Skip to content

PD Dataset Generation

Maxim Nikitin edited this page Jun 20, 2026 · 4 revisions

Dataset Generation

A comprehensive design space is generated using a Sobol sequence (218 ≈ 262,144 points) spanning 6 input parameters:

Parameter Symbol Range Units
Bed diameter $D$ 0.1 – 5 m
Bed length $L$ 1·D – 500·D m
Particle diameter $D_p$ D/500 – D/1.5 m
Particle sphericity $\phi_s$ 0.5 – 1
Bed void fraction $\varepsilon$ 0.36 – 0.9
Superficial gas velocity $U_{s,G,in}$ 0.01 – 5 m/s

Each source model is evaluated at every design point. Predictions are masked with NaN where the point falls outside the model's validity range. The output is the extended output matrix $\mathcal{Y}_\text{ext}$ (one column per model) and the initial validity matrix $\mathcal{V}_\text{ext}$ (1 = valid, NaN = out of range).

Clone this wiki locally