How a Base and a Foundation Change the Machine on Them

Why this matters

Two identical units, same model, same rotor, same balance grade, same drive, installed by the same crew in the same week. One is quiet. One shakes hard enough that the tenant below files a complaint. Nothing is wrong with either machine, and no amount of work on either machine will change the outcome, because the thing that vibrates is not the machine. It is the machine plus its base plus whatever that base stands on, and those three parts together decide the frequency the assembly prefers and how much it moves at any given force. This article takes one rule, applies it to the same fan in two locations, and gets opposite answers.

Before you get under or beside a unit that is being set

If a unit is being lifted or set during this work, nobody positions themselves under it or reaches under it while it hangs, and rigging is a qualified activity with rated gear, not something added to a service call. Lock and tag the disconnect and confirm the rotor has stopped rather than slowed before any indicator, hand or hammer goes near the assembly, per 29 CFR 1910.147 and its verification-of-isolation requirement. Most of this work happens on roofs and mechanical-room platforms, so fall protection is a duty rather than a judgment call: service work on an existing building sits under general industry at 29 CFR 1910.28(b)(1), while the same task performed as part of a construction project sits under 29 CFR 1926.501, and the two Parts trigger at different heights, so establish which one applies before you step out. Running readings come from outside an intact guard with a magnetic-base sensor on a stationary housing, since 29 CFR 1910.219 does not pause for data collection.

The vibrating body is the assembly, not the machine

Natural frequency rises with stiffness and falls with mass, and both of those belong to the whole load path: rotor, shaft, bearings, housings, rails, base, isolators, curb, deck or slab, and the beams under it. Change any element of that path and the frequency moves, even though nothing on the equipment nameplate changed.

That is why a machine can pass every check on the truck and still be wrong in a specific place, and why "it worked fine at the last site" is not evidence of anything. It also means a resonance can be created by work that nobody logged as mechanical: adding a second unit to a shared curb, removing a wall that was bracing a platform, replacing a deck, or bolting a new duct run to the same beam.

The gate

Establish the supporting structure's natural frequency, then check every strong excitation the machine emits against it. If any excitation lands within 15 to 20 percent of that frequency, the structure is part of the machine's behavior and no work on the machine will fix it.

Two ways to get the number, and they are for different moments.

On an existing installation, get the frequency directly: strike the structure with a soft-face hammer at the point of maximum motion, machine locked out and stopped, and read where it rings. This is faster and it measures the real assembly rather than a model of it.

In the planning case, before the unit is set, work from deflection. Natural frequency in cycles per minute is about 187.8 divided by the square root of the static deflection in inches under the applied mass. That expression is derived for a single mass on a linear spring, so applied to a deck panel or a beam it is a lumped approximation that gets the order of magnitude and the direction of change right and should not be quoted as a design figure. On a replacement you can often measure the deflection directly: set a dial indicator at the loaded point referenced to a column or bearing wall that is off the affected span, then read the difference as the crane takes the weight and sets it back. Set the indicator BEFORE the pick and read it from outside the fall zone: nobody positions themselves under a suspended load or reaches under it to adjust an instrument, and a reading is not worth standing in the drop path for.

Case A: the same fan on a curb over a bearing wall

A belt-driven rooftop fan, wheel at 890 rpm, sitting on a curb that lands over a bearing wall.

Measured deflection at the curb under the unit's weight: 0.010 in. Natural frequency: 187.8 divided by the square root of 0.010 = 1,878 cpm.

Ratio of running speed to structure frequency: 890 / 1,878 = 0.47. That is well below the peak, in the region where the structure behaves as effectively rigid. With light damping, on the order of 3 percent of critical for a bolted steel and masonry path and a figure that is assumed rather than measured, the magnification works out near 1.3.

The machine's residual unbalance produces the force it always produces, the structure barely amplifies it, and the installation reads as quiet.

Case B: the same fan mid-span on a long-span deck

Identical unit, identical rotor, set mid-span on a long-span steel deck.

Measured deflection at the same point of the unit under its weight: 0.045 in. Natural frequency: 187.8 divided by the square root of 0.045 = 885 cpm.

Ratio: 890 / 885 = 1.006. That is essentially on the peak. A bare steel deck is very lightly damped, so at around 3 percent of critical the magnification works out near 16.

Comparing the two: 16 against 1.3 is a factor of about twelve. The same wheel, the same residual unbalance, the same force, and twelve times the motion, because the deflection under the unit went from ten thousandths to forty-five thousandths of an inch. That is the entire difference between the two installations.

Note what the deflection figures are doing in that comparison: they are both measured under the same load in the same way, so the comparison is like for like. If you take one from a structural drawing and the other from a field measurement, you are comparing a design value against an as-built one, and the two routinely differ.

