Bearing Vibration Monitoring: ISO 10816, BPFO/BPFI Frequencies, Equipment & Practice

Bearing vibration monitoring is a predictive maintenance technique that uses accelerometer sensors and frequency spectrum analysis to detect early-stage bearing defects before unplanned machine shutdowns occur.

According to SKF Condition Monitoring, vibration monitoring can detect bearing defects 3–6 months before actual failure, reducing unplanned maintenance costs by up to 50% and significantly extending equipment service life. The ISO 10816 standard (superseded by ISO 20816) defines vibration severity limits for industrial machinery classes, while characteristic frequency formulas (BPFO, BPFI, BSF, FTF) enable precise identification of the damaged bearing component. This article covers the full spectrum of theory and practice — from spectral analysis fundamentals, calculation formulas, and equipment comparison, to practical interpretation of measurement results — drawing on documentation from SKF, FAG/Schaeffler OPTIME, NSK Condition Monitoring, and ISO 10816-3:2009.

Why Vibration Monitoring Is the Backbone of Predictive Maintenance

In modern bearing maintenance strategy, vibration monitoring holds a central role for three reasons:

Early detection: Vibration is the first signal to appear when a bearing begins to degrade — well before temperature rise, audible noise, or grease leakage. A micro-spall as small as 0.1 mm on a raceway generates characteristic impact pulses that an accelerometer can detect.

Fault localization: FFT spectrum analysis distinguishes between outer race damage (BPFO), inner race damage (BPFI), rolling element damage (BSF), and cage damage (FTF). This distinction determines whether the bearing must be replaced immediately or can wait for the next planned shutdown.

Severity quantification: ISO 10816 divides vibration into four zones (A, B, C, D) corresponding to severity levels, enabling maintenance engineers to make data-driven decisions rather than relying on intuition.

A steel mill in southern Vietnam deployed an online vibration monitoring system across 120 spherical roller bearings on its hot rolling line. Within the first 18 months, the system identified 14 bearings at stage-2 degradation — allowing planned replacements during scheduled shutdowns instead of emergency stops. The result: 60% reduction in unplanned downtime, with estimated annual savings of 2.8 billion VND.

Theory: Bearing Characteristic Frequencies

Each bearing component, when damaged, produces periodic impact pulses at a specific characteristic frequency. The four primary frequencies are:

BPFO — Ball Pass Frequency Outer Race

BPFO = (Z/2) × (1 - Bd/(Pd) × cos α) × (RPM / 60)

Where:

• Z = number of rolling elements (balls or rollers)
• Bd = rolling element diameter (mm)
• Pd = pitch diameter (mm)
• α = contact angle (degrees)
• RPM = shaft rotational speed (rev/min)

BPFO is the most commonly observed frequency because the outer race bears the full load zone and is most susceptible to fatigue spalling.

BPFI — Ball Pass Frequency Inner Race

BPFI = (Z/2) × (1 + Bd/(Pd) × cos α) × (RPM / 60)

BPFI is typically 20–30% higher than BPFO. Inner race defects are harder to detect because the inner race rotates with the shaft, causing the signal to be amplitude-modulated at the shaft speed.

BSF — Ball Spin Frequency

BSF = (Pd / (2 × Bd)) × (1 - (Bd/(Pd) × cos α)²) × (RPM / 60)

BSF appears when the rolling element surface is spalled, cracked, or flat-spotted. It is typically accompanied by the 2× BSF harmonic.

FTF — Fundamental Train Frequency

FTF = (1/2) × (1 - Bd/(Pd) × cos α) × (RPM / 60)

FTF is very low (typically 0.35–0.45 × RPM/60). Cage damage is rare but dangerous — it often leads to catastrophic bearing failure.

