Uploaded May 2026 | Updated September 2026, 3 weeks ago
Recorded 20 May 2026. Yunan Yang of Cornell University presents "Adjoint Direct Simulation Monte Carlo (DSMC) for Optimization and Control of Rarefied Flows" at IPAM's Multi-Fidelity Methods to Enable Robust Optimization and Real-Time Control of Fusion Processes Workshop.
Abstract: We derive adjoint equations for Direct Simulation Monte Carlo (DSMC) methods applied to the spatially inhomogeneous Boltzmann equation. The formulation incorporates spatial transport and a range of boundary conditions, including periodic boundaries, specular and thermal reflection, and prescribed inflow. The resulting adjoint system enables efficient gradient evaluation for quantities of interest defined through particle-based simulations. Numerical experiments validate the accuracy of the adjoint formulation and demonstrate its use in sensitivity analysis and gradient-based optimization. This work provides a foundation for integrating kinetic-scale particle methods into multi-fidelity optimization and control frameworks, with potential applications to rarefied plasma flows and edge physics in fusion devices.
Learn more online at: https://www.ipam.ucla.edu/programs/workshops/workshop-iv-multi-fidelity-methods-to-enable-robust-optimization-and-real-time-control-of-fusion-processes/?tab=overview
Recorded 20 May 2026. Yunan Yang of Cornell University presents "Adjoint Direct Simulation Monte Carlo (DSMC) for Optimization and Control of Rarefied Flows" at IPAM's Multi-Fidelity Methods to Enable Robust Optimization and Real-Time Control of Fusion Processes Workshop.
Abstract: We derive adjoint equations for Direct Simulation Monte Carlo (DSMC) methods applied to the spatially inhomogeneous Boltzmann equation. The formulation incorporates spatial transport and a range of boundary conditions, including periodic boundaries, specular and thermal reflection, and prescribed inflow. The resulting adjoint system enables efficient gradient evaluation for quantities of interest defined through particle-based simulations. Numerical experiments validate the accuracy of the adjoint formulation and demonstrate its use in sensitivity analysis and gradient-based optimization. This work provides a foundation for integrating kinetic-scale particle methods into multi-fidelity optimization and control frameworks, with potential applications to rarefied plasma flows and edge physics in fusion devices.
Learn more online at: https://www.ipam.ucla.edu/programs/workshops/workshop-iv-multi-fidelity-methods-to-enable-robust-optimization-and-real-time-control-of-fusion-processes/?tab=overview
