pasys doe¶

Run a Design of Experiments (DOE) batch study.

Synopsis¶

pasys doe <config> [options]

Description¶

The doe command generates a design space sample and evaluates all cases. Results are saved as Parquet files for further analysis.

Arguments¶

Argument	Required	Description
`config`	Yes	Path to configuration file (YAML or JSON)

Options¶

Option	Short	Default	Description
`--output`	`-o`	`./results`	Output directory
`--samples`	`-n`	`50`	Number of DOE samples
`--method`		`lhs`	Sampling method: `lhs`, `random`, `grid`
`--seed`		`42`	Random seed
`--workers`	`-j`	`1`	Parallel workers

Sampling Methods¶

Latin Hypercube Sampling (LHS)¶

Default and recommended. Space-filling design:

pasys doe config.yaml -n 100 --method lhs

Random Sampling¶

Uniform random:

pasys doe config.yaml -n 100 --method random

Grid Sampling¶

Full factorial (use with small design spaces):

pasys doe config.yaml --method grid
# Warning: Grid size grows exponentially!

Parallel Execution¶

Use multiple workers for faster evaluation:

# 4 parallel workers
pasys doe config.yaml -n 200 -j 4

Output¶

Results are saved to the output directory:

results/
└── results.parquet    # All case results

The Parquet file contains:

Case ID and configuration parameters
All computed metrics
Requirement verification results
Metadata (runtime, errors)

Examples¶

Basic DOE¶

pasys doe config.yaml -n 100

Output:

Design Space: 5 variables
Generating 100 samples using lhs...
Generated 100 cases using Latin Hypercube Sampling
Running batch evaluation...
  Progress: 10/100 (10%)
  Progress: 20/100 (20%)
  ...
  Progress: 100/100 (100%)

Completed: 100 cases
Feasible: 73 (73.0%)

Results saved to: ./results/results.parquet

Full Study Pipeline¶

# Run DOE with custom settings
pasys doe study_config.yaml \
  -n 200 \
  --method lhs \
  --seed 12345 \
  -j 4 \
  -o ./study_results

# Extract Pareto
pasys pareto study_results/results.parquet -x cost_usd -y eirp_dbw

# Generate report
pasys report study_results/results.parquet -o report.html

Reproducibility¶

Use the same seed for reproducible results:

# These produce identical results
pasys doe config.yaml -n 100 --seed 42
pasys doe config.yaml -n 100 --seed 42

Configuration Requirements¶

The configuration file must include a design_space section:

# config.yaml
name: "Trade Study"

architecture:
  # ... baseline configuration

scenario:
  # ... scenario definition

requirements:
  # ... requirements list

design_space:
  variables:
    # Variable parameters
    - name: array.nx
      type: categorical
      values: [4, 8, 16]

    - name: array.ny
      type: categorical
      values: [4, 8, 16]

    - name: rf.tx_power_w_per_elem
      type: float
      low: 0.5
      high: 3.0

    # Fixed parameters (low == high)
    - name: array.geometry
      type: categorical
      values: [rectangular]

    - name: rf.pa_efficiency
      type: float
      low: 0.3
      high: 0.3

Loading Results¶

import pandas as pd

results = pd.read_parquet("results/results.parquet")

# Feasible designs
feasible = results[results["verification.passes"] == 1.0]

# Analyze
print(f"Total: {len(results)}")
print(f"Feasible: {len(feasible)}")
print(f"Cost range: ${results['cost_usd'].min():,.0f} - ${results['cost_usd'].max():,.0f}")

Error Handling¶

# Missing design_space
pasys doe config_no_doe.yaml
# Error: Config must define a design_space for DOE

# Invalid variable specification
# Error messages indicate which variable has issues