Skip to main content

Table 2 Summary of the collision datasets in this work. Each simulation requires two unique models to serve as the target and projectile. In this work, we combined six distinct datasets to create a dataset of 14,856 collisions. The 12D_LHS200 dataset was additionally used to study the convergence of the post-impact parameters, as a convergence study the larger datasets was computationally infeasible. The 12D_LHSLOW dataset was simulated to study low asymptotic relative velocities from 0.1-1 \(v_{\mathrm{esc}}\)

From: Machine learning applied to simulations of collisions between rotating, differentiated planets

Dataset

Type

Collisions

Models

\(v_{\infty }\) (\(\mathrm{v}_{\mathrm{esc}}\))

12D_LHS10K

ARSM

10,000

20,000

1–10

12D_LHS500

LHS

500

1000

1–10

12D_LHS200

LHS

200

400

1–10

12D_LHSLOW

LHS

3384

6768

0.1–1

6D_LHS500

LHS

500

1000

1–10

4D_LHS500

LHS

500

1000

1–10

all

Composite

14,856

29,712

0.1–10

train

Composite

11,884

23,768

0.1–10

test

Composite

2972

5944

0.1–10