist is: the section rotates around the axis
Twist is torsional distortion: successive cross-sections are rotated relative to one another about the longitudinal axis. The axis itself need not move at all. A twisted square bar looks straight until you look closely — then the signs are unmistakable:
On the surface table. An angle iron or channel rests on three corners while the fourth rocks; a square bar laid on a face refuses to lie flat and rolls toward an edge; flanges lift at diagonal corners rather than along a bow.
End to end. Sight along the bar: an edge that starts at twelve o’clock at one end finishes at a different clock position at the other. The deformation is a gradual helical wind, not a bump.
On toothed parts. A rack’s tooth plane spirals gently, so teeth at one end face a slightly different direction than teeth at the other.
The functional consequences follow that rotation directly:
Racks drive a pinion along their whole length; a wound rack walks the contact pattern across the tooth face, producing noise, uneven loading and binding even though the axis is straight.
Angles, channels and structural profiles will not sit flat, will not weld up square, and bolt holes drift as fabricated frames twist in assembly.
Spline shafts and screw blanks carry keys, splines and nuts that must keep one angular reference; a spiral substrate rotates that reference under them.
Square and flat bars jam feeders and fixtures that expect a fixed orientation: the bar enters square and arrives turned.
Acceptable wind is set by your drawing — an allowed angle over a gauge length, or an edge-height difference — not by a generic figure.
04Why bars and profiles arrive twisted in the first place
Bend and twist share one root family — residual stresses rebalancing — but twist only appears when the section or the process is asymmetric. Four mechanisms do most of the damage:
Asymmetric rolling and drawing. A round section is rotationally symmetric: unequal stress in its longitudinal fibres can only relieve itself by bowing. A square, flat, angle or custom profile has no such symmetry — reduction and cooling differ face to face and corner to corner — so the same imbalance relieves itself partly by winding. Non-round mill stock carries twist as a matter of course.
Heat treatment and quenching — the great twist-maker for profiles. Flats and corners cool at different rates; transformation proceeds unevenly around the section, and the part winds as stresses rebalance. Racks are the textbook case — teeth one side, a solid web the other, almost no symmetry at all — which is why hardened rack stock arrives bowed and twisted together.
Uneven cooling and stacking. Long, slender, non-round stock cooling on poorly spaced supports, or stacked hot under its own weight, cools under a torsional bias and the wind freezes in.
Machining asymmetry. Milling teeth, cutting a keyway or machining stock from one side releases a stress field balanced in rotation; the part twists as it relaxes.
The lesson matches the bend world: incoming twist reflects the supplier’s whole process — two heats can arrive wound very differently — so an automatic cell measures every bar. And in reverse: round shafts are rotationally symmetric and rarely twist. Twist is above all a profile, flat, toothed-bar and non-round-section problem.
05How twist is measured on the shop floor
Twist is an angular difference between cross-sections at different axial positions, so measurement must establish the orientation of the section at several stations — not just its height. The bar rests in defined V-blocks or support rollers, the same fixturing discipline used for bend, then:
Square and flat bars. Dial gauges, LVDT or laser probes read the height of both edges at a station. With the section width known, the edge-height difference converts directly to a section rotation; comparing rotations between stations gives the twist over the span.
Angles and channels. Probes read both flanges (or each flange tip’s gap to the table); diagonal lift appears immediately as a rotating section rather than a curving axis.
Racks. The tooth-bearing plane — or a machined reference edge on the back — is probed at successive stations; the orientation of that plane along the length is the wind.
Complex extruded profiles. A laser or probe pair references two datum features of the section at each station, tracking their angular position down the bed.
As with bend mapping, short stiff parts need 2–3 stations; long bars get a 5–8 probe scan, producing a twist map — section angle versus axial position, the torsional twin of the deflection curve. A rotating TIR pass cannot report twist: on a non-round section the needle follows the section form, so twist needs a fixed datum — the table, or paired edge probes. Report results per your drawing — angle over a stated gauge length, stations named; we quote no generic twist figures, because acceptable wind between a structural angle and a precision rack is a chasm.
06Why a straightening press cannot press twist out
A press — manual, hydraulic or CNC — does one mechanical thing: it applies a force perpendicular to the axis at a chosen point, creating a bending moment in a plane. That moment rotates material around an axis perpendicular to the bar and curves the centerline. Everything — ram, supports, stroke, force cap — serves that one action.
Twist is held by residual shear stresses around the axis; releasing it needs a torque — a moment whose axis is the bar’s longitudinal axis. No arrangement of vertical ram force produces it: pressing a wound bar bends it, but the section rotation at the load point is unchanged. A profile pressed dead straight in every plane still rocks on the same three corners.
