H Shaped Steel Cutting: How the Machine Actually Cuts

• 1
  Why h shaped steel cutting is a 3D problem, not a 2D one
• 2
  Thermal cutting versus mechanical cutting on an H section
• 3
  What the machine has to control to cut an H profile accurately
• 4
  Calibration points that change the cut result
• 5
  Tolerances, edge quality and what the joint actually needs
• 6
  Process comparison for h shaped steel cutting

A flame or plasma torch cutting an H beam mostly works in one plane. The head travels along the web or over a flange edge, and the profile shape comes from a template or a stored program. That works well for straight cuts and simple copes. It falls apart the moment the cut has to pass through both flanges and the web in a single continuous path.

An H section has two parallel flanges joined by a thin web. The web is often only 6 to 12 mm thick, while the flanges may be 15 to 30 mm. That thickness ratio matters. Heat that soaks into a 30 mm flange will warp a 6 mm web if the cut sequence is wrong. The beam twists, the kerf closes behind the torch, and the finished cope no longer fits the mating plate.

The third dimension is access. A torch can approach from above, from the side, or at an angle, but it cannot reach behind a flange that is already in the way. A milling spindle with a long reach tool can. That is the basic reason h shaped steel cutting on a machining center costs more per meter but holds geometry a thermal process cannot.

So the first question is never which machine is faster. It is whether the cut path is open or closed, whether the section is symmetric, and whether the joint has to bear load.

• 1
  Open cutTorch or plasma is usually enough.
• 2
  Closed copeThe path wraps around a flange, so tool access decides the process.
• 3
  Thin web, thick flangeHeat input must be limited or sequenced.

Oxy-fuel cutting burns steel at roughly 1,200 to 1,500 °C in the kerf. The heat-affected zone on a 20 mm flange can reach 1 to 3 mm deep, and the edge hardness rises. On A36 or 1018 that is usually acceptable. On 4140 or 4340 the same edge can crack during cooling if the beam is not preheated.

Plasma cuts faster and puts in less total heat, but the kerf is wider and the edge has a slight bevel, typically 2 to 5 degrees depending on the nozzle and standoff. For a bolted connection with slotted holes that bevel is harmless. For a cope that has to slide into a mating member with 0.5 mm clearance, it is not.

Abrasive waterjet avoids heat entirely. It holds a near-square edge and cuts any hardness. The trade-off is speed: a 20 mm steel section cuts at roughly 100 to 150 mm per minute, and the abrasive stream tapers slightly through the depth. It is a good fit for one-off beams and for materials that cannot take heat.

Milling removes material with a rotating cutter. There is no heat-affected zone beyond a thin surface layer, the edge is square within the machine tolerance, and the same setup can drill, tap and counterbore the connection. That is why structural parts that need to assemble without field fitting usually end up on a mill.

• 1
  Oxy-fuelCheap, thick sections, wide HAZ, needs preheat on alloy steel.
• 2
  PlasmaFast, 2–5° edge bevel, good for slotted connections.
• 3
  WaterjetNo heat, near-square edge, slow on thick steel.
• 4
  MillingSquare edge, drilled and tapped in the same setup.

Three things decide whether h shaped steel cutting hits the drawing. The first is workholding. A 12 m beam is not stiff in torsion. If it is clamped only at the ends, the middle sags and the cut depth varies along the length. Support rollers every 1.5 to 2 m keep the beam straight, and the clamps should sit close to the cut, not far from it.

The second is thermal compensation. The ball screws and the beam itself grow as they warm up. A machine that runs a 4,000 mm part for two hours will drift if the control does not compensate. On a warm-up cycle of 20 to 30 minutes the axis reaches a stable length, and the first article should be measured after that, not before.

The third is tool reach and rigidity. Cutting the inside face of a flange means the cutter hangs out from the spindle. A 20 mm diameter end mill at 4× diameter reach will deflect under load. Depth of cut should drop to 0.3 to 0.5 mm per pass at that reach, and the feed has to come down with it. Long-reach tools are for finishing, not for roughing out a 30 mm flange.

None of this is exotic. It is the same set of rules you would apply to any long, thin, heat-sensitive part. The difference is that the part is also the fixture, so any error in support shows up directly in the cut.

Squareness between the spindle axis and the work table is the first check. On a large gantry or bridge mill, a 0.02 mm per 300 mm error becomes 0.27 mm over a 4,000 mm cut. That is enough to open a joint. The check is a dial indicator sweep on a granite square, and the correction is in the machine parameters, not in the program.

Backlash on the axis that moves along the beam length is the second. On a worn rack-and-pinion drive, reversal error of 0.05 mm shows up as a step every time the cutter changes direction. A bidirectional test with a laser interferometer or a ballbar finds it in minutes. Adjusting the preload or the compensation table fixes most of it.

Tool length and diameter offsets are the third. A 0.01 mm error in the length offset changes the depth of a shoulder by the same amount. After any tool change the first part should be measured at the feature, not at the datum. If the offset is wrong, the error repeats on every part in the run.

The last point is thermal drift during the run. Log the axis position at the start and the end of a long cut. If the difference is more than 0.01 mm, the machine needs more warm-up or active compensation before it is trusted with a ±0.005 mm feature.

• 1
  Spindle squareness0.02 mm per 300 mm is already 0.27 mm over 4,000 mm.
• 2
  Axis backlashShows as a step at every direction change.
• 3
  Tool offsetsVerify at the feature, not the datum.
• 4
  Thermal driftLog start and end position on long cuts.

Not every H beam joint needs ±0.005 mm. A bolted splice with 22 mm holes for M20 bolts has 2 mm of clearance built in. Cutting it to ±0.5 mm is wasted money. A welded cope has even more room, because the weld fills the gap. The tolerance should come from the joint, not from the machine's best number.

Where tight tolerance pays off is in assembled structures that are not welded, in machinery frames that carry a linear rail, and in parts that act as a datum for other components. In those cases the flange face, the web face and the hole pattern all have to sit in one coordinate system. Cutting them in one setup on a machining center removes the stack-up that comes from cutting, then drilling, then fitting.

Edge quality follows the same logic. A thermal cut leaves dross on the bottom edge and a slightly rounded top corner. For a painted structural member that is fine. For a sliding fit or a sealing face it is not. Milled edges on 1018 or 4140 come off the tool at Ra 1.6–3.2 μm as machined, and down to Ra 0.8–1.6 μm with a finishing pass.

Surface finish numbers are only useful if the drawing states them. If it does not, the shop will default to as-machined, and that is usually the right call for a structural part.