Main bearing temperature rise
Slow upward drift in main-bearing temperature is the earliest reliable signal of a developing main-bearing problem — usually visible months to a year before action becomes urgent. Replacement is one of the most expensive operations on a modern turbine, which makes early detection unusually high-leverage.
Reviewed by Michael Tegtmeier, Founder & Managing Director · Last reviewed: May 10, 2026
What happens
The main bearing carries the rotor's weight and the rotor-shaft thrust. Lubrication conditions, alignment, and contamination drive its operating temperature. As damage develops — typically as micropitting, fretting corrosion, or false brinelling progresses on the rolling elements or raceway — friction and the resulting heat both increase. The temperature differential to the model-predicted normal value grows long before any acute event, which is why SCADA temperature is such a reliable lead indicator.
Signs to look for
Bearing temperature trends upward relative to ambient and power output, while a healthy reference fleet of same-model turbines stays flat.[1,2]
Temperature standard-deviation differential widens — the bearing runs more variably for a given operating point.[1]
Grease analysis (when sampled) shows elevated iron particles or signs of additive depletion.[3]
Acoustic or low-frequency vibration peaks (when CMS data is available) — usually a later signal than SCADA temperature.[2]
Root causes
- Insufficient lubrication — clogged or undersized grease lines, blocked grease drains, or simply lubrication intervals too long for the duty cycle.[3,4]
- Contamination — water ingress through seals, abrasive particles, or grease aging.[3]
- Manufacturing or installation defects — misalignment, raceway brinelling from transport / lifting, residual stresses.[4]
- Operating-envelope drift — extended high-load or high-yaw operation outside the bearing's design assumptions.[2,4]
How Turbit detects this
Per-turbine neural networks learn each main bearing's individual normal-temperature signature given ambient + load + curtailment state. Deviations are flagged on the differential to the predicted value, not against a fixed threshold or fleet average. A second AI layer classifies the anomaly as drift (slow trend) vs. step (acute event) and predicts the most likely root cause from a library of 35+ failure modes.
From the Turbit fleet
In Turbit's monitored fleet, main-bearing temperature deviations are among the longest lead-time detections in our library — typical first-sighting-to-required-action windows run from 3 months to over a year, depending on damage propagation rate. Trend acceleration (the rate at which the differential widens) is a sharper severity signal than the absolute value.
On a 50-turbine portfolio, even one avoided main-bearing exchange — typically EUR 250k–450k all-in for a modern 3 MW+ machine before downtime cost — pays for years of monitoring across the entire fleet.
References
- Wind Turbine Main Bearing Fault Prognosis Based Solely on SCADA Data — Encalada-Dávila et al. (2021)DOI: 10.3390/s21062228
- SCADA Data-Driven Wind Turbine Main Bearing Fault Prognosis Based on Principal Component Analysis — Encalada-Dávila et al. (2022)
- Bearing temperature monitoring of a wind turbine using a physics-based model — Astolfi, Castellani, Terzi (2017)
- A Review of Research on Wind Turbine Bearings' Failure Analysis and Fault Diagnosis — Liu et al. (2023)
See Turbit on your fleet
Backtest Turbit on the turbines where you already know something happened — or read how operators like VSB, Energiequelle, Enercity and Teut use Turbit on their fleets today.