Accelerating fluid dynamics problems in planet formation with machine learning
| dc.contributor.author | Mao, Shunyuan | |
| dc.contributor.supervisor | Dong, Ruobing | |
| dc.date.accessioned | 2024-08-23T16:32:06Z | |
| dc.date.available | 2024-08-23T16:32:06Z | |
| dc.date.issued | 2024 | |
| dc.degree.department | Department of Physics and Astronomy | |
| dc.degree.level | Doctor of Philosophy PhD | |
| dc.description.abstract | I develop two machine learning tools for solving forward and inverse problems in protoplanetary disks. The first tool, Protoplanetary Disk Operator Network (PPDONet), predicts the solution of disk--planet interactions in real--time. PPDONet is based on Deep Operator Networks (DeepONets), a class of neural networks capable of learning non--linear operators to represent deterministic and stochastic differential equations. It maps three scalar parameters in a disk--planet system -- the Shakura \& Sunyaev viscosity $\alpha$, the disk aspect ratio $h_\mathrm{0}$, and the planet--star mass ratio $q$ -- to steady--state solutions of the disk surface density, radial velocity, and azimuthal velocity. Comprehensive testing demonstrates the accuracy of PPDONet, with predictions for one system made in less than a second on a laptop. A public implementation of PPDONet is available at \url{https://github.com/smao-astro/PPDONet}. The second tool, Disk2Planet, infers key parameters in disk-planet systems from observed disk structures. It processes two-dimensional density and velocity maps to output the Shakura--Sunyaev viscosity, disk aspect ratio, planet--star mass ratio, and the planet's location. Disk2Planet integrates the Covariance Matrix Adaptation Evolution Strategy (CMA-ES), an evolutionary algorithm for complex optimization problems, with PPDONet. Fully automated, Disk2Planet retrieves parameters within three minutes on an Nvidia A100 GPU, achieving accuracies ranging from thousandths to percentages. It effectively handles data with missing parts and unknown levels of noise. Together, these tools advance the field of planet formation by providing rapid, accurate solutions and parameter inferences for disk-planet systems, enhancing our understanding of the underlying physics of protoplanetary disks. | |
| dc.description.embargo | 2025-08-14 | |
| dc.description.scholarlevel | Graduate | |
| dc.identifier.bibliographicCitation | Mao, S., Dong, R., Lu, L., Yi, K. M., Wang, S., & Perdikaris, P. (2023). PPDONet: Deep Operator Networks for Fast Prediction of Steady-state Solutions in Disk–Planet Systems. The Astrophysical Journal Letters, 950(2), L12. | |
| dc.identifier.uri | https://hdl.handle.net/1828/20301 | |
| dc.language | English | eng |
| dc.language.iso | en | |
| dc.rights | Available to the World Wide Web | |
| dc.subject | protoplanetary disks | |
| dc.subject | planetary-disk interactions | |
| dc.subject | hydrodynamical simulations | |
| dc.subject | neural networks | |
| dc.title | Accelerating fluid dynamics problems in planet formation with machine learning | |
| dc.type | Thesis |