Control deflection as well as strength

A fixture can remain far below a material’s strength limit and still move enough to spoil the part. Strength asks whether a component survives a load under the relevant failure criteria. Stiffness asks how much its position changes while carrying that load.

For precision workholding, the second question can control the design. It also includes the workpiece: a part machined while bent may spring into an unacceptable shape after it is unclamped.

Start with a screening model

This is an illustration. A replaceable steel bridge in a proposed fixture is represented as a simply supported rectangular beam with a central load of 900 N. Its support span L is 100 mm, width b is 30 mm and thickness t is 10 mm. Assume an elastic modulus E of 200,000 N/mm².

The rectangular second moment of area about the bending axis is:

I = bt³ / 12 = 30 × 10³ / 12 = 2,500 mm⁴.

For the ideal central-load beam, maximum deflection is:

δ = PL³ / (48EI) = 900 × 100³ / (48 × 200,000 × 2,500) = 0.0375 mm.

The standard equation is documented in David Roylance’s Beam Displacements lecture note. It assumes small elastic deflection and ideal supports; it is not a complete model of a bolted fixture or a plate.

Check stress without confusing the decision

For the same ideal beam, the maximum bending moment is PL / 4 = 22,500 N·mm. The distance from the neutral axis to the outer surface is 5 mm, giving a nominal maximum bending stress of:

σ = Mc / I = 22,500 × 5 / 2,500 = 45 N/mm², or 45 MPa.

This stress calculation does not establish a permissible working stress for an unspecified steel, nor does it check holes, notches, contact stresses or fatigue. It simply shows that the stress and deflection calculations answer different questions.

If the fixture’s provisional deflection allocation is 0.010 mm at this location, the predicted 0.0375 mm fails that allocation regardless of whether a later strength check passes. A report containing only a favourable stress result would miss the dimensional problem.

Compare geometry changes

The beam model makes the sensitivity clear: deflection is proportional to load, to span cubed, and inversely proportional to thickness cubed when width and material remain fixed.

Increasing thickness from 10 to 16 mm changes the predicted deflection by the factor (10/16)³:

δ16 = 0.0375 × (10/16)³ = 0.009155 mm, approximately.

Keeping the 10 mm thickness but reducing span from 100 to 60 mm gives:

δ60 = 0.0375 × (60/100)³ = 0.008100 mm.

Figure 1Deflection of the support beam under 900 N

10 mm plate, 100 mm span0.0375 mm

16 mm plate, 100 mm span0.009155 mm

10 mm plate, 60 mm span0.0081 mm

Provisional allocation0.01 mm

Both changes bring the ideal beam result below the provisional 0.010 mm allocation. Neither establishes the complete fixture deflection. The shorter, deeper proportions also make neglected shear and actual support conditions more important to review.

Increasing width is less powerful in this model: doubling width halves deflection, while doubling thickness reduces it to one-eighth. This is a useful concept comparison, not a reason to increase thickness without checking tool access, mass and mounting.

Include the interfaces in the compliance path

A stiff bridge can sit on a compliant joint. As a separate teaching assumption, suppose a contact in the same series load path has stiffness 60,000 N/mm and carries the same 900 N.

Its displacement is 900 / 60,000 = 0.015 mm. Added to the original beam displacement, the total series displacement is 0.0525 mm. With the 16 mm bridge, it is still approximately 0.024155 mm, above the 0.010 mm allocation.

Figure 2Compliances in series add

1. Beam, 10 mm plate0.0375 mm · 71%
2. Compliant contact at 60,000 N/mm0.015 mm · 29%

The assumed contact stiffness is not a catalogue recommendation. The example shows why improving one member can leave the dominant movement elsewhere. Adding deflections is appropriate here only because the assumed elements carry the same load in series along the measured direction.

Real joints can have nonlinear contact behaviour, frictional slip and hysteresis. Several supports may share load in parallel. Use a model that represents the actual arrangement rather than adding every deformation number found in a simulation report.

The base joint is part of the fixture

The fixture’s connection to the machine table is a joint in series with everything else. A base plate that does not sit flat on the table, tee-slot bolts tightened in a random order, or chips trapped under the base all add compliance and can distort the fixture as it is bolted down.

Good practice includes cleaning the table and base before mounting, checking that the base sits without rocking, tightening the bolts in a stated sequence and to a stated method, and locating the fixture on the table with keys or pins rather than indicating it into position each time. A distorted base can move the locating contacts by more than the whole location budget, so it is worth checking the locators after mounting during the trial, not only on the bench.

Divide the deflection allocation among the members

A single 0.010 mm allocation for the whole fixture is hard to design to, because the movement comes from several places at once. Dividing it gives each member a target. As an illustration, the bracket fixture’s 0.010 mm might be split as:

• Support bridge or plate: 0.004 mm
• Contacts and joints in the load path: 0.003 mm
• Workpiece overhang at the critical feature: 0.003 mm

Each share then becomes a design check that can be passed or failed separately. The 16 mm bridge at 100 mm span, at about 0.0092 mm, fails its 0.004 mm share on its own, and the 60 mm span, at about 0.0081 mm, fails it too. That points the design towards a stiffer or shorter support, or a different allocation, before anything is built. Like the tolerance budget in a dimensional stack, the deflection budget should be recorded with the reasons for its split and revisited if a member cannot meet its share economically.

Consider the part while it is clamped

The workpiece can be the most flexible member. A thin wall may deflect under a clamp, be machined in that state, then return after release. Inspecting only while clamped can hide the resulting free-state error.

Compare measurements before clamping, under the defined clamping condition and after release where this is relevant. Keep the measurement method and temperature consistent. An observed change may include seating as well as elastic bending, so interpret it with the contact model.

