Ask ten designers what holds up a non-standard machine and nine will say strength, because strength is what the stress calc checks and what the FEA report prints. The one who says stiffness is the one whose machines do not vibrate, whose carriages do not bind, and whose measurement probes hold their zero across a shift. In this article I want to talk about frame and column design the way it actually matters on the shop floor: stiffness rules, strength follows, and the geometry of the section is worth more than the grade of the steel.
1. Why Stiffness, Not Strength, Is the Real Requirement
A machine frame that is strong enough to hold twenty tonnes without yielding can still deflect half a millimetre under a tool load it sees every cycle, and that half millimetre is the difference between the bore position staying put and the part coming out oval. Strength protects against failure; stiffness protects against the error, and on non-standard equipment it is the error that costs the customer money.
The classic way I explain it to a junior: strength is the fuse, stiffness is the ruler. The frame must not break, obviously, but the design requirement that actually drives the section sizes is the deflection limit, the carriage that must not tilt more than 0.05 mm per metre, the probe that must keep its aim. When in doubt, design the frame for deflection and check that it does not yield, and the machine will behave.
That ordering explains why a cast chassis and a welded fabrication of the same stiffness end up with completely different wall thicknesses, and why hollow sections routinely beat solid bars of the same weight: stiffness comes from placing the material far from the neutral axis, which is a geometry decision before it is a material decision.
2. Section Geometry: Flinging Material Away From the Neutral Axis
For a beam in bending, the deflection scales with the moment of inertia of the cross section, and the moment of inertia scales with the fourth power of the outer dimension and only the first power of the wall thickness. The engineering consequence is brutal and liberating at the same time: a slightly larger box section wins over a much thicker plate. Doubling the outer dimension of a square tube roughly multiplies its stiffness by sixteen, while adding wall thickness to the same section gives a modest linear bump.
The rule I hold in my head: for a given weight, the stiffest section is the one with the most material at the extreme fibres. A wide-flange section beats a square tube when the load is a pure bending moment in one plane, and the square tube wins when the load can come from any direction, which is most of the machine frame cases I meet. Round tube is the best torsion but worse in one-direction bending than a rectangle of equal material and envelope.
| Section | Best Load Case | Penalty |
|---|---|---|
| Wide flange (H-section) | Uniaxial bending, columns | Weak in torsion and weak axis |
| Square tube | Bi-axial bending, torsion | Weight per stiffness is fair |
| Rectangular tube laid broad | One dominant bending plane | Weak across the narrow way |
| Circular tube | Torsion, omnidirectional | Worse for pure uniaxial bending |
The honest engineering advice: on a non-standard frame, assume the load direction is never quite what you assumed, and pick a section whose stiffness does not collapse when the load angle shifts. That bias toward the square and rectangular tube is not fashion, it is insurance against the unknown direction of the next audit.
3. The Column: Buckling Is a Stiffness Story
The vertical column that holds a moving carriage is the classic stiffness problem dressed as a strength problem. Everyone computes the compressive stress and the Euler buckling load, and the frame passes, and then the machine sways at the top by more than the tolerance because the column is a stiffness member, not a strut, and the sway is the load times the inverse stiffness of the cantilever.
The deflection at the top of a cantilever column under a sideways load scales with the height cubed and inversely with the moment of inertia. This is the single most important scale law in machine design: double the height of the column and, for the same section, the top deflects eight times as much. No amount of steel grade fixes that, but widening the column footprint or adding a top tie-back does, because it changes the effective length and the load path.
delta_top = P * H^3 / (3 * E * I), where P is the sideways load, H the column height, E the modulus and I the moment of inertia. The cube of the height owns the deflection, and only I or the length term can fight it.
When I size a column, I start from the allowable top deflection, work out the required I, and then pick the section. If the required I forces a column too fat for the layout, the real fix is a structural one: tie the top to a wall, add diagonal bracing to a second frame, or carry the load lower. Fighting the cube with bracing is nearly always cheaper than fighting it with steel.
4. Welded Frames: The Joints Betray the Stiffness
A frame is only as stiff as its joints, and the welded fabrication has a particular weakness: the columns and beams are stiff, but the weld fillets at the corners are thin, and the joint flexibility adds up on a tall frame like a loose bolt on a stepladder. The classic failure is the frame that deflects correctly in a continuous-beam model and sags in reality because the model treated every corner as a rigid node and the shop made every corner a flexible hinge.
I model the welded corner with its real geometry or at least with a joint stiffness penalty, and I add gusset plates at the high-load corners before the shape is finalized, not after the vibration shows up. A diagonal gusset at the top corner of a column-to-beam connection is one of the cheapest stiffness upgrades in the business, an angle, four weld passes, and the corner stops being the weak link.
