Source-linked AI summary
Experimental and Signal Processing Techniques for Fault Diagnosis on a Small Horizontal-Axis Wind Turbine Generator
Francesco Natili, Francesco Castellani, Davide Astolfi, Matteo Becchetti
TL;DR
Small HAWT generator condition monitoring is overlooked despite demanding operating conditions and strong electromechanical coupling. This study analyzes experimental vibration data from a 3 kW HAWT and its permanent-magnet generator using wind-tunnel and test-rig measurements. Spectral coherence identifies damage compatible with a bearing-cage fault, while test-rig data better support precise spectrum localization.
Problem
Condition monitoring of small HAWT generators is overlooked despite complex flow conditions, high rotational speeds, and the generator’s important electromechanical coupling.
Method
The study analyzes time- and frequency-domain vibration data from a complete HAWT in wind-tunnel tests and generators driven at different speeds on a test rig.
Results
Spectral coherence identifies frequency content compatible with damage to the generator bearing cage in both test-rig and wind-tunnel data.
Takeaways & Limitations
Small HAWT condition monitoring requires sophisticated signal processing and experimental facilities, with test-rig data more adequate for investigating definite vibration-spectrum portions.
Takeaways & Limitations
Wind-tunnel measurements involve mixed whole-device spectral contributions, complicating precise damage localization; the study also does not estimate remaining useful life.
Abstract
from arXiv · showhide
Small HAWT is a technology characterized by non-trivial critical points, basically because it is targeted for domestic use and therefore cheap manufacturing and control must conjugate with good efficiency under possibly complex flow conditions (especially in urban environment). Therefore, dynamical control optimization and noise and vibration mitigation are pressing issues for this kind of technology. Despite it is peculiar of small HAWTs that the generator constitutes a non-negligible fraction of the total mass and therefore the electromechanical coupling is relevant, condition monitoring of small HAWT generators is an overlooked topic. The present work is a test case study of damage diagnosis on a permanent magnet generator of a HAWT having 3 kW of maximum power and 2 meters of rotor diameter. The experimental analysis is conducted through wind tunnel tests and on a generator test rig where a damaged and an undamaged generators have been driven at different rotational speeds. Vibration measurements are collected in the wind tunnel through radial accelerometers near the rear bearing of the shaft and in the test rig through uni-axial accelerometers (fixed in radial positions, in order to be aligned with front and rear bearings). The test rig data result being particularly useful for studying the low-frequency tail of the vibration spectrum, where the characteristic frequencies of the bearing are located. The experimental data are analyzed in the time and frequency domain for feature extraction: a fault in the cage of the bearing supporting the generator is diagnosed using in particular the spectral coherence analysis.
1. Introduction
Small HAWTs face complex operating conditions, high rotational speeds, and strong generator–structure coupling, yet condition monitoring of their generators remains underexplored. This study addresses the gap experimentally using wind-tunnel and test-rig measurements to diagnose generator damage.
- Small HAWTs must combine inexpensive domestic manufacturing and control with efficiency under complex, especially urban, flow conditions.
- Rotational speeds reaching several hundreds rpm make noise and vibration control a pressing prototyping concern, while these turbines typically lack condition-monitoring systems.
- The generator’s substantial share of total mass strengthens electromechanical coupling and can transmit powerful high-frequency vibrations to the HAWT structure.
- Existing bearing-damage studies mainly concern MW-scale turbines, while small-HAWT generator condition monitoring remains remarkably sparse.
- The study compares wind-tunnel measurements of a complete 3 kW, 2-meter-rotor HAWT with test-rig measurements of its generator to interpret whole-device vibration data.
- Spectral coherence and test-rig measurements enabled identification of damage at the generator bearing cage, whose characteristic frequencies lie in the low-frequency spectrum.
2. The test case and the facilities
The test case is a three-bladed, 3 kW small HAWT evaluated in a wind tunnel and with two nominally identical permanent-magnet generators on a test rig. Vibration, rotational, mechanical, and electrical measurements characterize the turbine and generator tests.
- The test HAWT has three blades, a 2-meter rotor diameter, 3 kW maximum power, and a 40 kg nacelle.
- The University of Perugia wind tunnel has a 2.2 × 2.2 m open chamber, maximum air speed of 45 m/s, and turbulence below 0.4%.
- Wind-tunnel vibration tests used a radial accelerometer near the rear shaft bearing, 5 kHz sampling, and tachometer-based rotor-speed measurement.