Support near the load path can reduce bending, but an added support must not lift the part off its locating references. The location and support decisions from Chapter 2 remain connected.

Avoid replacing stiffness analysis with maximum clamp force. More force can improve contact retention while worsening the feature you are trying to hold.

Check overhangs on the part itself

Wherever the cutter works beyond the last support, the part acts as a cantilever. The ideal deflection of a cantilever with a load at its free end is:

δ = PL³ / (3EI)

This is an illustration, using steel with E = 200,000 N/mm². Suppose the bracket’s full 80 mm width and 16 mm thickness overhang the nearest support by 40 mm, and an 800 N cutting force acts at the edge. Then I = 80 × 16³ / 12 ≈ 27,307 mm⁴, and:

δ = 800 × 40³ / (3 × 200,000 × 27,307) ≈ 0.0031 mm

That is within the 0.010 mm allocation. If the same region overhung by 80 mm, the cube of the length would multiply the deflection by eight, to about 0.025 mm, well outside it.

Figure 3The bracket as a cantilever under an 800 N cutting force

	40 mm overhang	80 mm overhang
Deflection at the edge	About 0.0031 mm	About 0.025 mm
Against the 0.010 mm allocation	Inside	Outside

The calculation points to a simple rule: put a support close to wherever the heaviest cut happens, and check the overhang at every machined feature, not only at the part’s centre. Thin features, such as a 6 mm tab, deserve their own check. A tab 30 mm wide and 6 mm thick, overhanging 20 mm and loaded with 300 N at its tip, deflects about 0.0074 mm by the same formula, most of the allocation from a small feature that is easy to overlook.

Stiffness and vibration go together

A fixture is a spring carrying a mass, and every spring-mass system has natural frequencies at which it vibrates readily. For a single effective stiffness k and mass m, the natural frequency is about:

f = (1 / 2π) × √(k / m)

This is an illustration. With an effective stiffness of 60,000 N/mm, or 6 × 10⁷ N/m, and an effective moving mass of 5 kg, the natural frequency is about 551 Hz. If the effective mass were 20 kg at the same stiffness, it would fall to about 276 Hz.

The cutting forces described in Chapter 3 pulse at the tooth-passing frequency, about 133 Hz for a four-flute cutter at 2,000 rpm, and at its multiples. A fixture whose natural frequency sits close to one of those multiples can resonate, giving chatter marks, poor finish and accelerated tool wear even though its static deflection looked acceptable. A stiffer and lighter fixture raises its natural frequencies, away from the strongest excitation. Where resonance is suspected, a change of spindle speed that moves the tooth-passing frequency away from the problem is often the quickest test.

Damping helps where stiffness cannot

Stiffness limits how far a fixture moves under a given force; damping limits how much it rings when excited. Materials and joints differ in how quickly they absorb vibration. Cast iron damps better than steel plate, and bolted joints with friction add damping as well as compliance. Polymer concrete and constrained-layer damping treatments are used where vibration is a persistent problem.

For a fixture like the bracket’s, stiffness is usually the first lever. Damping becomes worth considering when a stiff, well-supported design still chatters at a particular speed or tool, or when space and weight prevent further stiffening. The trial described in Chapter 7 is where such problems first show themselves, so it should include listening for chatter and inspecting the surface finish, not only measuring dimensions.

Temperature moves the fixture too

Deflection under load is not the only way a fixture changes shape. A fixture plate warms with the workshop through the day, with coolant temperature and with heat from the cut. Steel expands by roughly 12 millionths of its length per degree Celsius, using the same teaching coefficient as other Library books.

This is an illustration. If the end stop sits 150 mm from the fixture’s reference on the machine table, and the fixture warms by 5 °C between the first part of the morning and mid-afternoon, the stop moves by about 150 × 12 × 10⁻⁶ × 5 ≈ 0.009 mm relative to that reference. That is almost the whole 0.010 mm deflection allocation, from temperature alone.

The effect is easiest to manage by placing the fixture’s reference close to the features that matter, so that the distances that grow are short, by keeping coolant temperature stable, and by checking the first parts after a cold start against parts made once the machine and fixture have settled. The trial in Chapter 7 should include parts from both conditions, so that any warm-up shift is measured rather than discovered in production.

Use simulation with an explicit question

A finite-element model can help evaluate a complex fixture, but its boundary conditions are engineering assumptions. Fixing an entire base face rigidly can suppress movement that a real bolted interface permits. Bonding all contacts can prevent separation that occurs in service.

Begin with a question, such as the displacement of a controlled feature relative to its machining reference under a named load case. Check units, material data, load directions, contact assumptions and mesh sensitivity. Compare an idealised subcase with a hand calculation.

Report displacement in the functional direction, not only the largest colourful contour. A maximum elsewhere may be harmless, while a smaller movement at a locating feature consumes the tolerance allocation.

Plan a meaningful physical check

Use a suitable controlled test arrangement developed by the responsible engineer. Measure the relevant displacement relative to the intended reference while applying known loads within the safe limits of the equipment. Include unloading to investigate recovery or slip.

The measurement system must resolve the effect of interest. A noisy displacement reading cannot establish a small compliance. Record the load application point, direction and uncertainty, and keep the test away from unqualified machine operation.

The simulation guidance provides broader context for model verification. Here the objective is agreement between a physical load path and a functional displacement.

Exercise: evaluate an intermediate support span

Using the original 10 mm thick bridge, calculate the ideal deflection at an 80 mm span.

The result is 0.0375 × (80/100)³ = 0.0192 mm, still above the provisional 0.010 mm allocation. Explain why adding a compliant contact to the path would not improve that result. Its displacement adds to the total under the assumed series model.

The next chapter carries the fixture geometry into the toolpath and complete production cycle, where an apparently harmless obstruction can invalidate an otherwise sound concept.