The shop-floor signature is diagnostic: edge-height and flange-gap readings do not move, stroke after stroke, while bend readings improve; the operator presses harder, the twist stays, marks appear. Torque cannot be improvised from above — it requires the part gripped at two separated sections and one end rotated relative to the other, past the elastic torsional limit. That is a different machine on a different axis of correction — the same hard lesson as the tube case, where section shape needed rolls rather than a bigger press.
07How a twisting straightening machine removes torsion
A twisting straightening machine replaces ram-and-supports with two clamping heads along the bed. Each grips with jaw sets profiled to the part — square, flat, angle, channel, rack-back or custom — with soft faces protecting machined surfaces. One head holds; the other rotates through a controlled angle about the bar’s own axis. The cycle is the torsional cousin of press straightening:
Measure. The twist map (Section 5) locates where the wind is distributed.
Compute. The control sets the clamp stations to flank the wound zone and calculates the corrective rotation — including an over-angle: torsional correction has springback just as bending does, and the elastic share varies with material, hardness and section. The over-angle comes from measured response, refined on recheck — the springback compensation principle applied to torsion.
Torque. The rotating head winds the section past its elastic torsional yield, holds, releases.
Re-measure and repeat at successive stations — a distributed helical wind is worked out zone by zone, just as an S-curve is corrected lobe by lobe.
Gripping follows the same protection logic as screw work: racks clamp on the web or back, never on the teeth; finished faces meet soft profiled jaws; torque is capped per recipe. Profile-dedicated builds — the twisting straightening machine for profiles — use wider jaws and open-section fixturing. The family covers sections from Ø5 mm to 600 mm and 100 mm to 12 m, longer bars in sections.
08When a part needs both: twist first, or bend first?
Most heat-treated non-round stock arrives with both defects — the rack bar bowed and wound, the angle curving and rocking. The corrections interact, so three rules follow.
The sequence is part-dependent. A common order is twist-first, bend-last — leaving the axis correction last so the final measured number is straightness — but lightly wound parts may be bent first, and some bars need alternating passes with measurement between. The recipe follows the measured maps, not a fixed rule.
One fixturing beats two. Correcting twist on a twisting machine and then bend on a separate press means two setups, two datum chains — and each operation can quietly undo the other. Integrated cells put a press station and twisting heads on one bed, one control: measured once for both defects, corrected in a programmed sequence, re-measured until both converge.
Racks and square bars are the classic integrated application — hardened, asymmetric, almost always bent and wound. Our rack and square-bar straightening cells are built around that combined cycle.
The economics are the same argument that wins for bend automation: one handling, one datum chain, one converging loop.
09The parts that arrive twisted — and the ones that do not
Section symmetry predicts the defect: round stock can only bow; anything with a face, flange or tooth is a twist candidate.
| Workpiece | Section character | Typical defects after heat treatment | Correction notes |
|---|---|---|---|
| Square / flat bar | Non-round, milled faces | Bend + twist | Twisting heads plus press station; edge-height twist mapping |
| Angle iron, channel, structural profiles | Open, thin-walled, asymmetric | Twist dominant — diagonal rocking | Profile twisting machine; wide profiled jaws; check section rigidity (Section 11) |
| Rack bars | Teeth one side, web the other | Bend + twist, always together | Grip the web, never the teeth; integrated twist-and-press cell typical |
| Screw blanks, spline shafts, profile shafts | Helical or keyed features | Wind that rotates feature orientation | Torsion correction on the blank or between centres; protect finished features |
| Custom extruded / drawn profiles | Irregular section | Twist from asymmetric cooling | Custom jaw sets; paired-probe or laser datum tracking |
| Round bars and shafts | Rotationally symmetric | Bend only in practice | Press straightening; twist correction rarely required |
Inspection expectation: for the non-round families, “straightness” means two numbers — axis and angle — or a perfectly centered bar still rocks on the inspection table.
10The automatic closed loop with two defect maps
Manual twist correction — clamping to a floor plate, heating, winding with a wrench and judging by eye — still exists in repair shops, but it is slow, unrepeatable and record-free. An automatic cell runs the bend loop with torsion added as a second measured dimension:
Measure both. Probes map the deflection curve and the section-angle map in one fixturing — 2–3 stations on short bars, 5–8 probes on long ones.
Compute both. Press points and over-travel for bend; clamp stations and over-angle for twist; sequence from the part recipe.
Correct with limits. Force-capped press strokes and torque-capped rotations in turn; nothing relies on an operator’s feel.
Re-measure both and repeat until straightness and twist are inside target — the measure–compute–correct–recheck discipline from our working-principle guide, closing two loops at once.
The numbers follow the production family: 20–90 seconds per bar, typically 5–10 times faster than manual, with one operator tending 1–2 CNC machines or several linked automatic units; straightness lands in the usual bands — 0.10–0.30 mm/m general stock, 0.02–0.05 mm automotive and precision drive parts, ≤0.02 mm precision components — twist judged per drawing; and every bar ships with 100% inspection and 100% logged data for both defects. With steady volume the investment pays back often within 1–2 years with steady volume.