- Test-rig measurements used radial uni-axial accelerometers aligned with the generator’s front and rear bearings, plus torque, optical-rpm, voltage, and current measurements.
- Two identical 1.8 kW permanent-magnet generators were tested across wind-tunnel and several-shaft-speed test-rig conditions; one had anomalous vibrations and non-optimal performance.
3. The fault diagnosis
The diagnosis combines mechanical and electrical measurements with vibration analysis in the frequency and time domains. Spectral correlation identifies the bearing-cage fault more precisely than order-tracked Mahalanobis analysis, while test-rig measurements reveal the generator’s efficiency loss.
- Performance assessment: Torque, rotational speed, voltage, and current measurements support mechanical-power and efficiency estimates for damaged and undamaged generators.The test rig applies a resistive generator load and uses torque-meter, tachometer, and electrical measurements.
- Vibration spectra: Waterfall plots represent vibration amplitude across frequency and rotational-speed regimes, allowing features tied to shaft order and harmonics to be examined.The shaft frequency is labeled 1P, with harmonics labeled 2P, 3P, and higher orders.
- Bearing-fault framework: Bearing-fault diagnosis uses characteristic frequencies associated with inner-race, outer-race, rolling-element, and cage damage.The bearing framework uses shaft frequency, rolling-element diameter, pitch diameter, element count, and contact angle to define these characteristic orders.
- Spectral correlation: Spectral correlation is appropriate for rotating, non-stationary signals because periodic modulation produces discrete cyclic frequencies related to the fundamental 1P order.A bearing fault should generate non-zero spectral correlation at characteristic orders and their harmonics.
- Spectral-correlation result: A 0.4P contribution and its harmonics, especially 0.8P, appear for the damaged generator and are compatible with a bearing-cage defect.The same indication is observed in wind-tunnel and test-rig measurements at 400 rpm.
- Time-domain feature extraction: Raw time-domain Mahalanobis distances do not clearly separate generators, whereas order-tracking around bearing frequencies distinguishes damaged from undamaged data but localizes the fault less precisely.The cage- and inner-race-focused analyses are nearly indistinguishable, making them weaker than cyclostationarity for identifying the bearing location.
4. Discussion
The study compares wind-tunnel and test-rig measurements for diagnosing damage in a small HAWT generator. Spectral coherence can identify damage location, while test-rig data are clearer for interpreting the relevant vibration spectrum.
- Wind-tunnel tests capture the whole operating HAWT, whereas test-rig measurements isolate the generator subcomponent.
- Wind-tunnel spectra mix blade motion, rotor imbalance, electromechanical coupling, and bearing-damage contributions, complicating precise localization.
- Test-rig data are more suitable for examining definite vibration-spectrum portions because the subcomponent is isolated.
- Time-domain feature extraction can be prohibitive, whereas cyclostationary spectral correlation provides damage-location indications in both measurement settings.
- The damaged and undamaged generators produce distinguishable low-frequency wind-tunnel vibration scenarios, but localization requires suitable post-processing.
5. Conclusions
This case study evaluates condition monitoring for a small HAWT permanent-magnet generator using accelerometer measurements from wind-tunnel and test-rig experiments. Cyclostationary analysis reasonably localized a bearing-cage fault, while test-rig data supported clearer diagnosis; remaining useful life was not estimated.
- Condition monitoring of small HAWT generators remains overlooked despite their high rotational speed and substantial electromechanical coupling.
- The experiments compared suspected-damaged and undamaged 1.8 kW PMGs installed on a 3 kW, 2-meter-rotor HAWT.
- Test-rig spectra were easier to interpret than whole-device wind-tunnel spectra and were particularly useful for diagnosing bearing damage.
- Spectral correlation identified frequency content compatible with bearing-cage damage in both test-rig and wind-tunnel data.
- Small HAWT condition monitoring requires sophisticated signal-processing techniques and experimental facilities, with wind-tunnel tests providing system-level dynamic information.
- The study detected and reasonably localized bearing damage but did not estimate the subcomponent’s remaining useful life.
Abbreviations
The manuscript defines abbreviations used for condition monitoring, turbine type, generator type, and rotational speed.
- CM denotes Condition Monitoring.
- HAWT denotes Horizontal-Axis Wind Turbine.
- PMG denotes Permanent Magnet Generator.
- rpm denotes Revolutions Per Minute.