Workholding Strategy for Difficult CNC Parts
Plan how to locate and clamp hard parts, check holding force with real numbers, control distortion and deflection, and sequence operations so size holds after unclamping.
Table of contents
An easy part goes in a vise. A hard part makes you plan. It might be a thin ring, a long shaft, a plate that moves when you cut it, or a casting with no flat side. On these jobs the workholding decides whether the print can be held at all. Choosing it is a senior call, because the mistakes do not show up until the part is unclamped and springs out of tolerance. Reminder: a part that comes loose at speed is a thrown object. Never run a chuck, fixture or vacuum table beyond the limits its maker lists.
What You Will Be Able to Do
- Separate locating from clamping, and plan both from the print's datums.
- Estimate cutting force and check that the holding force is enough.
- Work out how speed reduces chuck grip.
- Prevent clamping distortion on thin parts.
- Control deflection and residual stress through setup and sequence.
- Prove a fixture before it goes into production.
Locating and Clamping Are Two Jobs
Locating puts the part in a known place. Clamping keeps it there. Mixing the two causes many fixture problems.
Good locating follows the 3-2-1 idea: three points on the primary surface, two on the secondary, and one on the tertiary. Each point should be solid and fixed. Use the print's datum features as your locators wherever you can. If the print measures from a bored hole and a face, locate from them, not from a convenient edge. Any error between the surface you located from and the datum goes straight onto the part.
Rules that hold on most jobs:
- Clamp against a locator. Push the part into its solid support, and never away from it.
- Clamp over support. A clamp with nothing under it bends the part.
- Point the cutting force at something solid. Let the fixed jaw or a stop take the load, rather than friction alone.
- Keep related features in one setup. Each time a part is reclamped, you add location error. If two features have a tight relationship, cut them without moving the part.
- Leave room for the tool and for chips. A locator buried in chips no longer locates.
How Much Force Is Enough?
Worked example. You rough a steel block in a vise, and the cut uses about 5 horsepower at the cutter, at 400 SFM.
Cutting force = 33,000 x horsepower / surface speed
Force = 33,000 x 5 / 400 = about 410 pounds at the tool.
Now say the force pushes along the vise jaws, so only friction holds the part. Two jaw faces share the work.
Holding force = 2 x friction factor x clamp force
Friction for steel on steel varies with finish and oil. Assume a low value, such as 0.15. To hold 410 pounds:
Clamp force = 410 / (2 x 0.15) = about 1,370 pounds
Add a safety factor for shock and interrupted cuts. A factor of 2.5 asks for about 3,400 pounds. That may be possible, but friction is a poor thing to rely on. If you turn the part so the force pushes into the fixed jaw, or add a stop at the end of the jaws, the metal takes the load, and friction only has to deal with what is left.
These numbers are estimates. They tell you whether you are close to a limit. Check what your vise can deliver at the torque its maker lists, and do not use a cheater bar, which can crack the vise and bend the part.
Chuck Grip and Speed
A lathe chuck loses grip as it spins, because the jaws are thrown outward. The force grows with the square of the RPM.
Force per jaw in pounds = 0.0000284 x jaw weight in pounds x radius in inches x RPM squared
Worked example. A chuck has jaws weighing 5 pounds each, with their center of mass 5 inches from the centerline. Its total static grip is 24,000 pounds.
| Speed | Force lost per jaw | Total lost, 3 jaws | Grip left |
|---|---|---|---|
| 1,500 RPM | About 1,600 lb | About 4,800 lb | About 19,200 lb |
| 3,000 RPM | About 6,400 lb | About 19,200 lb | About 4,800 lb |
Doubling the speed took the loss up four times, and most of the grip is gone. Heavy custom jaws and jaws set far out make it worse. Use the chuck maker's grip-versus-speed chart for your real numbers, and never go over its rated speed. Keep the chuck greased as its maker says, because a dry chuck grips far less than the gauge suggests.
