Vibration Fault Diagnosis in Wind Turbines Based on Automated Feature Learning

Abstract

A growing number of wind turbines are equipped with vibration measurement systems to enable the close monitoring and early detection of developing fault conditions. The vibration measurements are analyzed to continuously assess the component health and prevent failures that can result in downtimes. This study focuses on gearbox monitoring but is also applicable to other subsystems. The current state-of-the-art gearbox fault diagnosis algorithms rely on statistical or machine learning methods based on fault signatures that have been defined by human analysts. This has multiple disadvantages. Defining the fault signatures by human analysts is a time-intensive process that requires highly detailed knowledge of gearbox composition. This effort needs to be repeated for every new turbine, so it does not scale well with the increasing number of monitored turbines, especially in fast-growing portfolios. Moreover, fault signatures defined by human analysts can result in biased and imprecise decision boundaries that lead to imprecise and uncertain fault diagnosis decisions. We present a novel accurate fault diagnosis method for vibration-monitored wind turbine components that overcomes these disadvantages. Our approach combines autonomous data-driven learning of fault signatures and health state classification based on convolutional neural networks and isolation forests. We demonstrate its performance with vibration measurements from two wind turbine gearboxes. Unlike the state-of-the-art methods, our approach does not require gearbox-type-specific diagnosis expertise and is not restricted to predefined frequencies or spectral ranges but can monitor the full spectrum at once.

Cloud Based Fault Diagnosis by Convolutional Neural Network as Time–Frequency RGB Image Recognition of Industrial Machine Vibration with Internet of Things Connectivity

“…It can be an additional sensor only for fault diagnosis or a sensor already present in the system used by the control algorithms. The electromechanical machine or power system can be investigated by many different sensors and signals: current [ 71 , 72 ] and voltage [ 73 , 74 ], torque [ 75 , 76 ], angular velocity/position [ 77 , 78 ], linear 3-axis acceleration/speed/position [ 16 , 17 ], Doppler laser vibrometer [ 79 ], transmission coefficient and reflexion coefficient of omnidirectional antenna [ 80 ], strain/tension [ 81 , 82 , 83 , 84 ], power consumption [ 85 , 86 , 87 , 88 ], internal/external temperature at selected points [ 89 , 90 ] or surface temperature by thermal camera [ 91 , 92 ], depending on frequency range: displacement [ 93 ], vibrations [ 15 , 18 , 94 , 95 , 96 ], sound [ 97 , 98 , 99 ], sound from several microphones [ 100 ] or ultrasound [ 101 , 102 ], vibro-acoustic [ 103 ], chemical analysis of lubrication [ 104 , 105 ], chemical analysis by spectral imaging [ 106 , 107 , 108 , 109 ], camera imaging in human colour spectrum [ 110 , 111 , 112 , <...…”

A Fault Diagnosis Technique for Wind Turbine Gearbox: An Approach using Optimized BLSTM Neural Network with Undercomplete Autoencoder

“…Using fault features of convolution channels and frequency bands of wavelet coefficients, the residual network can be used to identify a fault in the gearbox of a wind turbine [13]. The methods of convolutional neural networks and isolation forests were applied to classify the health of the gearbox of a wind turbine in [14]. The neighborhood component analysis technique for best feature collection was used to evaluate the healthiness of wind turbine gearboxes in [15].…”

Research on gearbox compound fault diagnosis method and system development based on entire gearbox health maintenance

“…Step 4: Furthermore, the proposed method gets the real reciprocal spectrum C as shown in formula (5). By using the real reciprocal spectrum C, a whitening signal X whiten is obtained which retains the periodic transients and white noise, and its resonant frequency band is equally important everywhere, there is no need to select the optimal resonant frequency band consequently.…”

“…Firstly, the dynamic fault response of gearbox is derived through the fault mechanism, and then the feature information reflecting the fault state is obtained from the fault vibration signal, so as to identify and confirm the compound fault. 5,6 Generally, wavelet transform, 7 time-frequency representation technology, 8 adaptive decomposition technology 3,9 and its derivative algorithm, 1,6 improved envelope spectrum algorithm, 4 Kurtosis technique, 10 sparse model optimization characterization, 11,12 and other techniques are often applied to extract weak feature information which also provides the basis for subsequent fault mode classification, health monitoring, reliability evaluation, and life prediction. On the other hand, data-driven intelligent diagnosis technology 13 is a new developed subject in recent years.…”