How to Optimize Double-Sided Grinding Machines for Flatness and Finish
This guide is for process engineers and shop supervisors running double-disc grinders on hardened steel, bearing races, and thin plates. It covers the seven adjustments that decide flatness, parallelism, and surface finish, with parameter ranges you can try on the next setup. Read it before your next wheel change or coolant swap.

In this article
- 1
- 2
- 3
- 4
- 5
- 6
- 7
- 8
- 9
What matters most when you optimize double-sided grinding machines
Start with the part, not the machine
Before you touch a single dial, write down what the part actually needs. Flatness, parallelism, thickness tolerance, and surface finish each push the process in a different direction. A bearing race that needs 2 μm parallelism and Ra 0.4 μm will not run on the same setup as a 200 mm clutch plate that only needs Ra 1.6 μm. Nail the target numbers first.
Then look at the material. Hardened 52100 at 60 HRC grinds cool and slow. Soft 1018 steel loads the wheel and smears unless you open the grit. Aluminium and copper alloys need a coarse, open wheel and heavy coolant or they clog within a few parts. The wheel that worked on last week's job will burn this week's.
Finally, check the stock removal split. Double-disc grinding removes material from both faces at once, typically 0.05-0.15 mm per side per pass in roughing and 0.005-0.02 mm per side in finishing. If the blank comes in with 0.4 mm total stock and a 0.1 mm bow, no wheel adjustment fixes that. The part will rock and you will chase taper all day.
- 1Write the target numbers downFlatness, parallelism, thickness, and Ra, in that order.
- 2Match the wheel to the materialHardened steel and soft low-carbon steel need opposite wheel grades.
- 3Check stock and bow before setupExcess bow forces the part to rock between the wheels.
Choose the right abrasive and wheel grade
For hardened steels above 50 HRC, vitrified-bonded cubic boron nitride (CBN) or ceramic alumina wheels hold form and cut cool. CBN lasts far longer between dressings and holds edge geometry on bearing races and valve plates. For soft steels and cast iron, aluminium oxide in a vitrified bond is the practical choice. Silicon carbide suits non-ferrous work but breaks down fast under heavy load.
Grade is where most shops get it wrong. A wheel that is too hard glazes, rubs, and burns the part. You will see blue temper colours and a shiny, smeared finish. A wheel that is too soft sheds grit and loses size within a few hundred parts. Start one grade softer than you think you need, watch the spindle load and the finish, then step up if the wheel wears too fast.
Grit size sets the finish ceiling. For Ra 0.8-1.6 μm, 80-120 grit works. For Ra 0.2-0.8 μm, move to 180-240 grit with a softer grade and lighter infeed. Do not try to reach a fine finish by slowing the wheel alone. The grit has to be fine enough, or you will burn the surface without improving Ra.
Bond hardness matters as much as abrasive type. Resin bonds run cooler and suit high-speed double-disc setups. Vitrified bonds hold form better and suit precision work where flatness is the controlling spec. Match the bond to the machine's wheel speed range, not just to the part.
- 1CBN or ceramic for hardened steel above 50 HRCHolds form, cuts cool, and lasts longer between dressings.
- 2Aluminium oxide for soft steel and cast ironVitrified bond, 80-120 grit, moderate grade.
- 3Start one grade softer than you expectGlazing and burning are harder to recover from than fast wear.
Coolant flow, pressure, and filtration
Coolant does three jobs on a double-disc grinder: it removes heat, flushes chips out of the contact zone, and carries away the fine swarf that causes scratching. Flow rate alone will not do it. You need pressure at the contact point, typically 6-10 bar through nozzles aimed at both wheel-part interfaces. Low pressure leaves chips in the gap and you get random scratches and a Ra reading that jumps around.
Water-based coolant at 4-8% concentration covers most steel and cast iron work. For aluminium and copper alloys, use a higher concentration and a coolant designed for non-ferrous metals, or the chips will weld to the wheel face. Straight oil suits some hardened steel jobs where finish is critical, but it brings fire risk and higher cost.
Filtration is the part everyone ignores until it costs them a batch. A 20-30 μm filter removes most damaging particles. If your filter is clogged, pressure drops, chips recirculate, and you will see the same scratch pattern on every part. Change or clean filters on a schedule, not when the finish goes bad.
Temperature control matters on thin parts. If the coolant warms up through the shift, the part grows and your thickness drifts. A chiller holding coolant at 20 ± 2 °C keeps size stable across a long run.
- 1Aim for 6-10 bar at the contact zoneFlow rate without pressure leaves chips in the gap.
- 2Filter to 20-30 μmClogged filters cause repeat scratch patterns.
- 3Hold coolant at 20 ± 2 °CWarm coolant drifts part thickness through the shift.
