Engineering Prompt Library

Engineers often get poor AI answers because prompts lack context, assumptions, and verification requirements. This library helps CFD and simulation engineers ask better, safer, more verifiable questions.

Crucial Reminder: AI does not replace engineering judgment, experimental validation, solver documentation, or professional review. Always verify AI outputs against physics principles and trusted literature.

Prompt Categories

Engineering AI Safety

Published

A framework for knowing when to use AI and when human verification is strictly required.

CFD Setup Prompts

Published

Guides for structuring prompts related to domain, boundaries, and solver selection.

Turbulence Model Prompts

Published

Prompts to safely interrogate AI about k-epsilon, k-omega, LES, and wall functions.

Mesh and y+ Prompts

Published

Structured questions for mesh sizing, boundary layer resolution, and grid independence.

Convergence & Solver Logs

Published

How to paste solver logs and ask for debugging help without losing context.

Python Automation Prompts

Planned

Generating safe, verifiable scripts for Paraview, PyVista, and other flow-solver automation.

Validation & Verification

Planned

Prompts to cross-check results against analytical solutions and literature.

Engineering Report Prompts

Planned

Structuring AI requests to help draft methodology and result sections.

Literature/Research Prompts

Planned

How to query academic papers and extract numerical assumptions safely.

Manager Review Prompts

Planned

Reviewing AI-assisted work for hallucinations and physical impossibilities.

Bad Prompt vs. Better Prompt

Bad Prompt

"What do you know about the k-epsilon model?"

  • Too broad and ambiguous.
  • Will return generic Wikipedia-level information.
  • No context on the specific fluid, Reynolds number, or wall treatment.

Better Prompt

"I am running a 2D steady-state RANS simulation of water flow through a sudden pipe expansion (Re = 50,000). I plan to use the Realizable k-epsilon model with standard wall functions. What are the key assumptions of this model for this specific flow regime, and what specific validation checks should I perform on the reattachment length?"

  • Provides exact context and physics.
  • Asks for targeted assumptions and failure modes.
  • Requests specific validation checks.

Standard Prompt Structure

To get engineering-grade answers, your prompts should consistently include the following elements:

  • Context: What is the physical problem? (e.g., fluid type, dimensions)
  • Objective: What exactly are you trying to achieve?
  • Known Inputs: What parameters are already defined? (e.g., Re, Ma)
  • Constraints: What solver or computational limits exist?
  • Expected Output: Do you want a list, a table, or a script?
  • Verification Checklist: How will you prove the answer is correct?
  • Risk Level: Is this for a critical design decision or just brainstorming?

Use AI Safely in Engineering

  • 1.Never trust physics intuition from an LLM. They are language engines, not physics engines. Always calculate.
  • 2.Demand citations. Ask the AI to point to specific textbooks or solver manuals, then manually verify those sources.
  • 3.Strip proprietary data. Never paste confidential CAD geometry details, proprietary material data, or client IP into public AI tools.
  • 4.Validate equations. If the AI provides an equation or Python script, independently derive it or test it on a trivial benchmark case first.