Worked Example: 6308 Bearing on a 1,460 RPM Motor

Parameters for the 6308 bearing per SKF Product Data:

Calculating BPFO:

BPFO = (8/2) × (1 - 15.875/65 × cos 0°) × (1460/60)
     = 4 × (1 - 0.2442) × 24.333
     = 4 × 0.7558 × 24.333
     = 73.55 Hz

Calculating BPFI:

BPFI = (8/2) × (1 + 15.875/65 × cos 0°) × (1460/60)
     = 4 × (1 + 0.2442) × 24.333
     = 4 × 1.2442 × 24.333
     = 121.12 Hz

Calculating BSF:

BSF = (65 / (2 × 15.875)) × (1 - (15.875/65 × cos 0°)²) × (1460/60)
    = 2.0472 × (1 - 0.0596) × 24.333
    = 2.0472 × 0.9404 × 24.333
    = 46.84 Hz

Calculating FTF:

FTF = (1/2) × (1 - 15.875/65 × cos 0°) × (1460/60)
    = 0.5 × 0.7558 × 24.333
    = 9.19 Hz

When measuring vibration on an electric motor and observing a peak at 73.5 Hz or its harmonics 2×, 3× — this is a clear signature of outer race fatigue spalling.

ISO 10816 Vibration Severity Limits by Machine Class

ISO 10816 (now ISO 20816) classifies vibration severity by machine group and support condition (rigid or flexible). Values are measured as velocity RMS (mm/s) in the 10–1,000 Hz frequency range.

Vibration Severity Table by Machine Class

Zone definitions:

• Zone A: Newly commissioned or just-overhauled machinery — best achievable condition
• Zone B: Normal long-term operation, acceptable indefinitely
• Zone C: Maintenance action should be planned — continuous operation at this level is not advisable
• Zone D: Danger — shut down or reduce load immediately, risk of severe damage

At a cement plant in Hai Duong province, a 250 kW ID fan motor (Class III) measured 5.2 mm/s — falling within Zone C. Spectrum analysis revealed a BPFO peak at 78 Hz on the 22322 E spherical roller bearing. A replacement was scheduled during the kiln shutdown two weeks later — avoiding an emergency stop estimated at 800 million VND in losses.

Four Stages of Bearing Degradation

Based on research from SKF Bearing Damage and field experience across Vietnamese industrial facilities, bearing degradation follows four distinct stages:

Critical note: Stage 1 is only detectable with ultrasonic or envelope demodulation equipment — standard velocity meters are insufficient. This is why many facilities in Vietnam detect damage only at stages 3 or 4 when relying on "listening by ear" or basic measurement instruments.

Spectrum Analysis Methods

FFT (Fast Fourier Transform) Analysis

FFT converts the vibration signal from the time domain to the frequency domain, enabling identification of abnormal vibration sources.

Standard measurement setup:

• Frequency range: 0–1,000 Hz (sufficient for most industrial bearings running below 3,000 rpm)
• Resolution: ≥800 lines — 1,600 or 3,200 lines is preferable
• Window function: Hanning window
• Averaging: ≥4 samples (linear averaging)
• Units: mm/s RMS for velocity, m/s² peak for acceleration

FFT Peak Interpretation Guide

Envelope Analysis (Demodulation)

Envelope analysis is the most powerful tool for detecting stages 1 and 2 when the damage signal is still very weak:

1. High-pass filter at 5–20 kHz — where the bearing's natural resonance frequencies lie
2. Envelope detection (demodulation) — removes the high-frequency carrier, retaining the amplitude envelope
3. FFT on the envelope — the resulting spectrum shows peaks at BPFO, BPFI, BSF, and FTF far more clearly than the standard spectrum

Per the FAG/Schaeffler Application Note, envelope analysis detects defects 2–3 months earlier than velocity FFT on the same bearing.

At a paper mill in Binh Duong province, a maintenance engineer detected an envelope peak at 73 Hz (BPFO) on a 23124 CC/W33 spherical roller bearing in a dryer roll — while the overall velocity was still only 2.1 mm/s (Zone B per ISO 10816). Three weeks later, velocity had risen to 6.8 mm/s (Zone C). The early detection allowed the plant to prepare a replacement bearing and schedule the swap during the weekly maintenance window — preventing an unplanned line stoppage.

