Surface Grinding Tolerance: Realistic Flatness and Parallelism

A 200 × 150 × 20 mm tool steel plate arrives with a drawing callout of 0.01 mm flatness. The shop fly cuts both faces, checks it on a surface plate, and it reads 0.03 mm in the middle. They grind it, it reads 0.008 mm, they ship it. Two weeks later the customer measures it again and finds 0.04 mm of bow, and now nobody agrees whose measurement is correct.

Plate flatness is one of the few specifications where the measurement, the process sequence and the material's internal stress all matter as much as the machine. A grinder can produce a surface flat to a couple of microns. Whether the plate stays flat is a different question, and it is decided long before the part reaches the grinder.

This guide covers the three callouts buyers confuse, why plates move, what each process realistically delivers, and how to write a specification that a shop can actually meet.

Flatness, parallelism and thickness are not the same thing

These three callouts get used interchangeably in drawings and they should not be.

Flatness is a property of a single surface. It says the surface lies between two parallel planes a stated distance apart. Nothing else on the part is referenced.

Parallelism relates two surfaces. It says the top face must stay within a tolerance zone parallel to the bottom face, which is the datum. A plate can be perfectly flat on both faces and still be wedge-shaped, failing parallelism.

Thickness tolerance controls the distance between the faces. A plate can meet a ±0.02 mm thickness callout while both faces are bowed in the same direction.

A part that must sit without rocking on a fixture needs flatness. A part that must transmit load evenly between two mating surfaces needs parallelism. A spacer needs thickness. Writing all three when only one matters adds cost with no function.

Why a plate moves after you machine it

Rolled plate and bar carry residual stress from the mill. When you remove material from one face, you remove the layer that was balancing the stress below it, and the plate relieves itself by bowing. The thicker the plate and the more material you remove, the larger the movement.

A 20 mm plate with 3 mm removed from each face will typically move 0.02 to 0.08 mm. A 6 mm plate can move 0.1 mm or more, because there is less section left to resist the imbalance. Aluminum is worse than steel for this, since it is softer and relieves more readily; cast aluminum tooling plate is stress-relieved at the mill and moves far less.

The standard countermeasure is a three-step sequence: rough machine with 0.3 to 0.5 mm of stock left per face, stress relieve (either a thermal cycle or a natural ageing period), then finish machine or grind. Skipping the middle step is the most common reason a plate that measured flat in the shop is not flat at incoming inspection.

Fly cutting versus surface grinding

Each process has a different capability and a different cost structure.

Fly cutting on a VMC uses a single-point tool with a large radius sweeping across the face. It is fast, it can be done in the same setup as other features, and it produces a surface with a controlled scallop pattern. Flatness is limited by the machine's geometry and by how much the part deflects under clamping: 0.02 to 0.05 mm over 200 mm is typical, and 0.01 mm is achievable on a rigid part with a light finish pass.

Surface grinding uses a rotating vitrified wheel on a reciprocating table. It removes small amounts of material with low force, so the finished surface follows the magnetic chuck rather than the part's own stiffness. It is slower, requires a separate setup, and typically costs $8 to $25 per face depending on area. Flatness of 0.005 mm over 300 mm is routine.

Realistic values by thickness

The table below reflects what a competent shop delivers routinely, not what is theoretically possible on a demonstration part. Values assume steel, a part with a reasonable aspect ratio, and an adequate stock allowance.

Two things about this table matter more than the numbers. First, thin parts are not automatically easier — a 3 mm plate deflects under the magnetic chuck and springs back when released, so grinding it flat on the chuck does not make it flat off the chuck. Second, past about 60 mm thickness, the limiting factor is usually the grinder's own geometry and the part's weight, not the process.

The magnetic chuck trap

A surface grinder holds the part with magnetism, and that clamps the part to the table. If the plate is bowed, the chuck pulls it flat, the wheel grinds a flat surface, and the part springs back to its bowed shape when you release it. The ground face is flat only while it is being held.

The fix is to grind in stages: take a light pass, release, check, re-clamp and repeat with shims under the high spots until the part sits without being forced. Experienced hands do this automatically on thin work. If your part is under 8 mm thick and calls 0.01 mm flatness, expect the shop to quote extra time for it, and expect them to ask whether the callout applies free-standing or clamped.

Writing a specification the shop can meet

A few conventions remove most of the disputes.

State the datum. A flatness callout on a face that is also the datum is self-referential and unmeasurable in practice. Pick a real datum feature, usually the opposite face or a machined edge.

State the measurement condition. "Flatness 0.01 mm, free-standing, measured on a granite plate with a 0.001 mm indicator" is a specification. "Flatness 0.01 mm" alone is an argument waiting to happen.

Allow the sequence. If the part needs stress relief between roughing and finishing, write it into the drawing notes rather than assuming the shop will add it. It adds a day or two to lead time, and it is cheaper than remaking the part.

When we quote a plate like this at JSHS, the flatness callout is usually what decides whether the part goes on a grinder at all. A 0.02 mm requirement over 200 mm can often be met by fly cutting with a proper stress-relief sequence, which is roughly a third of the cost of grinding both faces.

Frequently asked questions

How flat can surface grinding get?

On a rigid part under 50 mm thick, 0.005 mm is routine and 0.002 mm is achievable with careful setup and temperature control. Below that, you are measuring the granite plate and the indicator as much as the part.

Why did my plate bow after grinding?

Almost always residual stress. Grinding removes a small amount of material from a stressed surface, and the balance shifts. The cure is stress relief before finishing, or a symmetric removal sequence that takes equal amounts from both faces.

Is fly cutting accurate enough instead of grinding?

For flatness of 0.02 mm and above over 200 mm, usually yes, especially on aluminum. Below 0.01 mm, or when parallelism between faces is critical, grinding is more reliable and often not much more expensive once you count the risk of a scrapped part.

Does anodizing change flatness?

It can. Type II anodizing builds 10 to 25 µm per surface and introduces growth stress that bows thin plates. A 3 mm aluminum plate anodized on one side only will bend noticeably. Mask the non-critical face or specify the flatness after coating.