Warning: Do not raise RPM on a job without checking the grip that is left. A part thrown from a chuck can carry enough energy to break through a machine window.
Thin Walls and Clamping Distortion
A thin part clamped hard changes shape. The tool then cuts a perfect shape on a distorted part. When you unclamp, the part springs back, and the perfect shape goes with it.
Worked example. A thin ring is held in a three-jaw chuck, and its bore is cut to perfect roundness on the machine. On the bench, it measures 0.003 out of round, with three high spots. Those three lobes are a fingerprint: each one lines up with a jaw. No offset can fix this.
Ways to reduce it:
- Spread the load. Use pie jaws, which wrap almost all the way around the part, or a full-contact collet.
- Drop the pressure for finishing. Rough at full clamp pressure, then release and reclamp at a much lower pressure for the finish cut. A lower pressure means less grip, so the finish cut has to be light, and the RPM has to respect the grip that is left.
- Hold where it is strong. Grip on a flange or thick section, not the thin wall.
- Support it from inside. Use a plug or a filler when you cut the outside.
- Clamp in the direction of the stiff section. On plates, clamp down over the supports.
Use a simple test. Put an indicator on the part and tighten the clamp. If the needle moves more than a small part of your tolerance, the clamp is bending the part. Do this test again with each change to the setup.
Thin Plate on a Vacuum Table
Vacuum spreads the load evenly, with no clamps in the way. But its limit is low, and it is easy to work out.
Worked example. A 6 x 8 inch plate has 48 square inches of sealed area. At about 10 psi of vacuum (around 20 inches of mercury), the down force is 48 x 10 = 480 pounds. With a friction factor of 0.2, the side force it can hold is about 96 pounds.
So that part suits light cuts only. If the cut breaks through and opens the seal, the holding force can vanish. Plan cut order to keep the seal, leave tabs or a skin, and keep the feed down near the end of the cut. A vacuum pump failure also has to stop the cycle, so confirm the interlock works.
Deflection on Long Parts
A long part bends away from the tool. That shows up as taper, chatter and poor finish. The stiffness falls with the cube of the length.
Deflection of a held part with a side force at its end = F x L cubed / (3 x E x I)
For a round steel bar, E is about 30,000,000 psi, and I = 3.14 x diameter to the fourth power / 64.
Worked example. A 1-inch steel bar with 200 pounds of side force:
| Stickout | Deflection |
|---|---|
| 6 inches | About 0.0098 inch |
| 3 inches | About 0.0012 inch |
Halving the length cuts the bending by eight times. That is why the first step is always to reduce the overhang, including tool overhang.
Working guide for turning, and sources differ, so treat it as a starting point:
| Length to diameter ratio | Usual support |
|---|---|
| Under about 3 to 1 | Chuck alone |
| About 3 to 1 up to 6 to 1 | Tailstock |
| Over about 6 to 1 | Steady rest, or a follow rest, plus tailstock |
Cut force, material and finish needs all shift these numbers. A tailstock pressed too hard can bow a slender part, so set it to what the job needs.
Residual Stress: Parts That Move After Cutting
Rolled plate, forgings and welded parts hold locked-in stress. When you remove material, the balance changes and the part moves. You can make a part perfectly flat in the machine, take it out, and watch it bow.
This cannot be solved by clamping harder. It is solved by sequence.
- Choose stable material. Stress-relieved plate, such as aluminum in the T651 temper, moves less. A stress-relief heat treatment may be needed for weldments.
- Rough all over first. Take most of the material off both sides, leaving even stock.
- Release and let it move. Unclamp fully and let the part settle. Some shops rest the part overnight on critical work.
- Re-establish the datum. Lightly clean up one face, then flip and clean up the other, using light clamp force.
- Finish with light cuts. Keep clamp pressure low, so you are not bending it flat.
