ArXiv Proposes Superposition-based Deep Neural Operator (SDNO) for Enhanced Turbine Endwall Cooling Prediction
Summary:ArXiv introduces a novel physics-enhanced neural operator framework called Superposition-based Deep Neural Operator (SDNO) to improve the prediction of turbine endwall cooling layouts with a variable number of film holes. The method divides the prediction task into two stages: first, a Transformer-based neural operator network, Calculate Net, predicts the temperature field for each sub-part of the endwall; then, another neural operator network, Super Net, combines these predictions to obtain the
Background and Motivation
In modern gas turbine design, the cooling layout of the turbine endwall is crucial for the performance and longevity of the equipment. However, traditional methods struggle with the generalization problem when predicting cooling effectiveness due to variations in the number and location of cooling holes. To address this, researchers propose a novel framework called Superposition-based Deep Neural Operator (SDNO).
Methodology and Innovations
-
Two-stage Prediction Model:
- First Stage: The cooling layout of the turbine endwall is divided into several sub-parts, and a Transformer-based neural operator network, Calculate Net, predicts the temperature field for each sub-part.
- Second Stage: Another neural operator network, Super Net, combines the temperature fields from the first stage to obtain the full cooling layout's temperature field.
-
Signed Distance Function (SDF) Encoding:
- Instead of directly using pixel plots to represent the location of cooling holes, the study introduces SDF to encode the location information, enhancing the model's robustness and generalization capabilities.
-
Training Data Design:
- The model is trained on samples with 1-5 cooling holes but can accurately predict the cooling effectiveness of layouts with 10-20 cooling holes, demonstrating its strong generalization ability for unseen data.
-
Performance Improvement:
- Compared to fully supervised baselines, SDNO significantly improves prediction accuracy, proving its potential for practical applications.
Technical Highlights
- Innovative Architecture Based on Superposition Principle: By decomposing complex tasks into multiple sub-tasks and integrating the results, SDNO demonstrates strong generalization capabilities.
- Application of Signed Distance Function: The use of SDF allows the model to handle variations in cooling hole locations more effectively.
- Introduction of Transformer Architecture: Leveraging the advantages of Transformers, the model performs excellently in processing complex sequential data.
Industry Impact and Developer Recommendations
- Impact on Gas Turbine Design: The application of SDNO is expected to improve the prediction accuracy of turbine endwall cooling layouts, thereby enhancing equipment performance and longevity.
- Research Implications for AI Model Generalization: This study provides new insights into how AI models can improve their generalization capabilities when dealing with complex physical problems.
- Developer Recommendations: Developers in related fields are encouraged to explore neural network architectures based on the superposition principle and apply similar methods in practice to enhance model performance.
Conclusion
The proposed SDNO framework offers an efficient and accurate method for predicting turbine endwall cooling layouts, showcasing the potential of AI in solving complex physical problems.
— END —Source: ArXiv Machine Learning (cs.LG) (2026-09-30)
Tags: #Deep Learning #Neural Networks #Turbine Cooling #Generalization #Transformer
Editorial & Fact-Checking Note: This article is compiled from primary research, official release documentation, and source papers by the ZICQ Newsroom pipeline with automated entity verification and human editorial review. If you notice any technical inaccuracy, please submit a correction via our corrections policy or email our editorial desk directly.
Community Comments