A 7.5 kW fan on a steel mezzanine floor made the floor shake. The fan ran at 1450 RPM — 24.2 Hz. The steel floor had a natural frequency around 18 Hz. The two weren’t close, so the floor shouldn’t have resonated. But the fan was bolted rigidly to the frame, and the frame transmitted every blade-passing pulse straight into the floor. The fix wasn’t a stiffer floor. It was isolating the fan. The isolators worked — but only after the first set was chosen wrong and made things worse.
The transmissibility rule
An isolator (rubber mount or spring) is a spring between the machine and the floor. It works when the natural frequency of the machine-on-isolators is well below the disturbance frequency. The transmissibility — how much of the vibration passes through — depends on the frequency ratio. At a ratio of disturbance-to-natural of 3:1, only about 12% of the vibration passes. At 2:1, about 33%. At 1:1 (resonance), the isolator amplifies the vibration — transmissibility above 100%. Below the natural frequency, the isolator does nothing useful; the machine sits on the floor through a soft spring and shakes more.
The first attempt used soft rubber mounts that gave a natural frequency of about 22 Hz — almost exactly the fan speed. The floor shook worse. The mounts amplified the fan’s 24.2 Hz pulse into the floor. The correct isolator: spring mounts sized so the machine-on-springs natural frequency was about 8 Hz. The ratio was 3:1, and the floor stopped shaking.
The static deflection rule of thumb
There’s a shortcut that avoids the frequency calculation. The natural frequency of a mass on a spring is set by the static deflection — how much the spring compresses under the machine’s weight. The rule: natural frequency in Hz ≈ 15.8 / √(static deflection in mm). For 8 Hz, that’s a deflection of about 4 mm. For a soft 4.5 Hz isolation (good for 1500 RPM machines), the deflection is about 12.5 mm. If the mount only deflects 1 mm under the machine, the natural frequency is about 15.8 Hz — useless for isolating a 24 Hz disturbance, and dangerous if the disturbance is near that.
The mistake people make: they pick a mount by load rating alone. A rubber mount rated for the machine’s weight might deflect only 1 mm. It isolates nothing. The selection must start with the required deflection, then find mounts that deflect that much under the actual weight. Spring mounts deflect far more than rubber ones — that’s why they’re used for low-frequency isolation. Rubber mounts are for high-frequency disturbances (motors, small pumps) or for noise, not for 1450 RPM machinery that needs real isolation.
What the calculation looks like for the fan
The fan weighed 350 kg, with the motor, on four mounts. Each mount carried about 88 kg. The target natural frequency was 8 Hz, giving a 3:1 ratio against 24.2 Hz. The required static deflection was (15.8/8)² = 3.9 mm. The selected spring mounts were rated for 100 kg each with a deflection of 4.2 mm at rated load. The installed natural frequency was about 7.7 Hz — transmissibility around 11%. The floor vibration dropped from 8 mm/s RMS to 1.2 mm/s RMS.
That 1.2 mm/s was acceptable for the floor and, more importantly, for the process — the mezzanine had a weighing station on it, and the floor shake was corrupting the scale readings. The isolation fixed the scale as a side benefit.
Start-up resonance and the snubber question
Every machine passes through its isolator natural frequency on start-up and coast-down. At 8 Hz natural frequency, the fan spends about two seconds crossing that speed band when starting. For two seconds, the isolators amplify. If the machine has a large unbalance or a big start-up torque, that resonance pass can be violent. Two fixes: snubbers (limiters) that let the mount deflect only so far, and for heavy machines, a temporary rigid tie-down for start-up that’s released at speed. For most machines, the two-second pass at resonance is harmless — the amplification is brief and the machine isn’t at full unbalance during spin-up. For large slow machines (compressors, big fans), the start-up pass deserves a check.
Also: the isolator natural frequency must be checked against the machine’s own running speed harmonics. A 4-pole motor at 1450 RPM has a 2x speed harmonic at 2900 RPM — 48.4 Hz. The isolator at 8 Hz is well below it, fine. But a machine with a strong 1x harmonic at 8 Hz (a slow rotary table, a large drum) would be sitting on an isolator that amplifies it. The check: the disturbance frequency you’re isolating is the lowest strong vibration frequency, and the isolator natural frequency must be below half of that.
Installing isolators wrong
Isolators fail silently in three ways. Bolted solid — the machine is leveled with shims between the mount and the floor, and the shims bridge the spring, shorting it out. The fix: level the machine with the isolator’s leveling screw, not with shims under the mount. Second: the isolator is bolted down so tight the rubber or spring is fully compressed and becomes a rigid block. Third: pipes and conduits connected to the machine carry the vibration straight to the floor. A flexible hose or loop in the pipe is part of the isolation system — a rigid pipe connection undoes everything the mounts do.
Isolators work on frequency ratio, not on load rating. A mount that deflects 1 mm isolates nothing at 24 Hz. Spring mounts deflecting 4 mm give an 8 Hz natural frequency and cut a 24 Hz disturbance to 11%. Choose by deflection, check start-up resonance, and don’t bridge the mounts with shims or rigid pipes. The shaking floor wasn’t a floor problem — it was a missing spring.
