Aerodynamic design and optimization of blade end wall profile of turbomachinery based on series convolutional neural network

Du, Qiuwan; Yang, Like; Li, Liangliang; Liu, Tianyuan; Zhang, Di; Xie, Yonghui

doi:10.1016/j.energy.2021.122617

Cited by 22 publications

(1 citation statement)

References 47 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Li [25] built a blade excitation analysis model based on machine learning, which could achieve the fast and accurate prediction of unsteady aerodynamic forces. Du [26] adopted the DCNN network which can precisely give rich flow field information and performance characteristics within 3 ms of training to enhance the design optimization problem of the end wall profile of a turbine stator blade. Du [27] also applied the DCNN structure in the field of airfoil design optimization to generate fundamental airfoil profiles with only three basic parameters, with high prediction accuracy when it was compared to the traditional machine learning methods.…”

Section: Introductionmentioning

confidence: 99%

Prediction Method of Unsteady Flow Load of Compressor Stator under Working Condition Disturbance

Liu

et al. 2022

Applied Sciences

Self Cite

View full text Add to dashboard Cite

Due to the complexity of the compressor operating conditions and the existence of various disturbances and unsteady effects in the flow field, the analysis of compressor stator vibration characteristics becomes particularly critical. The convolutional neural network model combined with a transient CFD method was introduced to solve the difficulty of analyzing the flow load of the compressor stator blade. This paper mainly focuses on two key points: the complex change of the aerodynamic load and the accurate prediction of the blade excitation. Considering the stator–rotor interference, the unsteady effects, and the variable working condition characteristics, the random disturbance analysis model of the flow field boundary was generated to simulate the unsteady flow excitation of the stator under complex working conditions. By establishing the neural network of boundary disturbance and flow excitation characteristics, the prediction model was trained and generated under the support of large-scale data. The most important role of the model was to establish the end-to-end data mapping between the disturbance condition and the aerodynamic load of the stator blade. The conclusions demonstrate that the introduction of an airflow disturbance is helpful to obtain the excitation characteristics of the stator under complex working conditions. The model established in this paper based on 1000 groups of disturbed working condition data can effectively predict the aerodynamic load of the blades under complex working conditions. In addition, the construction of the model is beneficial for saving a lot of computing resources, and the prediction accuracy also reaches a good level. The method presented in this paper provides a reference for the vibration analysis of the compressor stator.

show abstract

Section: Introductionmentioning

confidence: 99%