11Honest limits and your twist-straightening RFQ list
Torsional straightening yields the section in shear, so the part needs enough ductility and rigidity to rotate without damage. Three honest boundaries:
Very thin open sections — light-gauge angle or thin channel — can distort in section under torque rather than wind back, or crease a flange at the jaws; wide fixturing helps, but some gauges are better corrected at the mill.
Brittle, fully hardened complex sections may crack in torsional yielding; torque caps and light incremental corrections reduce but do not erase the risk.
Wind beyond the angular envelope, or bars cracked in quenching, are mill-side problems, not straightening problems.
Because the boundary is part-specific, the answer is a trial on your samples: wound bars from your own heat-treatment lot come back with before/after straightness and twist data you can verify. Send:
Your twist-straightening RFQ checklist
Section drawing — square, flat, angle, channel, rack or custom profile, with all section dimensions; shape decides jaw sets and torque. Envelope: Ø5–600 mm sections, 100 mm–12 m lengths.
Length and batch pattern — individual bars or bundled.
Material and heat-treatment condition — as-rolled, as-drawn, quenched, tempered; hardness drives torsional springback and crack risk.
Both tolerances — straightness with its basis (value + stations), and twist per the drawing (angle over a gauge length, or edge-height difference, stations named).
Incoming condition — measured bend and twist on real stock, so the envelope is sized to your worst bars.
Surfaces that must not be marked — teeth, machined datums, plated or ground faces.
Volume and line conditions — pieces per shift, 380–480 V three-phase supply, floor space, data-export needs.
Custom twisting and combined cells are built to order and typically deliver in 60–120 days, CE certified, by a manufacturer building straightening equipment since 2008.
12Frequently asked questions
No — the central point of this article. A press bends the centerline; it cannot rotate a cross-section. Twist is held by residual shear stresses around the axis and is released only by torque: gripping the bar at two stations and rotating one end past its elastic torsional limit. A twisted bar pressed dead straight still rocks on three corners. Twist needs a twisting straightening machine, not a bigger press. Lay it on a surface table. Bend shows as a gap following a bow that changes as the part is rolled; twist shows as rocking on corners, diagonal flange lift, or edges at the two ends sitting at different clock positions while the centerline looks straight. A square bar that rolls off its face, an angle resting on three corners, or a rack whose tooth plane spirals along its length is twisted. Confirm with paired edge-height probes at several stations rather than judging by eye. Rest the bar in defined V-supports and, at each axial station, measure the height of both edges or flanges with dial gauges, LVDTs or laser probes. Across a known section width, the edge-height difference converts to the section’s rotation; the change in rotation between two stations is the twist over that span. Short bars need 2–3 stations and long bars a 5–8 probe scan, building an angle-versus-position map. Accept against the drawing — a permitted angle over a stated gauge length or the equivalent edge-height difference, with stations named. The two corrections share one residual-stress field and can nudge each other: bending shifts the stresses holding twist, and torsional yielding can move the axis slightly. Integrated cells handle this by measuring both defects in one fixturing, correcting them in a programmed sequence — often twist first and bend last, recipe-dependent — and re-measuring both until each reaches target. Running them on separate machines risks the operations undoing each other. Any non-round section that arrives wound: square and flat bars, angle iron and channel, structural and extruded profiles, rack bars, screw and spline blanks, and profile shafts with keyed or helical features — especially after quenching, where asymmetric sections wind as they harden. Racks are the classic case because the toothed section is almost perfectly asymmetric. Round bars and shafts are rotationally symmetric; twist in round stock is rare and press straightening normally covers them. Most production bars and profiles, yes — but torsional correction yields the section in shear, so the material needs ductility: very thin open sections can deform or crease at the jaws, and brittle fully-hardened sections can crack; wind beyond the angular envelope or pre-cracked stock is not straightenable. The reliable answer is a trial on your sample bars, returning before/after twist and straightness data measured on your own metal.Can a regular hydraulic press straighten a twisted square bar?
How do I tell on the shop floor whether a bar is bent or twisted?
How is twist angle actually measured?
Does correcting twist change straightness, or vice versa?
Which workpieces need a twisting straightening machine?
Can every twisted bar or profile be salvaged?
Send your profile drawings — we will check bend and twist
SHANGDA has built automatic straightening machines since 2008, all CE certified — including twisting straightening machines and combined twist-and-press cells for square and flat bars, angles, channels, racks and custom profiles from Ø5 mm to 600 mm sections and 100 mm to 12 m. Send drawings, heat-treatment condition, straightness and twist callouts and incoming measurements; we will assess straightenability honestly and prove the result on your samples with before/after data for both defects.
Send Your Profile Drawings for a Free AssessmentSee Twisting Straightening Machines
