Yiren Shen

Yiren Shen

he/him/his

Postdoctoral Researcher at University of Washington NSF AI institute in Dynamical Systems

Posts by Collection

portfolio

publications

Comprehensive Study of the Initial Diameter for Combustion of n-Heptane/iso-Octane Mixture Droplets

Published in 10th U.S. National Combustion Meeting, 2017

This conference paper studies the effect of initial droplet diameter on the combustion of n-heptane/iso-octane mixture droplets.

Recommended citation: Xu, Y.; Farouk, T.; Shen, Y.; Hicks, M. C.; Avedisian, C. T.; Xie, Y.; Reeves, A. P.; Dryer, F. L. (2017). "Comprehensive Study of the Initial Diameter for Combustion of n-Heptane/iso-Octane Mixture Droplets." 10th U.S. National Combustion Meeting.

Burning of Algae-Derived Biofuel Droplets and Their Mixtures with Jet Fuel

Published in Energy & Fuels, 2020

This paper reports new results about the combustion dynamics of hydroprocessed renewable jet fuel droplets derived from heterotrophic algae (AHRJ) burning under conditions that promote one-dimensional transport dynamics and near spherically symmetric burning conditions. Results are compared between Jet-A and three AHRJ/Jet-A mixtures containing 25, 50, and 75% AHRJ by volume. The results show that AHRJ droplets burn slightly faster than Jet-A and have a significantly reduced sooting propensity.

Recommended citation: Xu, Y.; Brunson, J. D.; Avedisian, C. T.; Shen, Y.; Keresztes, I.; Condo, A. M.; Phillips, D. (2020). "Burning of Algae-Derived Biofuel Droplets and Their Mixtures with Jet Fuel." Energy & Fuels. 34, 929-935.
Download Paper

Quantitative Soot Volume Fraction Measurements from Droplet Combustion Experiments of Sooty Fuels

Published in 36th Annual Meeting of the American Society for Gravitational and Space Research (ASGSR), 2020

This presentation introduces research on quantitative soot volume fraction measurements from droplet combustion experiments of sooty fuels.

Recommended citation: Xu, Y.; Shen, Y.; Avedisian, C. T.; Hicks, M.; Suh, H. K.; Choi, M. (2020). "Quantitative Soot Volume Fraction Measurements from Droplet Combustion Experiments of Sooty Fuels." 36th Annual Meeting of the American Society for Gravitational and Space Research (ASGSR).
Download Paper

Soft Robot Skin With Conformal Adaptability for On-Body Tactile Perception of Collaborative Robots

Published in IEEE Robotics and Automation Letters, 2022

Robot skin, as the physical interface between collaborative robots (cobots) and the external environments, could mediate on-body tactile signals for cobots to respond timely once a collision occurs during tasks and thus avoid damages to both cobots and surroundings. In this letter, a soft robot skin consisting of an array of force-sensing units and a substrate with a serpentine structure was developed, both of which were made of porous material. The serpentine structure endows the developed skin with stretchable attributes, allowing it to be conformally attached to complex surfaces of a cobot with less local stress.

Recommended citation: Ye, Z.; Pang, G.; Xu, K.; Hou, Z.; Lv, H.; Shen, Y.; Yang, G. (2022). "Soft Robot Skin With Conformal Adaptability for On-Body Tactile Perception of Collaborative Robots." IEEE Robotics and Automation Letters. 7, 5127-5134.
Download Paper

Towards a Framework for Parametric Models and Geometric Sensitivity Computations with SU2 and pyCAPS

Published in AIAA Aviation 2023, 2023

Shape design optimization relies on obtaining objective and constraint functional sensitivities to geometric parameters in order to explore the design space effectively. To facilitate computer-aided design (CAD) applications in industry, this study presents a workflow for effectively handling geometric constraints and computing analytic sensitivities. The workflow is capable of computing parametric sensitivities by combining geometric sensitivities obtained using Engineering Sketch Pad with adjoint surface sensitivities from the SU2 adjoint solver through a dot product within the pyCAPS framework. The effectiveness and limitations of this workflow are evaluated through three test cases, including a preliminary demonstration of a shape optimization problem in wing design. The results indicate the conceptual effectiveness of the proposed workflow, while noting the need for further improvement in the mesh generation routine.

