Gas turbine fault diagnostics using fuzzy systems

Rahimi, Farbod; Lemma, Tamiru Alemu

doi:10.1109/iccsce.2017.8284426

Cited by 5 publications

(4 citation statements)

References 39 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Although the per-unit value is not strictly in the range [0, 1] owing to the influence of noise and fault pulses, this data processing method can preserve more valuable fault features. A bior 2D-wavelet transform was used to obtain highdimensional fault features with data dimensions of (16,8,4) in subsequent work, as shown in figure 5. The generator input of the DCGAN is one-dimensional random noise and the output is a multi-dimensional generative sample.…”

Section: Data Preprocessingmentioning

confidence: 99%

“…The smaller the FID, the better the quality and variety of the generated sample. The Inception v3 network of TensorFlow dspecific requirements for the input dimension, such as (299, 299, 3) or (150, 150, 3), rather than the obtained sample dimension (16,8,4). To solve this problem, bilinear interpolation and channel reduction were used to transform the data dimension.…”

Section: Evaluation Index Of Dcganmentioning

confidence: 99%

“…To verify the effect of the performance improvement of the data preprocessing method, that is, 2D-wavelet transform, we also constructed a training set using the initial data without the 2D-wavelet transform. The only difference between the diagnosis models based on the two training sets was the input dimension, that is, the initial data were (16,9) or (234,), whereas the transformed data were (16,8,4) or (512,). The accuracies of the validation set of these four models trained on the initial dataset were 43.29%, 82.15%, 33.01%, and 33.39%, respectively.…”

Section: Performance Improvement Effect Of Data Preprocessingmentioning

confidence: 99%

“…Fault diagnostic methods for gas turbines, such as gas path faults (GPFs), sensor faults, and life evaluation, have been widely studied [3]. A large number of algorithms, such as fuzzy set theory [4], support vector machines (SVMs) [5,6], and Kalman filtering [7], have been implemented in research on GPF diagnoses [8]. Kyriazis and Mathioudakis [9] proposed a GPF diagnostic method using fuzzy-based decision fusion.…”

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Improved and accurate fault diagnostic model for gas turbine based on 2D-wavelet transform and generative adversarial network

Yao

Wang

Song

et al. 2023

Meas. Sci. Technol.

View full text Add to dashboard Cite

severe working environment causes gas turbines to break down, which can directly affect their performance. Research on the diagnostic methods for gas turbine faults, such as, gas path faults and sensor failures, has always raised concerns. However, traditional fault diagnosis algorithms mostly use instantaneous data rather than time-series data, because they cannot efficiently use time-series analysis to extract fault features and improve algorithm accuracy. Problems with sparse fault samples and categories are also encountered with these algorithms. In this study, a gas turbine fault diagnostic method based on a 2D-wavelet transform and generative adversarial network (GAN) was proposed. The data preprocessing method, 2D-wavelet transform, of multiple time series (MTS) images was used to obtain fault features. Based on the Fréchet inception distance (FID), a performance evaluation index, an optimal generator built from a deep convolutional generative adversarial network model was selected to solve sparse or imbalanced datasets. The classification accuracy of the four algorithms, namely, random forest, support vector machine, convolutional neural network, and deep neural network, verified the performance of the data preprocessing and dataset building methods mentioned earlier. Compared with the original data, the 2D wavelet transform effectively improved the model accuracy. The generated samples also improved the misclassification issue caused by the imbalanced dataset; however, the ratio of real and generated samples in datasets still requires more attention.

show abstract

Section: Data Preprocessingmentioning

confidence: 99%

Section: Evaluation Index Of Dcganmentioning

confidence: 99%