Set wheel speed, feed rate, and infeed together
These three parameters only work as a set. Wheel surface speed for vitrified CBN on hardened steel usually sits in the 25-35 m/s range. Aluminium oxide on soft steel runs 30-40 m/s. Push the wheel speed too high and you burn the part even with good coolant. Too low and the wheel rubs instead of cutting.
Feed rate and infeed then follow. In roughing, infeed per side is often 0.05-0.15 mm with a feed that keeps spindle load at 60-75% of rated. In finishing, drop infeed to 0.005-0.02 mm per side and reduce feed so the wheel has time to spark out. If spindle load climbs above 85% in finishing, you are either feeding too fast or the wheel is glazing.
A common mistake is to push feed rate to hit cycle time while leaving wheel speed alone. That raises load, increases heat, and shows up as taper or a curved face. If you need a shorter cycle, raise wheel speed and feed together in small steps, and check flatness after each change, not just thickness.
Spark-out time at the end of the pass matters more than most operators think. Two to four seconds of dwell with no infeed lets the wheel and part reach the same depth and removes the spring-back that causes a tapered face.
- 1Wheel speed 25-35 m/s for CBN on hardened steelAluminium oxide on soft steel runs 30-40 m/s.
- 2Roughing infeed 0.05-0.15 mm per sideKeep spindle load at 60-75% of rated.
- 3Finishing infeed 0.005-0.02 mm per sideAdd 2-4 seconds spark-out to remove taper.
Find and kill vibration before it shows on the part
Vibration on a double-disc grinder almost never comes from the wheel alone. Check the mounting pads first. Worn or uneven pads let the machine rock, and the part picks up a chatter pattern that repeats a few millimetres apart. Level the machine and replace pads before you change any grinding parameter.
Next check spindle runout. A spindle running out more than 5 μm will produce thickness variation that no wheel adjustment can fix. Measure runout at the wheel face with a dial indicator, not at the spindle nose. If it is high, the wheel is not the problem and dressing will not help.
Part clamping is the third source. Thin parts that are not held evenly between the wheels will spring and grind uneven. Use the correct carrier plate thickness and check that the part sits flat before the wheels close. On parts under 2 mm thick, a carrier that is worn by 0.05 mm is enough to cause parallelism error.
Finally, listen. A double-disc grinder that is cutting well has a steady, low-pitched sound. A high whine or a rhythmic thump points to wheel imbalance, a glazed patch, or a loose mount. Fix the sound before you chase the numbers.
- 1Level the machine and check pads firstWorn pads cause repeating chatter marks.
- 2Measure spindle runout at the wheel faceAnything above 5 μm will show as thickness variation.
- 3Check carrier plate wear on thin parts0.05 mm of wear on a carrier can cause parallelism error.
Dress on a schedule, not when the finish fails
Dressing removes the glazed layer and re-exposes sharp grit. If you wait until the surface finish goes bad, you have already made scrap. Set a dressing interval based on parts ground or spindle load, not on how the wheel looks. On hardened steel with CBN, that might be every 500-1,000 parts. On softer steel with aluminium oxide, it could be every 100-200 parts.
Dressing depth per pass should be small, typically 0.01-0.03 mm, with a slow traverse. Taking too much at once loads the dresser and leaves a rough wheel face that shows up as a coarse finish for the next few parts. Two or three light passes work better than one heavy pass.
After dressing, always grind a test part and check flatness and Ra before going back into production. A freshly dressed wheel cuts differently for the first few parts. If you skip the test part, those first parts are likely out of spec.
Track dressing in the setup sheet. Write down the interval, the depth, and the test-part result. Over a few months, that record tells you whether the wheel grade is right or whether you are dressing too often to compensate for a wheel that is too hard.
- 1Set interval by parts or spindle load500-1,000 parts for CBN, 100-200 for aluminium oxide.
- 2Dress 0.01-0.03 mm per passLight passes beat one heavy pass.
- 3Always grind a test part after dressingThe first few parts after dressing run differently.
Seven steps to optimize double-sided grinding machines
- 1Record the part requirementsWrite flatness, parallelism, thickness, and Ra targets on the setup sheet. Note material, hardness, and incoming bow. If the blank is bowed more than 0.1 mm, flag it for flattening before grinding.
- 2Check machine level and mounting padsLevel the machine to the manufacturer's spec and replace any pad worn more than 0.2 mm. Uneven pads cause repeating chatter that looks like a wheel problem.
- 3Measure spindle runout at the wheel faceUse a dial indicator on the wheel face, not the spindle nose. Keep runout under 5 μm. If it is higher, fix the spindle or wheel mount before changing any grinding parameter.