Worked example. A 12-inch aluminum plate needs flatness of 0.002 after a deep pocket is cut on one side. Cut in one go, it bows 0.008. So you rough the pocket, leaving 0.030 stock, release it, and measure 0.006 of bow. Then you shim the low corners so the clamps do not bend it, and face the back side flat. You flip it, and finish the pocket with light cuts. The part now measures 0.0015. The time added is real, and it is the only way to make the print.
Balance also matters. Cutting evenly from both sides releases stress evenly, while cutting a deep pocket on one side only pulls the part toward it.
Parts With No Good Surface
Castings, forgings and odd shapes often have nothing flat to locate from. The first operation has to create a reference.
- Locate on the same points the print names as datum targets, if it has them.
- Use adjustable or self-seating supports, then lock them, so that you do not rock the part.
- Machine a good datum first, then locate everything else from it.
- On parts that bend easily, use extra supports that touch the part after it is located and do not force it.
Too many fixed locators on a rough surface makes the part rock or twist. Three fixed points on a casting are stable. Four can rock.
Proving a Fixture
Before a fixture goes into production, test it, because a fixture can pass a first part and fail on the fifth.
- Repeat test. Load the same part 10 times, and probe or indicate one point each time. The range of readings is the fixture's repeat error. It should be a small part of the tolerance.
- Clamp test. Check how far the part moves when you clamp.
- Free-state check. Measure the part clamped and unclamped.
- Load test. Have the real operators load it. If the part can go in backward, it will. Add a pin or a stop to prevent that.
- Chip test. Run a full cycle and look at where chips land. Add air blast or covers where they collect.
Worked example. Ten loads give readings across a 0.0012 range, and the feature tolerance is +/- 0.002, a band of 0.004. So the fixture alone takes 30 percent of the band, leaving little for tool wear and the machine. Find the loose part of the locating before you run the job.
Common Mistakes
| Mistake | What it costs |
|---|---|
| Locating from a convenient surface | Error between it and the datum |
| Relying on friction when a stop would work | Part moves under a heavy cut |
| Clamping thin parts hard for the finish cut | Part springs out of shape on release |
| Raising RPM without checking grip | A thrown workpiece |
| Finishing a stressed part in one setup | Part moves after unclamping |
| Clamping a bowed part flat | Bow comes back when released |
| More locators than the surface needs | Rocking or forced parts |
| Skipping the repeat test | A fixture that eats the tolerance |
Check Your Understanding
1. A cut draws about 3 horsepower at 300 SFM. Roughly what force does the tool see?
33,000 x 3 / 300 = about 330 pounds. Use that to judge the clamp force, or better, to aim that force at something solid.
2. A programmer wants to double the RPM on a lathe job, and the setup now uses heavier custom jaws. What do you check first?
The grip left at that speed, from the maker's chart. Doubling the RPM makes the loss four times larger, and heavier jaws add to it. Confirm the speed is under the rating too.
3. A round part measures perfect on the machine, and three-lobed on the bench. What is happening, and what do you try?
Clamping distortion. The jaws deformed the part. Try pie jaws, lower pressure on the finish cut, or holding at a stronger section.
4. A slender bar held by a chuck and tailstock chatters and tapers, even on light cuts. Its length is 10 times its diameter. What do you try?
A steady rest or follow rest. At that ratio, a tailstock alone is not enough, and pressing it harder can bow the bar.
5. A welded steel frame comes off the machine in tolerance and is out of flat the next day. What do you change before the next one?
Welds lock in stress, and cutting released it. Ask about a stress-relief heat treatment, then rough, let the frame settle, and finish with light cuts.
Key Takeaways
- Locating and clamping are separate jobs. Locate from the print's datums, and clamp against the locators.
- Run the force numbers, and point cutting force at something solid.
- Chuck grip falls with the square of speed, so use the maker's chart.
- A part clamped out of shape will spring back. Spread the load and lighten the finish clamp.
- Deflection grows with the cube of length, so reduce overhang first.
- Residual stress is solved by sequence, not clamp force.