The frames that surprise me are never the thin sections, they are the thick sections connected by geometry that cannot carry the moment. The weld does not have to break to ruin the stiffness; it only has to flex.
Weld distortion is the second joint problem. The shop pulls the frame out of square by a millimetre over the length, and then the carriage rails are installed on a frame that is already curved, and the binding starts before the machine runs. The cure is design for welding: balanced welds, symmetrical layout, minimal heat input on the critical long members, and a stress-relief or a machining pass on the mounting faces after welding. On precision frames, the rail faces are machined after welding, always, and buying the frame already rough saves nobody the rework.
5. Bolted and Modular Frames: The Joint Flex Again
Modular aluminium framing has taken over a big share of non-standard equipment, and it deserves it, but the stiffness story is different. The joint is a bolted connection through T-slots and corner plates, and the joint has a defined flexibility that shows up as a rotation under moment. A tall frame built from 40×40 profiles with corner brackets flexes far more than a welded tube frame of the same external size, because the bolted joint is the compliant part.
The numbers are sobering on the one hand and fixable on the other. I have measured a 2-metre 40×40 frame swaying close to 1 mm at the top under a modest hand push, where a welded 60×60 tube frame of similar weight stayed under 0.2 mm. The trick with modular frames is to stop pretending the joints are rigid and to add bracing, diagonal ties, gussets, or a steel core where the stiffness must be absolute.
| Frame Type | Relative Joint Flex | Best Use |
|---|---|---|
| Welded tube, gusseted | Low | Precision and heavy load |
| Cast chassis | Very low | High stiffness, high cost |
| Modular aluminium, corner brackets | Medium | Rapid reconfigurable guarding |
| Modular with steel core / bracing | Medium-low | Stiff but adjustable |
For the frames that carry a measurement or a machining relationship, I never trust the bare modular joint. I either buy the heavier profile range and the stiffened connectors, or I brace the bay diagonally, or I route the precision axis through a dedicated steel member. The aluminium frame becomes the envelope and the guarding; the steel member becomes the ruler, and the two jobs do not get mixed.
6. Damping and Dynamics: Stiffness Is Not the Whole Story
Stiffness fixes the static position, and dynamics fix the nuisance. A frame can be perfectly stiff and still ring like a bell at the start of every cycle, and the ringing shows up as chatter, as a probe that loses its reading mid-cycle, as an operator who learns to work around the vibration. The dynamic requirement is that the frame does not hit its resonant frequency in the operating range of the machine, or if it does, that the excitation at that frequency is small.
My working numbers for a non-standard machine: keep the first natural frequency of the frame above 1.5 times the highest forcing frequency of any reciprocating or stepping source, and when the source is a servo with a fast axis response, treat the frame as a structural spring in the servo loop, because the control system sees the frame flex as a phase lag. The flexible frame is why a stiff servo can feel mushy at the tool.
The practical damping upgrades are boring and effective: box sections filled with concrete or sand on the most sensitive bases, ribbed and cast members on the critical axis, rubber or polymer-injected mounts between the frame and the floor, and above all, avoid flat unsupported panels that drum. I have silenced a noisy frame by cutting a stiffening fold or adding a viscoelastic layer to a cover plate, and the machine got faster in the tuning session, which is the funniest metric: better damping looked like a servo improvement.
7. The Sizing Sequence I Use and a Worked Example
Here is the order I actually put on a new machine frame. Define the deflection limits at the functional points, the carriage, the probe, the tool tip, with numbers from the process, not from the guess. Compute the required stiffness of the load path to meet those limits under the worst operating load. Size the section members to that required stiffness using the cube-law and the moment of inertia tables. Add the joint and weld penalty to the model and check again, because the perfect members with flexible joints fail the budget. Then run the natural frequency check and add damping or bracing where it fails.
Worked example to make it real: a gantry picker needs the tool tip to hold plus or minus 0.1 mm under a 300 N side load, the gantry beam is 2.5 m long and the vertical column 1.4 m tall. Treating the column as a cantilever, a 60x60x4 square tube has a moment of inertia of about 430,000 mm4 across the axis, and the top deflection under a 300 N force at the tip of the 1.4 m column comes out near 0.3 mm with a typical steel modulus, three times the budget. The first fix is not a bigger column, it is a top tie-back that halves the effective length, which cuts the deflection by a factor of eight to a comfortable value, and the second fix is a slightly heavier section if the tie-back cannot be fitted. That is the whole arithmetic, and it lives on one page.
Try to notice what did not appear in that example: the stress. The 300 N load and 0.1 mm budget are so far below yield on any 60 mm tube that stress never enters the decision, which is exactly the point of this article. Design the frame for the ruler, check the fuse, and the non-standard machine will stand where it is put and measure what it is told.
