Gear racks, steering rack bars and square steel sections leave heat treatment bowed — and one face carries the teeth, the one surface a straightening press must never touch. Press it and a bruised flank turns a quiet rack into a noisy one before assembly. This guide explains why racks bend, why the square section changes support geometry and measurement (laser sensors scan the bar’s length instead of rotating it for TIR), and how a dedicated gear rack straightening machine corrects bars of 100 mm to 12 m across 5–600 mm equivalent stock while the teeth never meet a ram or support.
By SHANGDA Engineering TeamReading time: 12 minutesFor: rack makers & rack buyers
The golden rule for racks: the teeth sit on one flat face, and that face is never a working surface of the press. Lay the bar on its smooth back and a plain side with the tooth row pointing into open space, let laser sensors trace the bend along the length, and let the measured curve — not chalk or a straightedge — call every stroke.
A gear rack is a gear unrolled into a straight line: a long bar of square or rectangular section with a row of teeth milled, hobbed or ground along one face. That flat row turns a pinion’s rotation into linear motion wherever a carriage is driven — steering systems, elevators and lifting stages, gantry and machine-tool axis feeds, heavy actuators beyond the reach of screws or belts.
The same machines also straighten square and rectangular bar stock — the plain bar before teeth are cut, or non-toothed section steel for guide ways and frame rails — and the round-and-toothed steering rack shaft.
What sets the family apart from every shaft is the section and how bend is measured. A shaft is round: it turns on rollers and bend reads as runout around its circumference. A rack never turns, in service or in the machine; it is a flat-faced prism whose straightness is in-plane deviation of a face along the length — and the toothed face may never be touched. A rack machine is built around that prism: flat ways present the smooth faces to laser sensors and keep the tooth row clear of every contact, across 5–600 mm equivalent stock and 100 mm to 12 m lengths.
On a rack the teeth are the product. A bar can be bent and unbent on any other face, but a tooth row knocked about in the press cannot be undone — and the damage is not cosmetic:
A step in the track. A dent, flattened tip or corner mark on a flank makes the mating pinion strike it once per revolution — a rhythmic knock where motion ought to glide. In a steering gear it travels to the driver’s hands; in a feed drive it dominates the acoustic test.
Load piles onto two teeth. One bruised spot throws mesh force onto a couple of flanks, so the rack’s life runs out exactly where the press touched it.
Hardened flanks do not forgive. Carburised teeth dent or crack under a point load; tonnage the broad flat back absorbs without trace destroys a toothed face, and an over-hard stroke can split a tooth root before the mark is found.
So protection is first a question of which face carries the load — flat back, flat side, or round journal, never the row — with force size a distant second.
A bent rack is the normal output of heat treatment, not a process failure. Four mechanisms stack up (see our heat-treatment distortion guide for the full physics):
One-sided hardening. The toothed face is carburised for wear while the rest of the bar stays tough; a thin transformed layer grows on that face alone, and the prism bows.
Long and floppy. Bending stiffness falls with the cube of section height, so a residual moment a block would ignore curves a several-metre rack — often as an S-curve with two or three high points.
Mill stress in the stock. Rolled or drawn square bars carry frozen-in stresses; cutting teeth removes material asymmetrically, and the furnace relaxes the whole field as the bar cools.
Grinding and gravity. Flank grinding drags its own stressed skin across the bar, and a long hot rack on widely spaced supports droops in tempering. Hence the sequence rule: straighten once, after furnace and grinder have both had their turn.
What reaches the cell is a healthy tooth row on a crooked body — pitch and profile check out, but the bar points the wrong way. The fix flows through the smooth faces so the row stays as the gear-cutting machine left it.
The correction is ordinary three-point bending — two rests straddle a span, the ram loads the crown just past straight, and springback brings the bar to tolerance. What changes is the workpiece: a box, not a log, so the three working points are arranged around the box with the teeth pointing away:
Flat ways, not V-blocks. The bar beds on its broad smooth back over hardened ground ways, with a plain side against a location rail and the toothed face looking into open air. A resting face that doubles as the customer’s mounting datum rides on copper-, bronze- or polymer-faced pads.
