Nov 02, 2026
Floating Grinding, Rigid Cutting, and Compliant Contact: Why One Force Setting Cannot Finish a Casting
The shop-floor habit that gets the physics backwards
When a foundry looks at a grinding cell for the first time, two numbers usually dominate the conversation: arm payload and spindle power. Bigger looks safer. After commissioning, the program often fails for the opposite reason. The machine had enough force. It applied that force in the wrong place.
Cutting a remaining iron gate wants the tool held in, feed held steady, wheel or mill actually biting. Run the same press into a thin corner of an aluminum oil pan and the wall yields before the path ends. Waves show up. Local over-grind follows. A grinder with an angle grinder in hand backs off without thinking. A robot does not, unless that backing-off is written into the control.
So at least three contact states have to be kept separate: rigid cutting, floating grinding, and compliant (torque-limited) contact. They sound like brochure features. On the floor they are three different ways the tool meets metal.
Rigid cutting is for stock that still has to be removed
Gate stubs, heavy flash, local swell: the material is still there, and it is often thicker and harder than the casting body. The job is not to polish a surface. It is to take a lump off. The path has to stay put. The spindle has to bite. Table and arm cannot fold when the cut starts.
That is why a builder such as Anhui Neview talks about stiffness and payload together. On a DR-class iron cell, designing table and arm around a 200 kg class load is not a round number for a catalogue. Gate-root cuts can push contact force into a high measured range. If the structure is soft at that point, the path drifts, the wheel chatters, and consumable life collapses.
Rigid cutting has limits. If the locating scheme rocks, or the clamp only holds one corner, a heavy cut will walk the part in the fixture. If the stub sits next to a machining datum, an uncontrolled depth takes tomorrow’s stock. Those zones want rigidity. They do not want a floating head with its full stroke unlocked — extra float makes depth a guess.
Float keeps contact. It does not hand accuracy to chance
A floating spindle gives the tool a short allowed travel so it can stay on a surface that is not the CAD surface. Draft, shrink, and uneven flash mean a theoretical path kisses air in one spot and slams in the next. Float is there so contact continues inside a band, not so the robot can ignore coordinates.
Mega structural die castings make the split obvious. Parting lines, gates, and rib crossings on one shot can differ by a visible step. Neview’s large-part layouts put floating and rigid spindles on the same cell because one casting contains stock that must be cut off and skins that will not survive a hard pass. One spindle mode for the whole part either refuses the gate or breaks the wall.
Float is also misread as a substitute for path compensation. It is not. Float covers local height and continuous contact. Systematic shrink, a shifted datum, or a batch that leans the same way still needs measured offset applied to the path. Float is contact. Compensation is location. Running only one of them shows up quickly in scrap.
Compliant contact is loop control, not a soft machine
Some castings fight displacement sensors. Thin walls move when clamped and move again when cut. The offset you measured is not the offset that exists three seconds later. Rewriting the path from a stale measurement makes the motion worse.
Compliant contact takes another route. Instead of pinning the tool to a position target, current, position, and torque loops keep the contact force inside a band. Change the band and the same taught path behaves differently on a different wall thickness. Aluminum housings, some thin die-cast skins, and ribs that like to give away are the usual candidates.
Soft has a cost. Too little force and flash remains. A little too much and the wall dished in — shiny, and wrong. Technicians who only know how to lower torque end up with parts that look processed and still fail a fingernail check on the parting line. The aim is a bounded force, not an apology.
Change the alloy and the mix of the three modes has to change
Gray or ductile iron: gates, risers, and parting flash dominate. Wheels and burrs do most of the work. Rigid cutting usually takes the larger share. Iron is relatively brittle. The shape stays when you push. What you fear is vibration and a wheel that sheds, not a wall that folds.
Aluminum: the same flash sits on a wall an order of magnitude thinner. Over-grind, chatter marks, and a rib root cut too lean matter more than raw removal. Float and compliance take more of the cycle. Disc mills and profile cutters show up more often than a hard wheel leaned into the part.
Steel: harder metal, nastier scale and burned-on sand. Stiffness and spindle power come up before “soft touch.” Compliance can finish a contour. It will not remove a steel gate on a serious cycle time without killing the tool.
Copying a DR iron parameter set onto an MR aluminum cell, then blaming the robot for “not knowing aluminum,” is a category error. The contact strategy moved shops without moving with it.
Complexity is not a reason to open every float and soften every torque
A complex casting still has three kinds of real estate: process stock that must disappear, fits that must keep size, and blends that only have to look continuous. Maximum float plus minimum torque usually sacrifices the fit first and leaves the stock.
Zoning is dull and it works. Gate roots stay rigid. Long flash runs use float to keep the tool on the line. Thin corners switch to torque limit. The program looks less elegant. It matches how a good grinder already works by hand.
High-mix, low-volume shops should also keep zoning modest. Every zone is another teach and another buy-off. Some plants are better off locking 80 percent of the repeat surface with rigidity plus compensation, and leaving two awkward spots for a person, instead of specifying four spindles and full float on day one.
Over-grind, under-grind, and chatter usually point back to force
Over-grind likes thin walls, inside corners, and surfaces that already had a rough pass. The path still assumes the largest as-cast envelope, and the force never came down. Compensation that only fixes a bulk offset will still dig into a local dish.
Under-grind leaves a step on the gate or a wire of flash. Sometimes the wheel is dull. Often the cut needed rigidity and the float pushed the tool off at the exact place the metal was thickest.
Chatter can be structure, spindle balance, wheel runout, or a clamp that breathed. Adding float because a surface buzzed does not make a damper. It changes the shape of the vibration. Loose structure still has to be tightened.
Do not only ask whether the cell “has float”
Better questions: how much stroke, and can it lock to rigid; is torque one global value or zone-based; can compensation and float run together, and in what order; on a multi-spindle or mega-part cell, which spindles are rigid and which float; how many sample loops a thin-wall part usually needs before the path is released.
Neview lists measured error compensation and compliant contact as two capabilities because foundries meet two castings: parts you can measure, and parts that change while you cut. Measurable parts get a path rewrite. Unstable parts get a force loop. Asking how both are implemented is more useful than circling the word “smart” on a slide.
Some stations should not get a floating head
A single part number with tight as-cast consistency and a trustworthy datum can live on a rigid spindle and a honest fixture. Extra float is another mechanism in a dusty cabin and another parameter that can drift after a weekend.
A family of similar parts with wild flash, or die-cast batches whose shrink walks around, will burn teaching time if the tool cannot give or the path cannot shift. The practical test is still crude: does a skilled grinder ease off at this spot, or lean in because the metal will not yield? Ease off, and the cell needs give or a force cap. Lean in, and it is a rigid cut.
Close on selection, not on a slogan
In casting grinding, force shows up before brand. A gate has to come off, a thin wall has to stay, and a flash line has to be followed. Those three jobs rarely share one contact stiffness. Comparing rigid cutting, floating contact, and torque-limited blending side by side gets a buyer closer to the floor than comparing arm kilograms alone.
Cells that keep running are usually not the ones with every function turned to maximum. They are the ones that know which segment must stay hard, which segment may follow, and which segment must cap force. Spindle kilowatts, grit, and cycle time come after that argument is settled.
Read More
IPv6 network supported
