Predictive maintenance for wind turbines combines vibration, temperature, lubricant and load data from components such as the main bearing, gearbox, and generator. The goal is to detect and classify developing faults with appropriate confidence, track severity, and support a planned intervention before functional failure.
Done well, it can convert an emergency drivetrain repair into a planned intervention during a suitable weather and production window. Utility-scale nacelles are elevated, access-constrained environments, so installation, communications, safety, and lifecycle maintenance all affect the monitoring architecture.
Why Wind Turbines Require a Different Predictive Maintenance Approach
Most industrial condition monitoring assumes a technician can walk up to the asset, plug in a data logger, and walk away with a spectrum. A wind turbine nacelle does not offer that convenience.
Retrofitting a wired sensor network can require cable routing and protection, engineering approval, safe-access planning, and a scheduled outage. The cost and practicality vary by turbine design and site; crane mobilization is normally associated with major component work, not automatically with sensor installation. Wired and wireless options should therefore be compared on measurement performance, cybersecurity, reliability, maintainability, and total lifecycle cost.
A battery-powered wireless sensor can reduce new cabling, but mounting and communications must be engineered for the application.
Where Drivetrain Failure Costs Accumulate
Within NREL’s Gearbox Reliability Database of roughly 1,050 confirmed gearbox damage records, bearings account for 76.2% of recorded damage, gears for 17.3%, and other categories for 6.6%. Damage in both the bearing and gear categories is concentrated in the parallel section. These values describe the composition of reported gearbox damage records; they are not fleetwide turbine-failure probabilities.
The U.S. Department of Energy identifies axial cracking, also called white-etch cracking, as a major premature bearing-failure mode in wind-turbine gearboxes. DOE also notes that the definitive field cause is not settled; sliding or skidding, lubricant chemistry, and electrical current are among the investigated contributors.
Historical secondary sources published in 2015 cite about one gearbox failure per 145 turbines per year and repair costs of approximately $200,000-$300,000, with some cases up to $500,000. Actual cost and outage duration depend on turbine rating, onshore or offshore location, crane and vessel availability, logistics, contract terms, weather, and lost production.
Predictive Maintenance Diagnostics: From Raw Vibration to a Fault Class
A raw vibration waveform rarely identifies a fault by itself. Diagnosis combines time-waveform features, spectra, envelope analysis, order tracking, operating state, and machine geometry to test whether the observed components are consistent with a mechanical fault.
For gears, analysts track Gear Mesh Frequency (GMF), harmonics, and sidebands; these patterns can indicate modulation associated with wear, eccentricity, misalignment, load variation, or other mechanisms, so they are not uniquely diagnostic on their own. For rolling-element bearings, envelope analysis demodulates a machine-specific high-frequency resonance and reveals components near BPFI, BPFO, BSF, and FTF in the lower-frequency envelope spectrum. A fixed 2-6 kHz band is not universal. Variable-speed turbines also require reliable speed or order information and comparison within comparable load, speed, and operating states.
Edge processing can reduce transmitted data volume by calculating features, spectra, envelope indicators, and anomaly scores locally.
Standards Relevant to Wind-Turbine Condition Monitoring
Several standards contribute to a wind-turbine condition-monitoring specification, and they address different layers of the program.
| Standard | Scope | Relevance |
| ISO 20816-21:2025 | Vibration measurement and evaluation on horizontal-axis turbines rated above 200 kW, covering both geared and direct-drive designs (ISO) | Defines measurement and evaluation guidance for whole-machine vibration, including annex zone boundaries based on field data. ISO explicitly notes that these zones are generally not suitable for early fault detection and that the standard does not address diagnosis or fault detection. |
| IEC 61400-25-6 | Information models for condition-monitoring data exchange between turbines, CMS platforms, and SCADA/asset-management systems | Defines information models and data exchange for condition-monitoring information and is used with other parts of the IEC 61400-25 series. Conformance can support interoperability, but it does not by itself guarantee plug-and-play integration or eliminate OEM-specific engineering. |
| ISO 16079-2:2020 | Wind-turbine drivetrain monitoring: main bearings, gearboxes, generators, couplings, and lubrication systems | Wind-specific implementation guidance for failure-mode detection, diagnostics, and prognostics |
| ISO 13373-2:2016 | Vibration-data processing, analysis, and presentation for rotating machinery | Time- and frequency-domain methods for monitoring and diagnostics |
Architecture in the Field: DS-Track AIR, DS Hub AIR, and DS-Insight
A proposed wireless deployment could place sensors on selected main-bearing, gearbox, and generator locations, reporting to a nacelle-mounted gateway. Final sensor count, axes, mounting points, measurement settings, and gateway location should follow a failure-mode and effects analysis, structural signal-path assessment, radio survey, turbine-OEM constraints, and representative baseline measurements.