The ram works from the smooth side. A crown on the toothed side is pressed from the flat back directly behind it — bending stress depends on section and span, not on which face the shoe kisses. The control holds a tooth-zone map and blocks any stroke whose shoe would land over flanks.
Capture the prism; it will not self-centre. A round shaft on V-rollers centres itself; a bar on a flat bed can sit a few degrees out of square or twist under load. Side clamps and anti-flip stops pin the orientation so the shoe lands full-width and never pries the bar out of its seat.
Two planes, two orientations. Bow across the tooth plane and bow along it are separate curves; the fixture indexes the prism between passes so each curve is bent back through its own smooth face.
Steering rack shafts split the difference: smooth round journals on rollers, the toothed middle segment hung clear of every rest.
Tooth row into open air, never on a pad · back flat on ground ways, plain side against the rail · side clamps and anti-flip stops engaged · soft shoes only on back or side, flank zones blocked · journals rest rack shafts · force ceiling per recipe. Any contact on flanks is a fixturing error: re-lay the bar.
A dial gauge on centres belongs to the round world: spin a shaft, watch the needle swing, and the swing is runout (see our TIR and runout guide). A rack never spins, and an indicator dragged over the teeth reads the tooth form, not the bend. Rectangular stock is read the way a surface plate reads a rule:
A laser head walks the bar. Non-contact laser displacement sensors (or electronic probes on a travelling carriage) traverse the smooth back, then a side face, while the bar lies still. Each station reports the gap to a straight optical reference; plotted against position, the gaps trace that face’s bend line — height and sign both, so a crown reads differently from a sag.
One back, one side — two bend lines. The back trace governs strokes that change mesh depth; the side trace governs the lateral curve a round part cannot even display. The whole press plan is built from both curves together.
Stations follow the bar. A short section takes 2–3 readings; long racks get 5–8 laser or probe stations, dense enough that an S-curve shows two apices instead of blurring into one.
Twist is caught on the same pass. Comparing a face near the two ends flags a wound section, which the machine rejects or queues separately — bending strokes cannot unwind twist, and chasing it with the ram only worsens geometry.
The first trace sets the correction sequence; the final trace certifies it against the bar’s serial number. Good depends on the job: 0.10–0.30 mm/m for general racks and structural sections, 0.02–0.05 mm for automotive and steering work, ≤0.02 mm for precise feed racks. Quote each figure with a basis — traced face, support layout, plane — as our tolerance primer shows.
Pressing never stretches the tooth row, so pitch and profile come out exactly as the gear-cutting machine left them. A rack can pass every tooth-table inspection and still fail — the table never asks what shape the bar is:
The pinion is dragged along a crooked road. Its carriage travels straight only, so a bowed bar feeds toward it and away by the amount of bend: backlash yawns open on retreating sections and clamps shut where the bar bows into the pinion.
The driver feels noise and patchy effort. Steering racks show heavy and light patches and slow self-centring; feed drives show fluctuating torque, chatter and repeatability that drifts down the stroke — despite an immaculate pitch chart.
Lateral bow breaks the mounting. A curved side face cannot sit flat on its rail; bolting it down drags the bar straight by force, so the driven axis’ guides and seals wear one-sidedly.
Face straightness keeps the gear grinder’s accuracy alive after assembly — which is why rack prints carry a straightness callout beside the pitch class, and why the press is a planned operation, not a salvage booth.
Every bar runs the same four-beat cycle — the scan-press-rescan rhythm our five warning signs article shows hand pressing can never reproduce:
Scan. Sensors walk the back face and a side face — 2–3 stations on short sections, 5–8 on long bars — with the bar motionless on the ways; the control plots two bend lines from the gaps.
Decide. Each apex is located in its own plane; the control picks rest spans and ram stations on smooth faces only and works out how far past flat to drive each stroke. Carburised racks rebound hard and rebound shifts between heat lots, so over-drive follows the bar’s measured answer, not a dial set by habit (our springback guide unpacks why).
