MS008 - Advances in machine learning and data-driven techniques for aerodynamic optimization and uncertainty quantification
Keywords: machine learning; aerodynamics, uncertainty quantification
Over the past few years, the exponential growth in data generated by computational sciences has highlighted the potential to harness this data for extracting valuable insights and improving predictive capabilities. In the field of aerodynamics, extensive parametric studies, trade-off analyses, and optimizations generate vast amounts of valuable information. This data reservoir presents a significant opportunity to advance the use of data-driven and data-fusion models within engineering practice [1]. However, the development and integration of these models remain in their infancy, with best practices still being established.
While machine learning techniques, including neural networks, are widely recognized and offer a versatile toolkit for various tasks—such as cluster analysis, dimensionality reduction, classification, and regression—the challenges associated with processing and preparing aerodynamic and geometric data are substantial. The data handling processes are often complex and highly dependent on the specific objectives of a given task, leading to diverse interpretations and implementations of data-driven approaches. Moreover, the integration of machine learning techniques commonly employed in Artificial Intelligence (AI) and Data Mining (DM) has the potential to significantly reduce the computational costs associated with aerodynamic analysis and uncertainty quantification [2]. By leveraging these advanced methods, there is a promising pathway towards more efficient and accurate solutions in aerodynamic design and analysis, despite the current challenges in data preparation and model maturity.
This minisymposium seeks to gather and disseminate innovative approaches and recent advancements in the application of machine learning and data-driven techniques for aerodynamic analysis and uncertainty quantification, with a particular emphasis on addressing real-world challenges.
REFERENCES
[1] Le Clainche, Soledad, et al. "Improving aircraft performance using machine learning: A review." Aerospace Science and Technology 138 (2023): 108354.
[2] Li, Jichao, Xiaosong Du, and Joaquim RRA Martins. "Machine learning in aerodynamic shape optimization." Progress in Aerospace Sciences 134 (2022): 100849.
While machine learning techniques, including neural networks, are widely recognized and offer a versatile toolkit for various tasks—such as cluster analysis, dimensionality reduction, classification, and regression—the challenges associated with processing and preparing aerodynamic and geometric data are substantial. The data handling processes are often complex and highly dependent on the specific objectives of a given task, leading to diverse interpretations and implementations of data-driven approaches. Moreover, the integration of machine learning techniques commonly employed in Artificial Intelligence (AI) and Data Mining (DM) has the potential to significantly reduce the computational costs associated with aerodynamic analysis and uncertainty quantification [2]. By leveraging these advanced methods, there is a promising pathway towards more efficient and accurate solutions in aerodynamic design and analysis, despite the current challenges in data preparation and model maturity.
This minisymposium seeks to gather and disseminate innovative approaches and recent advancements in the application of machine learning and data-driven techniques for aerodynamic analysis and uncertainty quantification, with a particular emphasis on addressing real-world challenges.
REFERENCES
[1] Le Clainche, Soledad, et al. "Improving aircraft performance using machine learning: A review." Aerospace Science and Technology 138 (2023): 108354.
[2] Li, Jichao, Xiaosong Du, and Joaquim RRA Martins. "Machine learning in aerodynamic shape optimization." Progress in Aerospace Sciences 134 (2022): 100849.