A fleet dashboard can aggregate trends, filter transient events, and support risk-based prioritization.
What Field Diagnostic Data Can Show
The examples below are said to come from Delphisonic’s internal reports on rail and maritime machinery, not wind turbines. Bearing-kinematics and gear-mesh principles are transferable, but signal paths, structures, speeds, loads, controls, aerodynamic excitation, mounting, and operating regimes differ.
In the cited gear case, running-speed harmonics, sidebands, and kurtosis may be consistent with modulation or impulsiveness, but they do not uniquely establish tooth wear.
A separate bearing example reportedly showed BPFI-related harmonics across four severity levels.
Building a Wind Turbine Predictive Maintenance Program
Prioritize monitoring through asset criticality, failure-mode and effects analysis, fleet history, detectability, access constraints, and consequence of failure; do not assume the same two components dominate every fleet. Use vibration and envelope analysis with speed or order references, temperature, lubrication condition, SCADA context, and appropriate electrical or structural indicators. BPFI, BPFO, and GMF analysis covers selected drivetrain mechanisms, not most causes of turbine downtime.
From there, a fleet-wide dashboard with alarm filtering can support maintenance prioritization when alarms are validated, operating states are normalized, escalation rules are defined, and diagnostic findings are linked to the work-management process.
If you’re evaluating condition monitoring for a wind fleet, ask for wind-specific references, current product documentation, validation results, cybersecurity and integration details, and a pilot acceptance plan. Request a demo to review a deployment against representative turbines, operating states, failure modes, and measurable acceptance criteria.
References
– National Renewable Energy Laboratory, “Wind Turbine Gearbox Reliability Database, Operation and Maintenance” (2016) – https://docs.nrel.gov/docs/fy17osti/68347.pdf
– U.S. Department of Energy, “Zeroing In on the No. 1 Cause of Wind Turbine Gearbox Failures” – https://www.energy.gov/eere/wind/articles/zeroing-no-1-cause-wind-turbine-gearbox-failures
– International Organization for Standardization, “ISO 20816-21:2025” (supersedes the withdrawn ISO 10816-21:2015) – https://www.iso.org/standard/84280.html
– Pall Corporation, citing University of Strathclyde (2015) research on offshore wind turbine failure rates – https://www.pall.com/en/decarbonization/blog/wind-turbine-gearbox-failures.html
– GCube Insurance, “Grinding Gearboxes” (2015), as reported by Tamarindo – https://tamarindo.global/insight/analysis/wind-turbine-gearbox-failure/
– International Organization for Standardization, “ISO 16079-2:2020 – Condition monitoring and diagnostics of wind turbines – Part 2: Monitoring the drivetrain” – https://www.iso.org/standard/67618.html
– International Organization for Standardization, “ISO 13373-2:2016 – Vibration condition monitoring – Part 2: Processing, analysis and presentation of vibration data” – https://www.iso.org/standard/68128.html
– International Electrotechnical Commission, “IEC 61400-25-6:2016 – Logical node classes and data classes for condition monitoring” – https://webstore.iec.ch/en/publication/32580
Frequently Asked Questions (FAQ)
Predictive maintenance for wind turbines: sensors, standards, and deployment questions
Predictive maintenance for wind turbines uses vibration, temperature, and AI-based fault classification on components like the main bearing, gearbox, and generator to catch developing faults weeks or months before they cause an unplanned shutdown, instead of servicing a turbine on a fixed calendar interval regardless of its actual condition.
Envelope spectrum analysis on bearing and gear mesh frequencies typically surfaces early-stage faults while the fault energy is still concentrated in the 2 kHz to 6 kHz band, well before the defect produces an audible noise, giving maintenance teams a multi-week planning window in most field cases.
No. Edge AI sensors such as DS-Track AIR classify the fault on the sensor itself and only transmit the fault score and a spectral summary, which matters at wind sites where cellular or satellite connectivity is often the weakest link in the monitoring chain.
A wired sensor such as DS-Track offers a wider temperature range and sub-1ms latency but requires a cable run into the nacelle, which is expensive to retrofit on older turbines. A wireless sensor such as DS-Track AIR runs on battery for 3 to 5 years and mounts without any cabling, which is why it is the more common choice for retrofitting existing wind assets.
ISO 20816-21:2025 governs how broadband vibration severity is measured on wind turbines above 200 kW, covering both geared and direct-drive designs, while IEC 61400-25-6 defines how condition monitoring data is structured so it can be exchanged with SCADA and asset-management systems without a custom integration.