Press. The shoe lands on a flat face under a force ceiling — the ceiling is what keeps a hard case from opening at a tooth root.
Re-trace and close in. Both faces are scanned again and a stroke is offered only where the curve still demands it; the control tunes its rebound estimate as the bar converges. Cycle time is 20–90 seconds per bar — 5–10 times the pace of chalk-and-straightedge pressing.
A recipe per part number remembers rest positions, scan stations, blocked flank zones, force ceilings and the straightness target; every bar leaves with incoming and outgoing curves on the batch record — 100% inspection and 100% logged data by default. Two builds bracket the range: the laser-detection gear rack straightening machine for dense multi-station scanning, and the gear rack straightening machine with laser sensors for standard rack and square-steel production.
Rack straightening and screw straightening solve the same problem — correcting a bent axis while sparing a precision surface — but almost every physical decision differs, because one part is a rotating cylinder with a helix and the other a static prism with a flat tooth row.
| Aspect | Round threaded parts (ball screws, lead screws) | Square / rectangular toothed bars (gear racks) |
|---|---|---|
| Section & posture | Cylinder; turns on V-rollers | Prism; beds flat on ways, never turns |
| Protected surface | Helix wrapped around the part | One planar tooth row |
| Rests | V-blocks / rollers on round journals | Ground flat ways + side location rail |
| Ram contact | Soft shoes on smooth round sections | Wide flat shoe on the back or a side |
| Measurement | Indicator swing as the part rotates (TIR) | Laser gap trace along two faces, no rotation |
| Curves captured | Runout envelope in every direction | Two bend lines, in two planes |
| Additional risk | — | Twist of the section |
Hence a screw cell and a rack cell are different machines, even when both pledge a pristine working surface. Round fixturing cannot find a prism, and runout says nothing about a stationary bar. Rack lines are engineered around rectangular stock, with 1000-ton class frames for heavy sections, within 5–600 mm equivalent stock and 100 mm to 12 m lengths.
Hand pressing will rescue one bent bar for a careful fitter; it cannot run a rack batch, and its failure modes are specific to the box section:
The natural push point is the teeth. The furnace-bowed hump is most often visible on the toothed face, so the obvious stroke lands a block on the flanks — and loose prop blocks underneath are just as likely to rest on tooth tips. Either way the row takes the tonnage.
Loose blocks breed twist. A prism sitting a degree askew on improvised packing takes a stroke that partly winds the section instead of bending it. End-on it looks straighter; face to face it is twisted, and twist cannot be pressed back out.
One push too far is invisible. Carburised flanks accept almost no plastic strain; an over-stroke splits tooth roots with cracks the eye never finds, and a plain press answers to no force ceiling.
Half the bend is never checked. A straightedge on the back finds nothing sideways and nothing about twist, while a several-metre S-curve needs a placed correction sequence no hand routine repeats across a shift.
Nothing is recorded. Straightedge impressions and chalk arcs die with the operator; a buyer asking for straightness evidence per bar gets a shrug.
Long racks are awkward for a human and ideal for a robot: a rectangular section always presents the same orientation, and the tooth row gives the gripper a foolproof reference. A loader plucks each bar from the in-feed rack, lays it on the ways teeth-out in one determined orientation, and the cell runs the whole sequence — trace, decide, press each plane, re-trace, discharge — untouched. The gear rack straightening machine with robotic loading is the high-volume configuration of the range.
Staffing changes shape. One worker oversees 1–2 CNC machines when bars are hand-fed, or a row of linked cells where the only human tasks are feed replenishment and spot checks.
The recipe, not the person, sets the work. Identical scan stations, spans and force ceilings run bar after bar across every shift; results no longer vary by operator, and no know-how leaves when a fitter does.
The arithmetic follows the pace. At a 20–90-second cycle, 5–10 times hand throughput, a cell earns its cost back often within 1–2 years with steady volume: less loading labour, no bruised-rack scrap, and every shipment with measured bend lines.
