A turbine bearing doesn't fail without warning—it announces its deterioration weeks in advance through vibration signatures that are completely invisible to scheduled inspection rounds but unmistakable to a properly installed vibration sensor. The plants losing millions to unplanned turbine trips are the ones still relying on time-based maintenance intervals instead of the real-time condition data their equipment is already generating. Start monitoring your turbine health in OXmaint today and catch the next bearing failure 4 to 8 weeks before it forces a trip—or book a session with our vibration analytics team to see exactly what your current sensors are already telling you.
BEARING FAILURE PROGRESSION
Every Turbine Bearing Failure Goes Through These 5 Stages — Only Sensors Catch the First Three
Stage 1
Normal Operation
0–20% Life Used
Baseline vibration within normal envelope
Sensor Baseline
Stage 2
Sub-Surface Fatigue
4–8 Weeks Before Failure
High-frequency ultrasonic signals appear in bearing race
AI Detects First
Stage 3
Surface Defect Growth
2–4 Weeks Before Failure
Sideband frequencies emerge around bearing defect frequency
Alert Generated
Stage 4
Advanced Wear
Days Before Failure
Broadband noise floor rises, harmonics multiply across spectrum
Audible to Experienced Ear
Stage 5
Catastrophic Failure
Unplanned Outage
Unit trips, secondary damage to shaft, seals, and adjacent components
$2M+ Event
WHAT SENSORS MEASURE
The Vibration Parameters That Predict Turbine Bearing Failures
Overall Vibration (RMS)
10 Hz – 1,000 Hz
The broadband energy baseline. Rising RMS over weeks indicates progressive degradation even before specific fault frequencies appear. First alert threshold: 25% above established baseline.
Detects: General deterioration, imbalance, misalignment
Bearing Defect Frequency (BPFO/BPFI)
Calculated per bearing geometry
Unique spectral signatures tied to specific bearing races. When energy at Ball Pass Frequency Outer (BPFO) or Inner (BPFI) exceeds threshold, it means that race has a developing defect.
Detects: Race spalling, pitting, roller element damage
High-Frequency Envelope (HFE)
5 kHz – 40 kHz
The earliest detectable signal of bearing fatigue—micro-stress releases in bearing steel create ultrasonic pulses weeks before visible damage. HFE is the Stage 2 detection method.
Detects: Sub-surface fatigue before surface damage begins
1× and 2× Running Speed
Synchronous with shaft RPM
Energy at 1× RPM indicates rotor imbalance; energy at 2× indicates misalignment or looseness. Both cause accelerated bearing wear if unaddressed—catching them early prevents secondary failure.
Detects: Rotor imbalance, shaft misalignment, mechanical looseness
Temperature Correlation
Bearing housing surface °C
Bearing temperature rise lags vibration changes by days to weeks — making it a late-stage indicator alone. Combined with vibration, temperature correlation confirms severity and rate of progression.
Detects: Lubrication failure, heavy wear, friction-driven degradation
Kurtosis & Crest Factor
Statistical signal metrics
Statistical measures of the "peakiness" of the vibration signal. A rising kurtosis value on a previously smooth signal indicates the appearance of impact events—the first physical sign of bearing damage.
Detects: Impact-type bearing defects, early spall initiation
SENSOR PLACEMENT
Where Vibration Sensors Go on a Power Plant Turbine
Sensor placement determines whether you catch failures early or late. These are the validated measurement points for steam and gas turbines in power generation service.
| Measurement Location | Measurement Axis | Priority Frequency Band | Primary Defect Detected | Typical Lead Time |
|---|---|---|---|---|
| Drive-End Bearing Housing | Radial H + V, Axial | 10 Hz–20 kHz | Bearing race fatigue, imbalance | 4–8 weeks |
| Non-Drive-End Bearing | Radial H + V | 10 Hz–20 kHz | Bearing wear, looseness | 3–6 weeks |
| HP Turbine Bearing (#1, #2) | Radial + Axial | 100 Hz–40 kHz | Blade rub, shaft bow, bearing damage | 2–5 weeks |
| LP Turbine Bearing (#3, #4) | Radial H + V | 10 Hz–10 kHz | Imbalance, misalignment, looseness | 3–7 weeks |
| Generator Drive-End | Radial H + V, Axial | 10 Hz–20 kHz | Coupling misalignment, bearing wear | 4–8 weeks |
| Gearbox (if applicable) | Radial H + V | 500 Hz–40 kHz | Gear tooth wear, bearing failure | 2–6 weeks |
| Pump/Fan Bearing | Radial H + V | 10 Hz–10 kHz | Impeller imbalance, bearing wear | 3–5 weeks |
CONNECT YOUR SENSORS TO INTELLIGENCE
Raw Vibration Data Is Worthless Without the AI to Interpret It. OXmaint Provides Both.
OXmaint ingests vibration data from your existing ICP accelerometers, wireless sensors, or online monitoring systems and applies trained AI models to flag bearing fault frequencies, track progression rates, and generate prioritized maintenance alerts weeks before failure.
AI ANALYTICS PIPELINE
How Vibration Data Becomes a Maintenance Work Order in OXmaint
01
Continuous Data Capture
Accelerometers sample vibration at 25,600+ samples per second across all measurement points. OXmaint accepts data from wired ICP sensors, wireless IoT nodes, and existing online monitoring systems via OPC-UA, Modbus, or direct API.