Recommended citation: Shen, Y.; Patel, H.; Galbraith, M.; Alonso, J. (2023). "Towards a Framework for Parametric Models and Geometric Sensitivity Computations with SU2 and pyCAPS." AIAA Aviation 2023. AIAA 2023-3312.
Download Paper

Quantitative Investigation of Sooting Dynamics in Droplet Combustion Using an Automated Image Analysis Algorithm

Published in Fuel, 2023

This paper reports an image analysis approach using a newly-developed open-source program to extract quantitative measurements of soot volume fraction (SVF) from digital video images of burning n-heptane droplets. The automated program developed in this work can analyze images of fixed and untethered droplets to quantify sooting dynamics. The images analyzed in the program were taken from experiments carried out in the Multi-user Droplet Combustion Apparatus (MDCA) onboard the International Space Station (ISS).

Recommended citation: Xu, Y.; Shen, Y.; Avedisian, C. T.; Hicks, M. C.; Choi, M. Y. (2023). "Quantitative investigation of sooting dynamics in droplet combustion using an automated image analysis algorithm." Fuel. 354, 129313.
Download Paper

Application of Multi-Fidelity Transfer Learning with Autoencoders for Efficient Construction of Surrogate Models

Published in AIAA SCITECH 2024, 2024

The infusion of machine learning techniques into aerodynamic surrogate modeling stands out for its considerable reduction in computational costs, effectively capturing intricate flow dynamics, and complementing costly partial differential equation simulations. Despite their effectiveness, these machine learning-based models grapple with challenges related to generalization and computational demands. The primary objective of this study is to craft a surrogate model for aerodynamic prediction that strikes a balance between efficiency and accuracy. The strategy involves harnessing the power of transfer learning along with multi-fidelity data, aiming to curtail computational expenses while preserving precision. This is achieved by leveraging abundant low-fidelity data for initial training and subsequently fine-tuning the model with scarce high-fidelity data. The surrogate network, adopting an autoencoder-decoder architecture with super-resolution augmentation, is designed to map low-fidelity flow variables to their high-fidelity counterparts.

Recommended citation: Shen, Y.; Patel, H.; Xu, Z.; Alonso, J. (2024). "Application of Multi-Fidelity Transfer Learning with Autoencoders for Efficient Construction of Surrogate Models." AIAA SCITECH 2024. AIAA 2024-0013.
Download Paper

Performance Evaluation of a Graph Neural Network-Augmented Multi-Fidelity Workflow for Predicting Aerodynamic Coefficients on Delta Wings at Low Speed

Published in AIAA SCITECH 2025, 2025

Conceptual design of a stable and efficient supersonic transport (SST) aircraft during takeoff and landing, where higher angle-of-attack (AOA) induces complex aerodynamic phenomena such as vortex lift and flow separation, can be challenging. Existing analysis methods face a fidelity-cost trade-off: high-fidelity (HF) methods, such as computational fluid dynamics (CFD), offer prediction accuracy but are computationally expensive for solver-in-the-loop analyses, while low-fidelity (LF) methods, such as vortex lattice method (VLM), lack the ability to capture nonlinear flow physics. To address this gap, this paper proposes a multi-fidelity conceptual design analysis workflow that integrates a graph neural network (GNN)-based surrogate model into VLM to augment the analysis fidelity of LF tools. The surrogate model learns the discrepancies between LF and HF pressure fields, enabling accurate and efficient aerodynamic analyses on arbitrary quantities of interest. When evaluated on a dataset with various Delta wing geometries, the proposed workflow achieves an approximately fivefold reduction in the normalized root mean square error (NRMSE) for the predicted lift, drag, and pitching moment coefficients compared to using VLM alone. The results also highlight the proposed workflow’s generalizability across new flow conditions and wing geometries, while identifying its limitations in prediction accuracy variance across the test dataset. Overall, the proposed workflow provides an efficient and effective framework for aerodynamic assessment in conceptual design with improved fidelity.

Recommended citation: Shen, Y.; Alonso, J. (2025). "Performance Evaluation of a Graph Neural Network-Augmented Multi-Fidelity Workflow for Predicting Aerodynamic Coefficients on Delta Wings at Low Speed." AIAA SCITECH 2025..
Download Paper | Download Slides

VortexNet: A Graph Neural Network-Based Multi-Fidelity Surrogate Model for Field Predictions