Vibration Monitoring Equipment Comparison

Comparison of Equipment Commonly Used in Vietnam

Choosing Equipment by Plant Scale

Small plants (<50 measurement points): Start with a mid-range handheld unit (FAG Detector III or Fluke 810). Cost: 3,000–5,000 USD. Measure on a route basis weekly or monthly. Suited to mechanical workshops, woodworking plants, packaging facilities.

Medium plants (50–200 measurement points): Combine an advanced handheld analyzer (SKF Microlog) for critical equipment with IoT wireless sensors (Schaeffler OPTIME or SKF Enlight AI) for the rest. Total investment: 40,000–100,000 USD. Suited to steel mills, cement plants, medium-scale textile operations.

Large plants (>200 measurement points): Fixed online systems (SKF Multilog IMx-S) for critical assets plus wide-scale IoT. SCADA/CMMS integration. Total investment: 100,000–500,000 USD. Suited to power plants, chemical plants, oil refineries.

Practical Measurement Guide for Plant Engineers

Sensor Placement

The sensor must be placed directly on the bearing housing, on a flat machined surface, perpendicular to the shaft centerline. Three measurement directions:

• V (Vertical): Most sensitive to imbalance
• H (Horizontal): Most sensitive to misalignment
• A (Axial): Most sensitive to angular misalignment and thrust bearing faults

Avoid placing sensors on fan guards, protective covers, or thick paint surfaces — signal attenuation can reach 70–90%.

Standard Measurement Procedure

1. Machine must be at steady-state operation — normal speed and load, running for at least 15 minutes
2. Clean the measurement point — use a grinding stone or wire brush to create a flat ∅ 20 mm surface
3. Mount the sensor: magnet (most common, frequency response up to 2 kHz), adhesive stud (up to 5 kHz), or threaded stud (most accurate, up to 10 kHz+)
4. Measure in all 3 directions at each bearing housing
5. Record data with: machine ID, position, date/time, load, speed
6. Compare against baseline and ISO 10816 thresholds

Establishing Baseline

Baseline is the vibration level when the bearing is new or just replaced. Measure baseline within 24 hours after new installation:

• Measure at 50%, 75%, and 100% load
• Save FFT spectrum and overall values
• Set alert threshold = 2.5× baseline, danger threshold = 10× baseline (per ISO 13373-1 guidance)

Real-World Case Studies from Vietnamese Factories

Case 1: Cement Plant in Quang Ninh Province

Equipment: 315 kW clinker cooler fan, 22328 E/VA405 spherical roller bearing, 980 rpm.

Issue: Monthly route measurement using an SKF Microlog showed vibration velocity at the DE (drive end) bearing housing rising from 2.4 mm/s (baseline) to 4.8 mm/s over 3 months.

Analysis: FFT spectrum showed a clear peak at 57 Hz — matching the calculated BPFO frequency for the 22328 E (9 rollers, Pd = 195 mm, Bd = 36 mm). The 2× BPFO harmonic (114 Hz) was also clearly visible. Envelope spectrum confirmed BPFO peak with ±16.3 Hz sidebands (= 1× RPM).

Diagnosis: Outer race fatigue spalling, stage 3.

Action: Bearing replaced during a scheduled kiln shutdown 10 days later. Upon removal, the outer race showed a spall track measuring 25 mm × 8 mm on the raceway — confirming the diagnosis. Root cause: insufficient lubrication (re-lubrication interval was too long).

Case 2: Textile Mill in Binh Duong Province

Equipment: 45 kW spinning machine drive motor, 6312 C3 deep groove ball bearing, 1,465 rpm.

Issue: Schaeffler OPTIME IoT system sent an automatic alert — vibration level increased 3× baseline within 2 weeks.

