Engineers running parameter studies on Rescale now have two new approaches for setting up and executing Design of Experiments, both powered by the Rescale Assistant and Workflows. Depending on the study’s complexity, engineers can choose between Workflow Runs for maximum flexibility or a single Workflow Template for centralized control. 

This post walks through both methods and when to use each.

Design of Experiments on Rescale Today

Rescale has long supported a built-in Design of Experiments (DOE) job type. Engineers create a single job, define variables in Parallel Settings, upload a templated input file, and submit. Rescale fans out child runs across a shared cluster and consolidates results into a DOE table.

This works well for straightforward parameter sweeps with no dependencies between runs and short runtimes that benefit from shared compute. But it has limits. Child runs don’t support Custom Fields or Automations, which means results can’t easily feed into AI Physics model training or be used by the Rescale Assistant. And the available DOE algorithms are limited to Cross Product, Monte Carlo, and CSV upload.

What Rescale Assistant Brings to DOE

Rescale Assistant can design the full DOE matrix. Describe your parameters, their ranges, and your study goals, and the Rescale Assistant recommends and generates the appropriate design:

• Full Factorial
• Latin Hypercube Sampling
• Fractional Factorial
• Box-Behnken
• Central Composite Design
• Monte Carlo sampling

Once the matrix is approved, Rescale Assistant submits the study directly through Workflows. No CSV export, no manual configuration.

Method 1: Rescale Assistant + Workflow Runs

Build a Workflow template with your simulation parameters exposed as Workflow parameters. Rescale Assistant generates the DOE matrix, then creates one Workflow Run per design point, passing each row’s values automatically.

How it works:

1. Create a Workflow template with solver job nodes. Expose the variables to sweep (inlet velocity, angle of attack, mesh density) as Workflow parameters.
2. Ask Rescale Assistant to generate your DOE matrix.
3. Rescale Assistant creates one Workflow Run per design point.

Each run is a fully independent Rescale job with Custom Fields, Automations, and license queuing.

When to use Workflow Runs:

• Pre- and post-processing per run. Each Workflow Run can include nodes before and after the solver — mesh generation, file prep, result extraction. In a single Workflow Template with dozens of jobs, that wiring gets complicated fast.
• All runs start simultaneously. Every Workflow Run launches independently, so the full DOE kicks off at once.
• Fault isolation. If one run fails or gets spot-killed, the rest keep going. Re-run just the failed design point.

Method 2: Rescale Assistant + Workflow Template

Rescale Assistant generates the DOE matrix and builds a single Workflow Template containing all design points as job nodes on one canvas.

How it works:

1. Set up an example job with the solver, input files, hardware, and command configuration.
2. Ask Rescale Assistant to generate the DOE matrix.
3. Rescale Assistant builds a Workflow Template with one job node per design point, each configured with that row’s parameter values.
4. Optionally, connect job nodes to chain simulations where one run’s output initializes the next.
5. Click Run Workflow.

When to use a Workflow Template:

• Review before submitting. The entire DOE is visible on a single canvas before any compute starts. Engineers can verify parameter assignments and job dependencies before committing resources.
• A single record of the study. One Workflow Run captures the full DOE — what was submitted, how it was configured, and what happened.
• Centralized monitoring. Every design point’s status (queued, running, completed, failed) is visible on one canvas. This live monitoring surface is especially useful for studies with many jobs. 
• Concurrency limits. Control how many DOE jobs run simultaneously. Useful when license tokens are limited or when compute budgets need to be managed across a study.

Choosing Between Different Options

	DOE Job Type	Workflow Runs	Workflow Template
Pre/post-processing per run	Pre/post-processing as part of Job execution	Pre/post-processing in a pipeline as separate compute (Jobs, Workstation)	Pre/post-processing as part of Job execution
All runs start at once	Depends on Slots or number of parallel DOE task executions specified	Yes	Depends on graph structure
Review DOE before submitting	No	No — runs launch individually	Yes — full canvas visible first
Single monitoring surface	Yes (DOE results table)	No	Yes (Workflow canvas)
Concurrency control	Yes (by specifying Slots)	No	Yes
Record of full study	Single parent job	Scattered across runs	One Workflow Run
Custom Fields & Automations	No	Yes	Yes
AI Physics readiness	No	Yes	Yes
Advanced DOE algorithms	CSV upload only	Yes (via Rescale Assistant)	Yes (via Rescale Assistant)
Fault isolation	Must re-run entire DOE	Re-run one Workflow Run	Re-run individual nodes

See It in Action

Getting Started

Any solver on Rescale works with both methods. The DOE approach is independent of the simulation software — Rescale Assistant and Workflows handle the parameter variation; the solver runs inside the job as usual.

To try it:

1. Start with Rescale Assistant. Describe your parameters, ranges, and study goals. Rescale Assistant recommends a DOE algorithm and generates the matrix.
2. Pick your method. Workflow Runs for per-run flexibility and parallel launch. Workflow Template for review, monitoring, and concurrency control.
3. Submit and monitor. Rescale Assistant handles the setup. Results land in Custom Fields, ready for analysis or AI Physics model training.

For detailed setup instructions, see the Workflows documentation and the DOE & Parameter Sweeps guide.