Published in AIAA SCITECH 2025, 2025

Aircraft design relies on accurate aerodynamic predictions. High-fidelity (HF) methods, such as computational fluid dynamics (CFD), provide accurate aerodynamic analyses but are computationally expensive for early-stage design. Conversely, low-fidelity (LF) methods, such as the vortex lattice method (VLM), offer cost-effective solutions but struggle to capture complex flow phenomena, limiting their predictive accuracy. The conceptual design process thus presents a fidelity-cost trade-off, requiring a balance between high-fidelity (HF) and LF data in early design phases. This study introduces VortexNet, a graph neural network (GNN)- based surrogate model designed to bridge the fidelity gap between LF and HF aerodynamic predictions. VortexNet learns corrections to LF panel-wise local loading coefficient field data using data-driven insights from HF CFD simulations, enabling pressure coefficient field predictions across a range of Delta wing geometries and free-stream conditions. The model demonstrates strong prediction accuracy and generalizability across configurations, effectively capturing nonlinear flow features under geometric variations. A hyper-parameter sensitivity study and a preliminary prediction mechanism explanation, leveraging the latent space ablation technique, are conducted to rationalize the model’s predictive capabilities and provide guidance for future improvements in VortexNet-like surrogate modeling. These results indicate that VortexNet has potential as a valuable tool for conceptual design in multidisciplinary design optimization (MDO), while emphasizing the need for further validation and refinement.

Recommended citation: Shen, Y.; Needels, J.; Alonso, J. (2025). "VortexNet: A Graph Neural Network-Based Multi-Fidelity Surrogate Model for Field Predictions." AIAA SCITECH 2025. 1(0494).
Download Paper | Download Slides

ShapeBench: A Scalable Benchmark and Diagnostic Suite for Standardized Evaluation in Aerodynamic Shape Optimization

Published in Preprint, 2025

Aerodynamic shape optimization remains difficult to benchmark and compare across methods because datasets, evaluation protocols, and quality metrics vary widely across the literature. ShapeBench addresses this gap by introducing a large-scale benchmark and diagnostic suite for aerodynamic shape optimization. The benchmark includes standardized datasets, physics-informed evaluation metrics, and a framework for comparing optimization outcomes across representative aerodynamic design problems. We highlight how ShapeBench enables more consistent performance analysis and better diagnosis of optimization behavior across architectures and problem settings.

Recommended citation: Fazliani, S.; Chawla, K.; Guo, J.; Shen, Y.; Ihme, M.; Udell, M. (2025). "ShapeBench: A Scalable Benchmark and Diagnostic Suite for Standardized Evaluation in Aerodynamic Shape Optimization." Preprint.
Download Paper

Graph Neural Network-Guided Aerodynamic Shape Optimization for Conceptual Design of Supersonic Transport Wings

Published in AIAA Aviation 2025, 2025

Low-speed stability constraints have historically been difficult to incorporate during the conceptual design phase due to the limited accuracy of low-fidelity simulation methods. This paper presents a graph neural network (GNN)-guided multi-fidelity optimization framework for the aerodynamic shape design of supersonic transport wings at the conceptual level. A GNN-based field prediction surrogate model is integrated into the SUAVE aircraft design environment, enabling real-time, high-fidelity-informed aerodynamic shape optimization across a defined design space. Using a two-point, stability-constrained shape optimization problem, we demonstrate that the GNN-guided workflow produces a design that satisfies both stability and aerodynamic performance requirements, while the traditional vortex lattice method (VLM)- based workflow results in a configuration that violates the intended design objectives. The proposed approach improves the fidelity of aerodynamic analysis and enhances confidence in conceptual design, increasing the likelihood that the resulting configuration will perform well in later design phases.

Recommended citation: Shen, Y.; Alonso, J. (2025). "Graph Neural Network-Guided Aerodynamic Shape Optimization for Conceptual Design of Supersonic Transport Wings." AIAA Aviation 2025
Download Paper | Download Slides

A Multi-Fidelity Double-Delta Wing Dataset and Empirical Scaling Laws for GNN-based Aerodynamic Field Surrogates

Published in AIAA SCITECH 2026, 2026

Data-driven surrogate models are increasingly adopted to accelerate vehicle design, but open-source multi-fidelity datasets and empirical guidelines linking dataset size to model performance remain limited. This study investigates the relationship between training data size and prediction accuracy for a graph neural network (GNN)-based surrogate model for aerodynamic field prediction. We release an open-source, multi-fidelity aerodynamic dataset for double-delta wings spanning 272 geometries and 2448 flow snapshots evaluated with both vortex lattice method (VLM) and Reynolds-averaged Navier–Stokes (RANS) solvers. Using the dataset, we conduct a preliminary empirical scaling study of the MF-VortexNet surrogate and find that test error decreases with dataset size according to a power-law relation, providing guidance for future data collection and model design in multi-fidelity aerodynamic surrogate modeling.