Analysis: Spectrum downloaded from cloud showed a very high 1× RPM (24.4 Hz) peak in the radial direction, accompanied by 2× RPM. No BPFO/BPFI peaks present.

Diagnosis: Not a bearing defect but rotor imbalance — the coupling was worn.

Action: Coupling replaced and laser-aligned. Vibration dropped from 7.2 mm/s to 1.8 mm/s. The bearing remained in good condition — avoiding an unnecessary replacement (saving approximately 4,500,000 VND + 6 hours of downtime).

Case 3: Steel Mill in Hai Phong

Equipment: Cold rolling line gearbox, 32220 tapered roller bearing + 22222 EK spherical roller bearing, 720 rpm.

Issue: SKF IMx-S online monitoring detected a gradual upward trend on the intermediate bearing housing over 6 weeks.

Analysis: Envelope spectrum showed a peak at BPFI (88 Hz) with strong ±12 Hz sidebands. The 2× BPFI (176 Hz) harmonic was also clear. Acceleration gSE rose from 0.4 gSE to 1.8 gSE. Overall velocity was 3.5 mm/s (Zone B — not yet alarming per ISO 10816).

Diagnosis: Inner race of the 22222 EK beginning to spall, transitioning from stage 2 to stage 3. Envelope analysis detected the issue 4 weeks before the ISO overall threshold would have triggered.

Action: Replacement bearing ordered at competitive pricing from a reputable supplier, scheduled for the following month's maintenance window. Upon disassembly: inner race showed three small spall marks (~5 mm each) evenly distributed — classic signs of surface fatigue from sustained high loading.

Monitoring Specific Bearing Types

Spherical Roller Bearings (SRB)

Spherical roller bearings (22205–22340 series) have two rows of rollers — producing more complex vibration spectra than deep groove ball bearings. BPFO and BPFI can differ between the two rows if the load distribution is uneven. C3 or C4 clearance affects baseline vibration amplitude — larger clearance produces higher baseline vibration that does not necessarily indicate damage.

Note: Always reference exact characteristic frequencies from manufacturer databases (SKF Bearing Calculator, Schaeffler medias) rather than using approximate formulas for SRBs.

Tapered Roller Bearings

Tapered roller bearings are typically mounted in pairs with preload. Axial clearance significantly affects vibration amplitude:

• Excessive clearance: 1× RPM peak increases in the axial direction
• Insufficient clearance (excessive preload): temperature rises, higher BSF harmonics increase

When monitoring tapered roller bearings in gearboxes, always measure in the axial direction — this is the most sensitive orientation.

Common Measurement Mistakes in Vietnamese Plants

1. Wrong sensor location: Mounting on a fan guard or protective cover instead of the bearing housing. Signal attenuation reaches 70–90%, making BPFO/BPFI peaks invisible.

2. Measuring during transients: Taking readings immediately after startup or during load changes. Results are not repeatable.

3. No baseline recorded: Without a reference value, all numbers are meaningless. Baseline must be measured immediately after new bearing installation.

4. Relying solely on overall values: Many plants read only overall mm/s without analyzing the spectrum. An overall value in Zone B does not mean the bearing is healthy — characteristic peaks may already be present on the envelope spectrum.

5. Misidentifying vibration source: A 1× RPM peak is not always caused by the bearing — most often it indicates imbalance or misalignment. Replacing a bearing without confirming characteristic frequencies is wasteful.

6. Insufficient frequency range: Measuring only 0–200 Hz on high-speed ball bearings (>3,000 rpm) will miss BPFI peaks at higher frequencies.

