Bearing defects produce vibration at mathematically predictable frequencies determined by the bearing's geometry and shaft speed. Knowing exactly where to look in the vibration spectrum, before a fault appears, is the foundation of condition-based bearing monitoring.
Enter Bearing Parameters
| Fault Component | Frequency (Hz) | Frequency (CPM) | Order (× shaft) | Indicates |
|---|
How Rolling Element Bearing Fault Frequencies Work
Every rolling element bearing produces vibration at specific, calculable frequencies when a surface defect is present. These fault frequencies arise from the bearing's internal geometry. As the shaft rotates, rolling elements pass over a defect on the outer race, inner race, or cage at a rate determined purely by geometry and speed. No estimation is required. The physics is exact.
The outer race fault frequency (BPFO) reflects how often rolling elements impact a stationary defect on the fixed outer ring. Because the outer race does not rotate, this frequency is relatively stable and often the clearest early indicator in a vibration spectrum. The inner race fault frequency (BPFI) is more complex: the inner race rotates with the shaft, so the defect moves through a load zone once per revolution. This produces amplitude modulation of the BPFI frequency at 1x shaft speed. Sidebands spaced at shaft frequency appear around the BPFI peak, which is a reliable diagnostic signature.
The ball spin frequency (BSF) describes how fast each rolling element rotates about its own axis. A defect on a ball or roller produces a lower-energy signal than race defects because rolling elements do not strike the same fixed point on every pass. They precess around the bearing. BSF peaks are often less prominent and may appear at 2x BSF (twice per revolution of the rolling element) due to the double-contact geometry. The fundamental train frequency (FTF) represents the rotational speed of the cage, or the rate at which balls pass a fixed point if the cage has a defect or uneven spacing. FTF is sub-synchronous (always below 1x shaft frequency) and is typically the lowest-amplitude signal of the four.
In practice, bearing fault diagnosis works by comparing your current vibration spectrum against a baseline taken when the bearing was new. An amplitude rise at one of these exact calculated frequencies, particularly if it grows over successive readings, confirms a developing fault at that component. Frequency location tells you which component is damaged; amplitude trending over time tells you how severe the damage is and how fast it is progressing.
Finding Your Bearing Geometry Inputs
All four inputs (d, D, N, and θ) must come from the bearing manufacturer's published datasheet for your specific bearing designation. Do not measure these values from the physical bearing or estimate them from another bearing of a similar size. Small errors compound across all four fault frequencies.
Major manufacturers publish complete bearing geometry data in their engineering handbooks and online catalogs: SKF uses the designation system in their Interactive Engineering Catalogue, Timken publishes geometry coefficients in their bearing tables, NSK and Schaeffler (INA/FAG) provide the same in their product documentation. Search for your bearing's part number (e.g. 6310, 22218, NU 318) on the manufacturer's website and look for the table of internal geometry or "bearing factors." Some manufacturers publish BPFO/BPFI multipliers directly (as fractions per revolution), which you can multiply by your shaft frequency instead of computing from raw geometry.
If you cannot find your exact bearing's datasheet, contact the original equipment manufacturer. Using approximate values introduces frequency errors that reduce diagnostic confidence. A 5% error in ball diameter shifts all four fault frequencies by a corresponding amount.
Bearing Fault Frequency Formulas
These are the standard rolling element bearing fault frequency equations published in SKF, Schaeffler, and ISO documentation. All major vibration analysis software (Emerson CSi, Fluke 810, Pruftechnik OMNITREND, National Instruments DIAdem) implement these identically.
BPFO = (N/2) × f × (1 − (d/D)×cos(θ)) // Ball Pass Frequency, Outer Race (Hz)
BPFI = (N/2) × f × (1 + (d/D)×cos(θ)) // Ball Pass Frequency, Inner Race (Hz)
BSF = (D/(2×d)) × f × (1 − ((d/D)×cos(θ))²) // Ball Spin Frequency (Hz)
FTF = (f/2) × (1 − (d/D)×cos(θ)) // Fundamental Train Frequency / Cage (Hz)
where θ is in radians: θrad = θdeg × π / 180
CPM = Hz × 60 // Convert any frequency to cycles per minute