Recommended citation: Shen, Y.; Alonso, J. (2026). "A Multi-Fidelity Double-Delta Wing Dataset and Empirical Scaling Laws for GNN-based Aerodynamic Field Surrogates." AIAA SCITECH 2026.
Download Paper

A Multi-Fidelity Workflow for Conceptual Design Using Graph Neural Network-Based Field Prediction Surrogate Model

Published in Computers & Fluids, 2026

This journal article presents a multi-fidelity conceptual design workflow that uses a graph neural network-based field prediction surrogate model to augment low-fidelity aerodynamic analysis with high-fidelity CFD information.

Recommended citation: Shen, Y.; Needels, J.; Alonso, J. (2026). "A Multi-Fidelity Workflow for Conceptual Design Using Graph Neural Network-Based Field Prediction Surrogate Model." Computers & Fluids. 317, 107153.
Download Paper

Multi-Fidelity Aerodynamic Surrogate Modeling for Conceptual Design Using Graph Neural Networks

Published in Stanford University PhD Thesis, 2026

This dissertation develops multi-fidelity aerodynamic surrogate modeling frameworks for conceptual design using graph neural networks. It introduces methods for augmenting low-fidelity aerodynamic tools with high-fidelity CFD field information, builds and validates graph-based prediction models for delta-wing aerodynamics, and demonstrates their use in design-space exploration and optimization under realistic aerodynamic constraints.

Recommended citation: Shen, Y. (2026). "Multi-Fidelity Aerodynamic Surrogate Modeling for Conceptual Design Using Graph Neural Networks." Stanford University PhD Thesis.
Download Paper

talks

AI-driven Simulations for Automotive Industry

Published:

This talk surveys how AI-driven simulation, including machine learning-based surrogate models and multi-fidelity analysis, can accelerate engineering design and assessment workflows in the automotive industry.

VortexNet: A Graph Neural Network-Based Multi-Fidelity Surrogate Model for Field Predictions

Published:

This talk introduces VortexNet, a graph neural network (GNN)-based surrogate model designed to bridge the fidelity gap between low-fidelity (LF) and high-fidelity (HF) aerodynamic predictions. VortexNet learns corrections to vortex lattice method (VLM) panel-wise loading coefficient fields using data from high-fidelity CFD simulations, enabling pressure coefficient field predictions across a range of delta wing geometries and free-stream conditions. The talk covers the model architecture, its prediction accuracy and generalizability across configurations, and its potential role in conceptual design and multidisciplinary design optimization (MDO).

Multi-Fidelity Aerodynamic Surrogate Modeling for Conceptual Design

Published:

This talk presents MF-VortexNet, a multi-fidelity aerodynamic surrogate framework that uses graph neural networks to learn field-level corrections to low-fidelity pressure predictions from high-fidelity CFD data. By operating on pressure-field information, MF-VortexNet improves interpretability, enables the evaluation of arbitrary aerodynamic quantities, and supports generalization across a prescribed design space. The talk has strong focus on introducing an open-source multi-fidelity aerodynamic dataset, empirical scaling behavior of surrogate error with training data size, and integration of the framework into a conceptual design environment for stability-constrained shape optimization.

Multi-Fidelity Aerodynamic Surrogate Modeling for Conceptual Design Using GNNs

Published:

Conceptual aircraft design requires major engineering decisions under limited information. In practice, low-fidelity aerodynamic tools are routinely used because higher-fidelity simulations are often unavailable or too expensive for early-stage design. This creates a fidelity-cost gap: the level of analysis fidelity needed for confident decision-making is not available at the stage with the greatest downstream impact. This challenge is especially important in supersonic transport design, where nonlinear aerodynamic effects make considerations such as low-speed stability difficult to assess using conventional conceptual design tools.

teaching

Sustainable Aviation (AA260)

Graduate course, Teaching Assistant, Stanford University, Department of Aeronautics and Astronautics, 2023

Teaching Assistant, Spring 2023 (Apr. 2023 - Jun. 2023).

Engineering of Systems (ENGR21)

Undergraduate course, Teaching Assistant, Stanford University, School of Engineering, 2024

Teaching Assistant, Spring 2024 (Apr. 2024 - Jun. 2024).

Applied Aerodynamics (AA200)

Graduate course, Teaching Assistant, Stanford University, Department of Aeronautics and Astronautics, 2026

Teaching Assistant, Winter 2026 (Jan. 2026 - Mar. 2026).