- 4Select wheel grade and grit for the materialCBN or ceramic for steel above 50 HRC, 80-120 grit for Ra 0.8-1.6 μm, 180-240 grit for Ra 0.2-0.8 μm. Start one grade softer than your first guess.
- 5Set coolant pressure and filtrationAim for 6-10 bar at both wheel-part contacts. Filter to 20-30 μm and hold coolant at 20 ± 2 °C. Check nozzle aim before every run.
- 6Set wheel speed, infeed, and feed togetherWheel speed 25-35 m/s for CBN. Roughing infeed 0.05-0.15 mm per side at 60-75% spindle load. Finishing 0.005-0.02 mm per side with 2-4 seconds spark-out.
- 7Dress on schedule and verify with a test partDress 0.01-0.03 mm per pass at the interval you set. Grind a test part and check flatness, parallelism, thickness, and Ra before releasing the run.
Starting parameters by material and target
Use these as starting points, then adjust to spindle load and finish.
| Material / target | Wheel and grit | Wheel speed | Infeed per side |
|---|---|---|---|
| Hardened steel, Ra 0.8-1.6 μm | CBN, 80-120 grit, vitrified | 25-30 m/s | 0.05-0.10 mm rough, 0.01-0.02 mm finish |
| Hardened steel, Ra 0.2-0.8 μm | CBN, 180-240 grit, vitrified | 25-35 m/s | 0.005-0.015 mm finish |
| Soft low-carbon steel | Aluminium oxide, 60-80 grit | 30-40 m/s | 0.08-0.15 mm rough, 0.02-0.03 mm finish |
| Cast iron | Aluminium oxide, 60-100 grit | 30-35 m/s | 0.08-0.15 mm rough, 0.02-0.03 mm finish |
| Aluminium and copper alloys | Silicon carbide or coarse Al2O3, 60 grit | 30-40 m/s | 0.05-0.12 mm rough, 0.01-0.02 mm finish |
| Thin plates under 2 mm | Fine grit, soft grade | 25-30 m/s | 0.005-0.01 mm finish, long spark-out |
Fix the machine before you tune the wheel
Most flatness and finish problems on a double-disc grinder trace back to leveling, runout, coolant pressure, or clamping, not to the abrasive. Check those four first. If the part still misses spec after that, the wheel grade and dressing schedule are the next levers.
Questions we hear from process engineers
Why does my flatness drift through the shift even when nothing changed?
Thermal growth is the usual cause. The machine, the part, and the coolant all warm up over the first two hours. If coolant temperature is not controlled, the part grows and thickness drifts.
Hold coolant at 20 ± 2 °C with a chiller and let the machine warm up for 30-45 minutes before grinding production parts. Check the first part against the last part of the previous shift.
How do I know if the wheel is too hard or too soft?
A wheel that is too hard glazes. The finish looks shiny and smeared, spindle load climbs, and you see temper colours. Dressing helps for a few parts, then the problem returns.
A wheel that is too soft sheds grit and loses size fast. You will dress often and still see size drift. Move one grade harder and re-check after 100 parts.
Can I reach Ra 0.4 μm on a double-disc grinder?
Yes, on the right part and material. You need a fine-grit wheel, light infeed, good coolant pressure, and low vibration. Thin or flexible parts are the limit, not the machine.
If Ra stalls at 0.8 μm, check filtration and spindle runout before changing the wheel. Dirty coolant and 10 μm of runout will cap your finish no matter what grit you fit.
How often should I dress the wheel?
Set the interval by parts ground or by spindle load, not by how the wheel looks. On hardened steel with CBN, 500-1,000 parts is a common starting point. On softer steel with aluminium oxide, 100-200 parts.
If you find yourself dressing more often to hold size or finish, the wheel grade is probably too hard. Change the grade rather than shortening the dressing interval.
What causes a tapered face on a double-disc grinder?
Taper usually comes from spring-back, not from the wheel. If the part is not held evenly or the spark-out is too short, one side grinds deeper than the other.
Check carrier plate wear, increase spark-out to 2-4 seconds, and confirm the part sits flat before the wheels close. On parts under 2 mm thick, 0.05 mm of carrier wear is enough to cause it.
Should I switch to CBN to improve cycle time?
CBN helps most on hardened steel above 50 HRC because it holds form and cuts cooler than aluminium oxide. That can mean longer dress intervals and fewer size corrections.
It is not automatically faster. Fitting CBN without fixing coolant pressure, filtration, or spindle runout will not solve a cycle-time problem. Fix the basics first, then try CBN on one cell and compare flatness, Ra, and cost per part.
Send us your grinding challenge
Share your part drawing and target flatness, parallelism, or Ra. Our engineers will review the setup and come back with a quotation and a free DFM analysis within 12 hours.
12-hour quote100% inspectionNDA on request