Research on Classification and Identification of Crack Faults in Steam Turbine Blades Based on Supervised Contrastive Learning DOI Creative Commons
Qinglei Zhang, Longfei Tang,

Jiyun Qin

et al.

Entropy, Journal Year: 2024, Volume and Issue: 26(11), P. 956 - 956

Published: Nov. 6, 2024

Steam turbine blades may crack, break, or suffer other failures due to high temperatures, pressures, and high-speed rotation, which seriously threatens the safety reliability of equipment. The signal characteristics different fault types are slightly different, making it difficult accurately classify faults rotating directly through vibration signals. This method combines a one-dimensional convolutional neural network (1DCNN) channel attention mechanism (CAM). 1DCNN can effectively extract local features time series data, while CAM assigns weights each highlight key features. To further enhance efficacy feature extraction classification accuracy, projection head is introduced in this paper systematically map all sample into normalized space, thereby improving model's capacity distinguish between distinct types. Finally, optimization supervised contrastive learning (SCL) strategy, model better capture subtle differences Experimental results show that proposed has an accuracy 99.61%, 97.48%, 96.22% task multiple crack at three speeds, significantly than Multilayer Perceptron (MLP), Residual Network (ResNet), Momentum Contrast (MoCo), Transformer methods.

Language: Английский

Efficient large-scale graph learning for predicting the 3D multi-physics flow fields of axial compressor Rotor37 with variable geometry DOI
Yichen Hao, Q Liu, Jia Li

et al.

Energy, Journal Year: 2025, Volume and Issue: unknown, P. 136083 - 136083

Published: April 1, 2025

Language: Английский

Citations

0
Review of deep learning-based aerodynamic shape surrogate models and optimization for airfoils and blade profiles DOI
Xiaogang Liu,

S-C Yang,

Haifeng Sun

et al.

Physics of Fluids, Journal Year: 2025, Volume and Issue: 37(4)

Published: April 1, 2025

In recent years, deep learning technology has developed rapidly and shown great potential in the optimization of complex systems. aerodynamic shape optimization, traditional computational fluid dynamics experimental methods are limited due to issues efficiency cost. contrast, surrogate models have gradually become a new alternative their advantages nonlinear modeling, efficient computation, flexible design. These offer novel approaches through such as data regression, automatic differentiation, operator learning. This paper presents comprehensive review latest research progress field based on models, focusing key technologies, application cases, future development trends. The article first elaborates importance context airfoil blade profile introducing background motivation. Then, it discusses technologies challenges faced optimization. Subsequently, introduces detail model, including data- physics-drisven neural networks, Physics-Informed Neural Networks Deep Operator Networks, practical cases these networks Finally, looks into pointing out Kolmogorov–Arnold improving model accuracy interpretability, well types summarizes development.

Language: Английский

Citations

0
Approximate analytical model of mistuned hysteresis effect for multi-joint surfaces DOI
Xuanen Kan, Junhui Chen, Quandai Wang

et al.

Thin-Walled Structures, Journal Year: 2025, Volume and Issue: unknown, P. 113535 - 113535

Published: June 1, 2025

Language: Английский

Citations

0
Aerodynamic force prediction of compressor blade surfaces based on machine learning DOI
Yan Niu,

Kainuo Zhao,

Minghui Yao

et al.

Physics of Fluids, Journal Year: 2024, Volume and Issue: 36(8)

Published: Aug. 1, 2024

The flow field distribution of compressor blades is critical to the performance aero-engine. To efficiently obtain aerodynamic loads on blades, this study employs machine learning models predict characteristics blade surfaces. predictive performances these are evaluated by applying random forest, multi-layer perceptron (MLP), one-dimensional convolutional neural network, and long short-term memory network based simulation data computational fluid dynamics (CFD). results indicate that MLP model performs exceptionally well among all test metrics, with its predictions closely matching CFD results. Further analysis using SHapley Additive exPlanations methods performed interpret reveal importance various input features. research demonstrates significant potential in predicting aerodynamics providing accurate reliable

Language: Английский

Citations

2
Comparison of vibration values of rotating discs with variable parameters obtained by finite element analysis modeling with different machine learning algorithms DOI

Hasan Çallıoğlu,

Said Müftü,

Candaş Nuri Koplay

et al.

Multidiscipline Modeling in Materials and Structures, Journal Year: 2024, Volume and Issue: unknown

Published: Nov. 8, 2024

Purpose Rotating functionally graded (FG) disks of variable thickness generates vibration. This study aims to analyze the vibration generated by rotating using a finite element program and compare results obtained with regression methods. Design/methodology/approach Transverse values FG were modeled different The accuracies models are compared. In context comparing methods, multiple linear (MLR), extreme learning machine (ELM), artificial neural networks (ANNs) radial basis function (RBF) used in this study. error graph between observed value predicted each method was obtained. methods scientific measures calculated. Findings analysis transverse is consistent studies literature. When variables determined on disk ELM, MLR, ANN RBF it concluded that most accurate model order RBF, ANN, MLR ELM. Originality/value There discs literature, but there very few modeling. shows which can be modeling discs.

Language: Английский

Citations

2
Integration of deep learning and computational fluid dynamics for rapid aerodynamic force prediction of compressor blades DOI
Yan Niu,

Kainuo Zhao,

Yuejuan Yang

et al.

Physics of Fluids, Journal Year: 2024, Volume and Issue: 36(10)

Published: Oct. 1, 2024

The distribution of flow fields around compressor blades is crucial for the performance and reliability aircraft engines. To effectively obtain aerodynamic loads, this study combines deep learning with computational fluid dynamics (CFD) to develop an efficient prediction model. Initially, CFD used acquire detailed field data blade surface its surrounding environment. Subsequently, a distance parameterization method applied process geometry, models are capture complex relationship between geometry parameters high precision. results indicate that proposed model can predict loads within seconds mean squared error less than 2%. Compared traditional methods other approaches, exhibits higher accuracy. findings highlight effectiveness integrating enhance predictions provide promising approach future modeling research.

Language: Английский

Citations

1
Research on Classification and Identification of Crack Faults in Steam Turbine Blades Based on Supervised Contrastive Learning DOI Creative Commons
Qinglei Zhang, Longfei Tang,

Jiyun Qin

et al.

Entropy, Journal Year: 2024, Volume and Issue: 26(11), P. 956 - 956

Published: Nov. 6, 2024

Steam turbine blades may crack, break, or suffer other failures due to high temperatures, pressures, and high-speed rotation, which seriously threatens the safety reliability of equipment. The signal characteristics different fault types are slightly different, making it difficult accurately classify faults rotating directly through vibration signals. This method combines a one-dimensional convolutional neural network (1DCNN) channel attention mechanism (CAM). 1DCNN can effectively extract local features time series data, while CAM assigns weights each highlight key features. To further enhance efficacy feature extraction classification accuracy, projection head is introduced in this paper systematically map all sample into normalized space, thereby improving model's capacity distinguish between distinct types. Finally, optimization supervised contrastive learning (SCL) strategy, model better capture subtle differences Experimental results show that proposed has an accuracy 99.61%, 97.48%, 96.22% task multiple crack at three speeds, significantly than Multilayer Perceptron (MLP), Residual Network (ResNet), Momentum Contrast (MoCo), Transformer methods.

Language: Английский

Citations

0