▶
02
Spectral Decomposition
Fast Fourier Transform (FFT) converts raw time-domain signals into frequency spectra. OXmaint's AI compares live spectra against equipment-specific baseline profiles, hunting for bearing defect frequencies, sideband growth, and statistical anomalies.
▶
03
Fault Classification
Detected anomalies are classified by fault type (imbalance, misalignment, bearing wear, looseness) and severity level. The AI cross-references temperature trends, load data, and historical failure patterns to filter noise and confirm genuine degradation signals.
▶
04
Maintenance Work Order
Confirmed faults automatically generate prioritized maintenance alerts with fault type, severity, affected component, trend data, and recommended action. OXmaint creates the CMMS work order, assigns the right technician, and schedules the repair within your maintenance calendar.
PROVEN OUTCOMES
What Vibration-Based Predictive Maintenance Delivers in Real Plants
87%
of turbine bearing failures are detectable by vibration analysis 3+ weeks before catastrophic failure
12:1
average ROI on vibration monitoring programs — for every $1 spent on sensors and analytics, plants recover $12 in avoided downtime costs
$3.4M
Avoided per prevented major turbine failure
Combining lost generation, emergency parts (40% premium), replacement power, and regulatory penalties avoided when a bearing replacement is scheduled vs emergency
68%
Reduction in unplanned turbine outages
Plants with continuous vibration monitoring and AI analytics report forced outage frequency dropping by two-thirds within the first full year of operation
30 Days
Faster parts procurement on average
Early detection gives procurement teams enough lead time to source replacement bearings at standard pricing instead of emergency rates
FREQUENTLY ASKED
Vibration Monitoring for Turbines: Key Questions
What type of vibration sensor is best for turbine bearing monitoring?
For most power plant turbine applications, triaxial ICP (Integrated Circuit Piezoelectric) accelerometers mounted directly on bearing housings deliver the best combination of frequency range and signal quality. For high-temperature locations on steam turbines, high-temperature-rated accelerometers rated to 150°C+ are required. Wireless MEMS sensors offer easier installation but have narrower frequency ranges suitable for lower-speed equipment. Book a sensor selection session where OXmaint's engineers recommend the right sensor type and specifications for each measurement point on your specific turbine model, or sign up to access our sensor configuration library.
How does OXmaint's AI know when a vibration reading indicates a real problem versus normal variation?
OXmaint's AI builds a statistical baseline for each measurement point over an initial learning period of 2 to 4 weeks, capturing normal variation across different load conditions, ambient temperatures, and operating modes. Alerts are generated when vibration parameters deviate from this personalized baseline by statistically significant margins — not from generic threshold tables. This approach eliminates the false positives that plague fixed-threshold monitoring systems and reduces alert fatigue for maintenance teams. Sign up to see how the baseline learning process works for your equipment, or request a demo using data from a turbine similar to yours.
Can OXmaint use data from existing vibration sensors and online monitoring systems already installed at the plant?
Yes. OXmaint integrates with vibration data from existing permanently installed sensors via OPC-UA, Modbus TCP, and direct API connections, as well as route-based data collected with portable analyzers (CSV/Excel import supported). You do not need to replace existing sensors or monitoring hardware to start using OXmaint's AI analytics — most plants begin generating actionable predictions from their existing sensor infrastructure within days of connection. Schedule an integration review to confirm compatibility with your current monitoring hardware, or sign up and upload a sample data file to see the analytics output immediately.
How many vibration sensors does a typical 200MW steam turbine unit require for comprehensive bearing monitoring?
A 200MW steam turbine train with HP, IP, and LP sections plus generator typically requires 16 to 24 measurement points for comprehensive bearing health coverage — generally two radial directions (horizontal and vertical) plus one axial measurement per bearing. The exact count depends on turbine configuration, shaft arrangement, and whether auxiliary equipment like boiler feed pumps and fans are included in scope. Book a sensor scoping session where OXmaint's team develops a complete measurement point plan for your turbine train, or access our turbine monitoring templates as a starting framework.
What is the difference between online continuous vibration monitoring and periodic route-based measurement?
Online continuous monitoring captures vibration data 24/7 from permanently mounted sensors, detecting changes that occur between inspection rounds and catching rapid degradation events in real time. Route-based measurement uses a portable analyzer on a monthly or quarterly schedule — useful for trending but unable to catch failures that develop between rounds, which accounts for the majority of unplanned bearing failures. OXmaint supports both approaches in the same platform, allowing plants to start with route-based data and expand to continuous monitoring on their highest-risk assets. Sign up to configure your monitoring strategy or discuss the right mix of monitoring approaches for your plant's risk profile and budget.
YOUR TURBINES ARE ALREADY TALKING — START LISTENING
Every Unplanned Turbine Trip Is a Bearing Failure Your Vibration Sensors Could Have Predicted. OXmaint Makes Sure You Never Miss One.
Connect your sensor data, let OXmaint's AI learn your equipment's normal signature, and start receiving bearing fault warnings weeks before they escalate. Setup takes less than 48 hours. The first prevented failure pays for years of monitoring.
