How Machine Base Flatness Is Actually Measured

Straightedge, precision level, autocollimator, laser and CMM compared for large machined surfaces, plus the support and temperature effects that invalidate results.

Flatness is the most-specified and least-defined requirement on a machine base drawing. “Flat within 0.05” appears constantly, and on a large surface it is not yet a specification.

What flatness means, precisely

Flatness is the distance between two parallel planes that just contain the surface. That is all. It needs no datum, it says nothing about orientation, and a surface can be perfectly flat while sitting at an angle to everything else on the part.

This matters because engineers often write flatness when they mean parallelism to a mounting face, or squareness to a datum. Those are different controls, they need a datum, and they are measured differently. Getting this right in GD&T removes most of the argument before it starts.

The methods, and where each one fits

Straightedge and feeler gauge. Fast, cheap, and coarse. Useful for a quick check and for finding an obvious high spot. It measures along a line, not across an area, so it cannot report a real flatness value for a surface.

Precision level. A sensitive level walked across the surface in a grid, accumulating angular readings into a height map. The workhorse for large machine bases, because it needs no reference larger than the instrument and works in place. Slow, and it measures relative to gravity, so it cannot see a uniform tilt.

Autocollimator. An optical instrument reading the angle of a mirror on a carriage moved across the surface. Very accurate over long distances, well established for machine tool work, and needs a clear optical path and a stable setup.

Laser interferometer or laser tracker. Fast over large areas, produces a dense map, handles awkward geometry and multiple faces in one setup. Expensive, and not every fabricator has one. A tracker gives you the whole part; an interferometer gives higher accuracy along a line.

CMM. Ideal when the part fits, which for a machine base often means it does not. Where a large-envelope CMM exists it is the cleanest answer, giving flatness, parallelism and squareness against a real datum scheme in one operation.

Surface plate and dial indicator. The reference method for small parts. Needs a plate larger than the part and the means to get it there, so it is rarely available for a base of any size.

The things that invalidate a measurement

This is the part that causes disputes, and none of it is instrument error.

Support scheme. A large structure sags between its supports under its own weight. Measured on three points it takes one shape; on six it takes another; grouted to a foundation, another again. All three are correct measurements of different conditions. Unless the drawing says how the part is supported during inspection, the number is not repeatable.

Temperature. Steel and iron expand appreciably with temperature, and a large base in a shop that swings ten degrees between night and day is a different size in the morning than in the afternoon. Worse, a temperature gradient through the structure bends it. A measurement taken while the part is still warm from machining is measuring the wrong shape.

Settling time. A part craned onto supports needs time to settle before it reads consistently. Measure it immediately and again some hours later and the numbers will differ.

Point density. Flatness from nine points and flatness from a hundred are different numbers on the same surface, because more points find more deviation. State the grid.

Cleanliness. On a large machined face, a chip under a foot or a burr at an edge produces a reading that has nothing to do with the surface.

Writing a requirement that can be met and checked

Specify all six of these:

1. Which surface, identified unambiguously, each pad separately or the whole face as one.
2. The value, and over what length. 0.05 mm over 300 mm and 0.05 mm over 2000 mm are entirely different jobs at entirely different prices.
3. The method, or an agreed list of acceptable methods.
4. The support condition during inspection.
5. The temperature, or an acceptable range, and whether a soak period is required.
6. The point grid or scan density.

Six lines of notes. They convert an argument into an inspection, and they let three fabricators quote the same job rather than three different ones.

Be honest about what the machine needs

Tightening flatness is not linear in cost. Moving from 0.10 mm to 0.05 mm over a metre is often another cut. Moving from 0.05 mm to 0.02 mm can mean a different machine, an extra setup, a temperature-controlled bay and a second stress relief cycle.

The most expensive drawings in this business are the ones where a tight default tolerance was applied to every surface because nobody asked which surfaces actually matter. Rails sitting on a grouted plate do not need what a metrology frame needs. Decide which faces are functional, specify those properly, and let the rest be as-machined.

See also: flatness measurement methods compared, with grid patterns and report contents.

Common questions

← All guides