Whether a given rack will reach tolerance with its teeth untouched can be judged from data already in your technical file. Send the following and the machine specification largely writes itself:
Section drawing and length — square or rectangular dimensions, overall length, weight, and end features such as round journals. The standard envelope is 5–600 mm equivalent stock at 100 mm to 12 m lengths.
Steel and heat-treatment state — case depth and hardening process on the teeth, through-hardened or plain body stock; which faces leave the grinder hard and which arrive as-machined.
Tooth layout — which face carries the row, tooth size, and journal or smooth-segment positions; a rough cross-section sketch fixes the fixture design.
Straightness target with its basis — the face to be traced, the support layout and the plane; steering bars usually land in the 0.02–0.05 mm band, precision feed racks at ≤0.02 mm.
Yearly volume and batch sizes — this single input decides between a stand-alone CNC press and a robotic-loading line.
Plant utilities — floor length available for the bar run, 380–480 V three-phase supply, and any data-logging or line-interface requirements.
Drawing questions end with physical bars: a trial on your own steel demonstrates rebound and returns verified before/after curves from our floor. Custom gear rack straightening machines are built to the part and ship within 60–120 days.
Not on a purpose-built rack machine: the bar rests on its plain back and side, the ram presses only those plain faces opposite the measured high point, tooth zones are locked out of the program, and every stroke is force-capped. Damage happens on general-purpose presses, where a prop block sits under a tooth tip or the ram comes down on the visible high point — the toothed face; one bruised flank means noise, uneven loading and rejection. No. TIR is a rotational reading for round parts; a rack does not rotate, and a probe across the teeth reads the tooth form, not the bend. Laser sensors or electronic probes scan the plain back and a side face along the stationary bar — 2–3 stations for short bars, 5–8 for long racks — producing a bending curve in each plane: the back curve covers bow that changes mesh depth, the side curve reveals lateral bend a one-plane check misses. No. Pitch and profile are fixed at the hobbing or grinding stage; pressing bends the bar locally and never touches the flanks. What erodes accuracy in practice is leaving the curve in place: a crooked track makes the pinion climb and dip, backlash varies along the stroke, steering feel turns patchy and feed drives chatter. The press defends the accuracy the grinder delivered. One-offs with a very careful fitter, yes; batches, no. The natural push point on a furnace-bowed bar is the toothed face itself, improvised packing leaves blocks under tooth tips or winds the section, strokes without a force ceiling can split hardened roots, side bow and twist never enter the straightedge check, and nothing is recorded — while a several-metre S-curve demands a placed correction sequence. An automatic cell traces, presses and re-traces each bar in 20–90 seconds, 5–10 times the hand pace, with curves filed per piece. General racks and structural sections live with 0.10–0.30 mm/m; automotive steering bars and comparable precision work land in the 0.02–0.05 mm band; the tightest machine-tool feed racks call for ≤0.02 mm. State the figure with its basis — traced face, support layout and plane — so acceptance cannot be argued over and the lateral trace stays on the inspection sheet. Yes, and with fewer constraints: with no tooth row to protect, rests, shoes and scan traces use whichever faces fit the bend. Plain bars are corrected before and after teeth are cut, and the heaviest sections run on frames up to the 1000-ton class. Put the section, length and steel family on the inquiry and the cell is configured around the stock.Does straightening damage the gear rack teeth?
How is straightness measured on a rack — is it TIR?
Does straightening change the tooth pitch or tooth profile?
Can a hydraulic shop press straighten long gear racks?
What straightness tolerance should a straightened rack hold?
Can the laser machine also straighten plain square and rectangular bar stock?
Since 2008, SHANGDA has built CE-certified automatic straightening cells for racks, bars, shafts, sections and tubes — 5–600 mm equivalent stock and 100 mm to 12 m lengths — including laser-scanning lines for gear racks, steering rack bars and square steel that load the prism on flat ways and press only through smooth faces. Send your section drawing, heat-treatment state and straightness target; our engineering team will lay out the fixturing, scan stations and loading option and prove the result on your sample bars, with before/after bend curves on every piece.
Ask for a Free Rack AssessmentSee Gear Rack Straightening Machines