Integrating Vibration Monitoring into Maintenance Systems

Route-Based Monitoring

Suitable for plants beginning their predictive maintenance journey:

• Create a critical equipment list
• Assign measurement point codes for each bearing housing (e.g., PP-FAN01-DE-V = PP fan, drive end bearing, vertical direction)
• Establish measurement intervals: weekly for critical equipment, monthly for secondary
• Use management software (SKF @ptitude, Schaeffler ConditionAnalyzer) for trend tracking

Online + IoT Monitoring

The evolutionary step beyond route-based monitoring:

• Sensors permanently installed at each measurement point
• Data automatically transmitted to server/cloud at regular intervals (hourly to daily)
• AI algorithms automatically detect abnormal trends and send alerts
• CMMS (Computerized Maintenance Management System) integration for automatic work order generation

At a packaging manufacturer in Dong Nai province, 80 Schaeffler OPTIME sensors were deployed across all motors and gearboxes on the flexographic printing lines. Total investment: approximately 32,000 USD (400 USD/point). In the first year, the system detected 6 bearings at stage-2 degradation, 2 cases of severe misalignment, and 1 motor with electrical issues. ROI was achieved within 8 months.

The Link Between Vibration Monitoring and Lubrication

Vibration monitoring and bearing lubrication are closely interconnected:

• Grease starvation: Ultrasonic acceleration (dBμV) rises before characteristic peaks appear. Apply grease using the G = 0.005 × D × B formula while monitoring ultrasonic values — when the value drops back to baseline, sufficient grease has been applied.
• Over-greasing: Temperature rises, overall low-frequency vibration increases, characteristic peaks remain unclear.
• Wrong or degraded grease: Abnormal BSF peaks, noise floor rises uniformly across the spectrum.

Combining vibration monitoring with condition-based lubrication is a significant advancement — lubricating when needed rather than on a fixed schedule. A chemical plant in Long An province applied this method across 45 centrifugal pumps, reducing grease consumption by 40% while extending average bearing life by 30%.

Technology Trends in Vibration Monitoring

MEMS Sensors and IoT

MEMS (Micro-Electro-Mechanical Systems) sensor costs continue to decline — enabling deployment of hundreds of measurement points at just 300–600 USD per point. Solutions such as Schaeffler OPTIME, SKF Enlight AI, and NTN Condition Monitoring use wireless sensors communicating via Bluetooth or LoRa, with 3–5 year battery life and zero wiring requirements.

AI and Machine Learning Analytics

Machine learning algorithms trained on millions of data points from failed bearings enable:

• Automatic anomaly detection without manually configuring characteristic frequencies
• Remaining Useful Life (RUL) prediction
• Automatic vibration source classification (bearing vs. imbalance vs. misalignment)

However, AI does not replace the analysis engineer — it is a support tool. Experience in understanding the machine, the process, and the maintenance history remains the decisive factor.

Digital Twins and Life Prediction

Digital twin models combine real-time vibration data with physics-based models (load calculation, thermal, lubrication) to predict remaining bearing life more accurately. SKF and Schaeffler are currently deploying these services for major industrial clients across Southeast Asia.

Key Takeaways

• Vibration monitoring detects bearing defects 3–6 months before actual failure — enabling planned replacements instead of emergency shutdowns.
• The four characteristic frequencies BPFO, BPFI, BSF, and FTF pinpoint exactly which bearing component is damaged — preventing unnecessary or premature replacements.
• ISO 10816 divides vibration into four zones (A–D) by machine class — engineers must know their machine class to apply the correct thresholds.
• Envelope analysis detects stages 1–2 defects 2–3 months earlier than velocity FFT — investing in envelope-capable equipment is mandatory for serious predictive maintenance.
• IoT wireless sensors (300–600 USD/point) have made wide-scale monitoring feasible for small and medium Vietnamese plants — there is no longer any justification for relying solely on audible inspection.
• Combining vibration monitoring with condition-based lubrication simultaneously reduces grease consumption and extends bearing service life.
• Always analyze FFT/envelope spectra rather than relying solely on overall values — a normal overall reading does not guarantee the bearing is defect-free.
• Recording a baseline after new installation is mandatory — without a baseline, there is no predictive maintenance.