Simulation-Informed Decisions
Simulation as a Decision Tool
Simulation in engineering isn't new. What's new is using simulation systematically within a digital engineering framework — where simulations are connected to the system model, configured from model data, and their results feed back into engineering decisions automatically.
The Simulation Pyramid
Not all simulations serve the same purpose. Think of them in layers:
Screening (Fast, Low Fidelity)
- Analytical equations, parametric models, lookup tables
- Runs in milliseconds
- Used to: Eliminate obviously bad alternatives quickly
- Example: Mass and power budget calculations
Exploration (Medium Fidelity)
- Reduced-order models, simplified FEA, 1D thermal analysis
- Runs in minutes to hours
- Used to: Compare design alternatives, identify sensitivities
- Example: Simplified structural analysis of 50 bracket variants
Validation (High Fidelity)
- Full FEA, CFD, multi-physics, Monte Carlo
- Runs in hours to days
- Used to: Validate the selected design against requirements
- Example: Full thermal-structural analysis of the final bracket design
Certification (Highest Fidelity)
- Calibrated, validated models with documented uncertainty
- May require physical test correlation
- Used to: Satisfy regulatory or contractual requirements
- Example: Crash simulation for automotive safety certification
The key insight: use the right fidelity for the decision at hand. Don't run a 3-day CFD simulation to decide between two concepts that differ by 40% in performance — a 5-minute analytical estimate is sufficient.
Connecting Simulations to the Model
In a digital engineering workflow, simulations aren't standalone activities. They're connected to the system model:
Inputs from the Model
- Geometry from the CAD model
- Material properties from the system model
- Load cases from the requirements model
- Interface conditions from the architecture model
Outputs to the Model
- Performance predictions (temperature, stress, deflection)
- Margin calculations (how close to the limit?)
- Pass/fail verdicts against requirements
- Sensitivity data (which parameters matter most?)
Traceability
Each simulation run is traceable to:
- The model version it used as input
- The simulation tool and configuration
- The results and their interpretation
- The decision it informed
Simulation Automation
Manual simulation workflows — export geometry, set up mesh, define boundary conditions, run solver, post-process results — are labor-intensive and error-prone. Automation addresses both:
Parametric Sweeps
Define the parameters to vary and their ranges. The automation framework generates simulation configurations, submits jobs, collects results, and presents them in a structured format.
Design of Experiments (DoE)
Instead of sweeping every combination (expensive), DoE methods (Latin Hypercube, Sobol sequences) sample the design space efficiently — maximizing information per simulation run.
Surrogate Model Training
Run a DoE with the full-fidelity simulator, then train a surrogate model (Gaussian process, neural network) on the results. The surrogate predicts simulation outputs in milliseconds, enabling optimization and exploration at a scale the full simulator can't support.
Common Traps
Analysis Paralysis
Running more and more simulations without making a decision. Define decision criteria before running the simulation — what result would change your mind?
False Precision
A simulation reports stress to six decimal places. That doesn't mean it's accurate to six decimal places. Understand uncertainty and communicate results with appropriate confidence.
Simulation as Validation Theater
Running simulations to confirm a decision already made, ignoring results that contradict the preferred outcome. The value of simulation is in surprise — results that challenge assumptions.