What the fix is in Case B. Not the fan. Either raise the deck's frequency by adding support under the span, which means a post or a beam that SHORTENS the span rather than plate added alongside it. The reason is not the square-root relation, since plate adds stiffness too: bending stiffness goes as one over the span cubed, so halving the span beats almost anything you can bolt to it, and added plate brings added mass that works against you; or move the unit to a stiffer point, typically over a beam line or a wall; or change the fan speed enough to clear the mode by 15 to 20 percent, which at 885 cpm means dropping to about 750 rpm and losing roughly 16 percent of the airflow. The first two are changes to a building's load path and belong to a structural engineer, not to a field decision made on the roof, and where new steel is welded in, the fume is an inhalation hazard in its own right: coated or galvanized members release zinc oxide fume and stainless releases hexavalent chromium, calling for local exhaust and respiratory protection under a written 29 CFR 1910.134 program, with 29 CFR 1910.1026 where chromium-bearing metal is in the joint. All three are structural or system decisions. None of them is a mechanical repair.

The two bodies you are tuning want opposite things

This is where people get turned around, because both are called "vibration" and the advice is inverted.

The machine-on-isolators system wants a LOW natural frequency, far below running speed, so the ratio is high and transmission into the building is small. Softer, more deflection, better.

The supporting structure wants a HIGH natural frequency, well above every strong excitation, so it behaves as rigid. Stiffer, less deflection, better.

Confusing them produces the classic bad call: a machine transmitting into a soft deck gets softer isolators, which is correct for the mount system and does nothing about the deck mode that is doing the amplifying. The deck was never the isolation path; it was the thing being excited.

What an inertia base actually buys, and what it does not

An inertia base is a mass, usually a concrete-filled frame, sitting between the machine and its isolators, and the folk explanation is that the added mass lowers the natural frequency. That explanation is wrong in the case it is most often used in, and the correction matters.

Natural frequency falls with added mass only if stiffness stays the same. When an inertia base is added, the isolators are re-selected to carry the heavier load at their rated deflection, and since the frequency depends on the deflection, the new system sits at the same natural frequency as the old one. You added mass and stiffness together and the ratio did not move.

What the inertia base genuinely buys is three things:

• Amplitude. Above resonance, the same force moves more mass less. A machine plus a base weighing one and a half times the equipment is two and a half times the mass, so the assembly's own motion under the same rotating force drops to about 40 percent of what it was. The building sees the same transmitted fraction; the machine itself moves less, and everything mounted on it lives longer.
• Rocking resistance. It lowers the composite center of gravity relative to the mount plane and widens the effective footprint, which suppresses the rocking modes that plague tall, top-heavy machines on soft mounts.
• Alignment stability. A driver and a driven machine on a single rigid base cannot move relative to each other, which protects a coupling alignment that a flexing rail assembly would slowly destroy.

Common practice sizes an inertia base at roughly one to two times the equipment weight for rotating machinery and heavier for reciprocating equipment, but that is a starting convention; the isolation supplier or the equipment manufacturer sizes it for the actual machine and mount selection.

The base whose stiffness changes while you tighten it

One last mechanism, because it produces readings that make no sense. A base is only as stiff as its contact with what it sits on. Three conditions break that:

• A soft foot, where one of the machine's feet does not sit flat on the base, so the stiffness of the assembly changes as the hold-down bolt is torqued and the frame is distorted rather than supported. The library has a dedicated article on why soft foot defeats an alignment; the same condition also makes every vibration reading depend on bolt torque.
• Ungrouted or voided grout under a base. The load path is then through the anchor bolts alone, which is far softer than the design assumed, and the frequency drops accordingly.
• Anchors that have loosened, which is self-accelerating: lower stiffness means more motion, more motion loosens more, and the frequency walks downward over months until it meets an excitation.

The tell for all three is a natural frequency that changes between visits. A structural mode is a stable number. If your impact test rings at one frequency this month and noticeably lower next month with no work done in between, the load path is coming apart and the answer is at the interface, not in the machine.

How to verify you got this right

Measure the same mode from two directions and at two points on the assembly. A structural mode that reads the same from a strike on the rail and a strike on the curb is one body moving; two different frequencies mean two bodies, and you need to know which one your excitation is landing on before you stiffen anything.

After a structural change, re-run the impact test and confirm the frequency moved by roughly what the stiffness change predicted through the square-root relationship. A stiffness addition that produces no frequency shift was added at a point the mode does not move, which is the usual result of bracing at the middle of a span that is already still there and leaving the ends alone.

References

• 29 CFR 1910.147 for isolation, lock and tag, and stopped-rotor verification before impact testing or indicator work on a mounted machine
• 29 CFR 1910.28(b)(1) for general-industry fall protection during rooftop and platform work, and 29 CFR 1926.501 where the same task falls under construction
• 29 CFR 1910.219 for guarding during running measurements on a mounted assembly
• 29 CFR 1910.134 and 29 CFR 1910.1026 for respiratory protection and hexavalent chromium exposure where structural steel is welded during a stiffening change
• Isolation supplier and equipment manufacturer documentation for inertia base sizing and per-corner mount selection
• See related: What a Flexible Mount Does and What It Cannot Do; What Resonance Is and Why It Finds Your Machine; Why Soft Foot Defeats an Alignment