When a rolling-element bearing starts to fail, it vibrates at frequencies that are not simple multiples of shaft speed but are precisely predictable from the bearing's own geometry. Each part of the bearing, the outer race, the inner race, the rolling elements, and the cage, generates its own characteristic frequency as the defect on it is repeatedly struck. These are the bearing defect frequencies, and because each is unique, matching a peak in a vibration spectrum to one of them tells you not just that the bearing is failing but exactly which element is failing.
Bearing defect frequency in one line: A bearing defect frequency is the rate at which a defect on a specific part of a rolling-element bearing is struck as the bearing turns, calculated from the bearing's geometry and the shaft speed. The four characteristic frequencies are BPFO for the outer race, BPFI for the inner race, BSF for the rolling elements, and FTF for the cage. A spectrum peak at one of these frequencies identifies which element is defective.
A rolling-element bearing has four parts that can develop defects, and each produces a distinct frequency. The Ball Pass Frequency Outer race, BPFO, is the rate at which rolling elements pass a fixed point on the outer race, so a spall on the outer race is struck at BPFO. The Ball Pass Frequency Inner race, BPFI, is the rate at which elements pass a point on the inner race, which rotates with the shaft, so an inner-race defect shows at BPFI. The Ball Spin Frequency, BSF, is the rate at which a single rolling element rotates about its own axis, so a defect on a ball or roller is struck at BSF, usually seen at twice BSF because the defect contacts both races per spin. The Fundamental Train Frequency, FTF, is the rotation rate of the cage that carries the elements, and it appears when the cage itself is damaged.
The reason these frequencies are so useful diagnostically is that they are non-synchronous, they are not integer multiples of shaft speed. Imbalance, misalignment, and looseness all produce peaks at exact multiples of running speed, 1x, 2x, 3x. Bearing defect frequencies land at fractional, non-integer multiples determined by the bearing's internal geometry, so they stand apart from the shaft-related peaks in the spectrum. A peak that does not line up with a whole multiple of running speed but does match a calculated bearing frequency is a strong indicator of a genuine bearing defect rather than another fault.
In practice the defect frequencies rarely appear alone. A developing race defect typically shows the fundamental defect frequency plus a train of its harmonics, and inner-race defects usually carry sidebands spaced at shaft speed, because the defect moves in and out of the load zone once per revolution as the inner race turns. Recognizing not just the fundamental but this harmonic-and-sideband pattern is part of confirming that a peak is truly a bearing frequency and gauging how far the damage has progressed.
Each defect frequency is derived from the bearing's physical dimensions and the shaft speed. The inputs are the number of rolling elements, the rolling element diameter, the pitch diameter of the bearing, the contact angle, and the relative speed between the inner and outer races, which for a stationary housing is just the shaft speed. From these, standard formulas give BPFO, BPFI, BSF, and FTF. The formulas capture the geometry directly: more rolling elements and a larger element-to-pitch-diameter ratio raise the ball-pass frequencies, and the contact angle shifts the values as the bearing's load path changes.
A convenient consequence of the geometry is that BPFO and BPFI together relate simply to the element count. For a given bearing, the sum of BPFO and BPFI, expressed in orders of shaft speed, equals the number of rolling elements, and BPFO is somewhat less than half that count in orders while BPFI is somewhat more than half. That relationship is a quick sanity check on calculated values and a reminder that these frequencies depend on the specific bearing, not on the machine in general. This is why analysts need the bearing part number or its geometry to compute the correct target frequencies.
Because the frequencies scale with shaft speed, they are often expressed and stored in orders, multiples of running speed, rather than in fixed hertz. Storing them as orders means the same set of values applies whether the machine runs fast or slow, and the monitoring system multiplies by the current speed to place them on the spectrum. Many analysis tools and bearing databases provide the order values directly from the bearing designation, so the calculation reduces to looking up the bearing and applying its known multipliers to the measured speed.
The diagnostic payoff is precision: because each element has its own frequency, a matched peak names the culprit. A peak at BPFO with harmonics says the outer race is spalling; a peak at BPFI with shaft-speed sidebands says the inner race is damaged; energy at twice BSF says a rolling element is defective; a peak at FTF points to a cage problem. This is far more actionable than a generic high-vibration alarm, because it tells maintenance not only that the bearing is failing but how, which informs urgency and whether the failure mode is one that can propagate quickly.
This matching can be automated, which is what makes it practical across many machines. If the bearing geometry is known, the defect frequencies can be pre-computed as orders and monitored automatically. A monitoring system defines bands at BPFO, BPFI, BSF, and FTF, scaled by the live shaft speed, and tracks the amplitude in each. Merobix can trend those band amplitudes for each monitored bearing, so a rising BPFO band flags an outer-race defect specifically, and an operator sees the affected element identified rather than just a rising overall number. For remote and unmanned sites, this pushes expert-level bearing diagnosis out to machines no analyst visits routinely.
Trending the defect-frequency bands over time turns identification into prognosis. The first appearance of a defect frequency, often picked up first through envelope analysis, marks the onset; the growth of its amplitude and the emergence of more harmonics track the progression toward failure. Historized band trends let the platform establish a clean baseline and raise a confident alarm when a specific bearing frequency emerges and climbs, and correlating that with speed and load confirms the fault is real and tied to operation. The result is a maintenance decision anchored to a named failing element and a measured rate of deterioration.
BPFO is Ball Pass Frequency Outer race, the rate at which rolling elements pass a point on the outer race. BPFI is Ball Pass Frequency Inner race, the equivalent for the inner race. BSF is Ball Spin Frequency, the rotation rate of a rolling element about its own axis. FTF is Fundamental Train Frequency, the rotation rate of the cage. Each corresponds to a defect on a different part of the bearing.
The defect frequencies are calculated from the bearing geometry, the number of rolling elements, the element and pitch diameters, and the contact angle, together with the shaft speed. Standard formulas produce BPFO, BPFI, BSF, and FTF from these inputs. Because they scale with shaft speed, the values are often expressed as orders, multiples of running speed, so they apply at any operating speed once multiplied by the current shaft speed.
They are set by the bearing's internal geometry rather than by the shaft alone, so they land at fractional, non-integer multiples of running speed. This distinguishes them from imbalance, misalignment, and looseness, which produce peaks at exact whole multiples like 1x and 2x. A peak that does not line up with a whole multiple of running speed but does match a calculated bearing frequency is a strong indicator of a real bearing defect.
Merobix reads your field devices into a cloud SCADA - the real thing behind these terms, live in days from any browser.