scholarly journals Machine-Learning-Based Intelligent Mechanical Fault Detection and Diagnosis of Wind Turbines

2021 ◽  
Vol 2021 ◽  
pp. 1-11
Author(s):  
Qiang Gao ◽  
Xinhong Wu ◽  
Junhui Guo ◽  
Hongqing Zhou ◽  
Wei Ruan

Wind power has gained wide popularity due to the increasingly serious energy and environmental crisis. However, the severe operational conditions often bring faults and failures in the wind turbines, which may significantly degrade the security and reliability of large-scale wind farms. In practice, accurate and efficient fault detection and diagnosis are crucial for safe and reliable system operation. This work develops an effective deep learning solution using a convolutional neural network to address the said problem. In addition, the linear discriminant criterion-based metric learning technique is adopted in the model training process of the proposed solution to improve the algorithmic robustness under noisy conditions. The proposed solution can efficiently extract the features of the mechanical faults. The proposed algorithmic solution is implemented and assessed through a range of experiments for different scenarios of faults. The numerical results demonstrated that the proposed solution can well detect and diagnose the multiple coexisting faults of the operating wind turbine gearbox.

2014 ◽  
Vol 8 (4) ◽  
pp. 380-389 ◽  
Author(s):  
Donatella Zappalá ◽  
Peter J. Tavner ◽  
Christopher J. Crabtree ◽  
Shuangwen Sheng

2014 ◽  
Vol 16 (6) ◽  
pp. 1029-1037 ◽  

<div> <p>The structure of the wind turbines nowadays is a critical element due to their importance from the reliability, availability, safety, and cost points of view. This is more relevant when the offshore wind turbine is considered. This paper introduces a novel design of a Fault Detection and Diagnosis (FDD) model based on ultrasound technique. The FDD model will be able to detect fault/failures via the pulse-echo technique. The pulse-echo is got via piezoelectric transducers that are also employed as sensors. The signal processing is based on two steps. Firstly, a wavelet transform is applied to the measured signals with filtering purposes, in order to enhance the signal to noise ratio. Secondly, a time series modeling approach, as an autoregressive with exogenous input model, is employed for pattern recognition by minimizing the Akaike information criterion. An experimental platform is proposed to test the procedure, where pulse-echo experiments were employed before and after a fault occurred. The results from this paper lead to the identification of an early indication of structural problems induced by internal (material, shape, age, etc.) and external (temperature, humidity, pressure, etc.) factors. The model can anticipate catastrophic faults, reducing the preventive/corrective tasks and costs, etc, and increasing the availability of the wind turbine, and therefore the energy production.</p> </div> <p>&nbsp;</p>


IEEE Access ◽  
2020 ◽  
Vol 8 ◽  
pp. 78343-78353 ◽  
Author(s):  
Radhia Fezai ◽  
Kamaleldin Abodayeh ◽  
Majdi Mansouri ◽  
Abdelmalek Kouadri ◽  
Mohamed-Faouzi Harkat ◽  
...  

Author(s):  
T. Agami Reddy

Research has been ongoing during the last several years on developing robust automated fault detecting and diagnosing (FDD) methods applicable for process faults in chillers used in commercial buildings. These FDD methods involve using sensor data from available thermal, pressure and electrical measurements from commercial chillers to compute characteristic features (CF) which allow more robust and sensitive fault detection than using the basic sensor data itself. One of the proposed methods is based on the analytical redundancy approach using polynomial black-box multiple linear regression models for each CF that are identified from fault-free data in conjunction with a diagnosis table. The second method is based on a classification approach involving linear discriminant analysis to identify the classification models whereby both the detection and diagnosis can be done simultaneously. This paper describes the mathematical basis of both methods, illustrates how they are to be tuned using the same fault-free data set in conjunction with limited faulty data, and then compares their performance when applied to different fault severity levels. The relative advantages and disadvantages of each method are highlighted and future development needs are pointed out.


Sign in / Sign up

Export Citation Format

Share Document