How Stiff Does a Machine Base Need to Be?

Setting deflection and natural frequency targets for a machine base, why stiffness and damping are different problems, and how to check a design against both.

“Make it stiff” is not a specification. Two engineers will size the same base very differently from that instruction, and both quotes will come back at prices you cannot compare. Stiffness needs numbers, and there are only a few that matter.

Two different problems, constantly confused

Stiffness is how far the structure moves under a steady load. Cutting force pushes, the base deflects, the tool is not where the control thinks it is, and the part is wrong.

Damping is how fast the structure stops moving after a disturbance. A cut starts, the structure rings, and the ripple appears in the surface finish.

They are independent. A welded steel base can be extremely stiff and almost undamped: it barely moves under load and rings like a bell when struck. Cast iron, mineral cast and filled weldments trade some of the stiffness-per-volume for damping.

So the first question on any base is which failure you actually have. Adding plate to a chatter problem is the most common expensive mistake in this field, and it is common because both problems present as “the machine is not accurate enough”.

Setting a deflection target

Start from an error budget rather than a rule of thumb. Your machine has a total allowable error. Spindle, ways, thermal drift, control and the structure each get a share. Decide what share the base is allowed, and that is your deflection limit under the governing load case.

Having done that, sanity-check against normal practice. Machine tool work commonly applies a deflection limit of span divided by 3000, very much tighter than the L/250 or L/360 of building structures, because the consequence is a scrapped part rather than a cracked plaster ceiling.

Some points that decide whether the number means anything:

• Which load case? Maximum cutting force, worst-case tool position, a moving axis at the end of its travel, or the machine’s own weight distribution. State it.
• Measured between what? Deflection of the base in isolation is not interesting. What matters is relative displacement between the tool and the work, which means the whole loop through the base, the ways and the column.
• Supported how? A base on three levelling points behaves differently from the same base on six, and differently again when grouted. Analyse it as it will actually sit.

Setting a dynamic target

Static stiffness is the easy half. The dynamic behaviour is what usually limits the machine.

The design aim is to keep the structure’s first natural frequency clear of the frequencies the machine excites, with margin. Targets above 80 to 100 Hz are commonly quoted for machine tool structures, and are a reasonable starting point.

But the real requirement is separation from your specific excitations:

• Spindle rotational frequency across the working speed range.
• Tooth passing frequency, which is spindle speed times the number of flutes and is usually the troublemaker.
• Servo bandwidth and any reversal or dwell.
• Pumps, fans and anything else with a rotating unbalance bolted to the structure.

Map those against the first few modes. A structure whose first mode falls between two operating speeds is worse than one with a slightly lower first mode that sits below everything.

Checking the design

Hand calculation first. Treat the base as a beam or a plate and use classical beam theory to get an order of magnitude for deflection before anyone opens a CAD model. It takes minutes and it catches designs that are wrong by a factor of five, which no amount of mesh refinement will rescue.

Then FEA for the real geometry, the modal results and the load cases hand calculation cannot cover.

Then a physical check on the first unit. A tap test on the finished base tells you the actual modes and gives you the damping the model could only estimate. It is a cheap measurement and it is the only one that includes the joints, the welds and the real support conditions.

If the numbers do not pass

In rough order of cost effectiveness:

1. Deepen the section. Stiffness goes with the cube of depth. Free if the envelope allows it.
2. Improve the rib layout. Triangulate against the governing twist. Often buys stiffness and saves weight at the same time.
3. Change the support points. Moving levelling points can transform deflection with no change to the structure at all.
4. Add mass or damping. Fill the sections if the problem is dynamic rather than static.
5. Change material. Cast iron or mineral cast, with all the tooling cost that implies.
6. Thicken the plate. Last, not first. Most cost per unit of stiffness gained.

That order is worth memorising, because designs under schedule pressure tend to reach for the last option first.

What to hand a fabricator

They are not going to re-analyse your machine, and you should not want them to. Give them the geometry, the plate grades, the weld specification, the machined faces and tolerances with their datums, and the stress relief requirement. Your analysis stays yours. What they need to quote is the structure, not the reasoning behind it.

Common questions

← All guides