03.48 · Uncertainty & sensitivity

A sampling study turns explicitly stated input variations into a distribution of calculated responses. This exercise teaches you to distinguish the assumed variation, the numerical sampling procedure and the interpretation of the resulting spread.

Illustrative calibration only. Start with a 16-ply cross-ply reference laminate; this is a separate, reference-state study.

Open this exercise in Workbench

Prepare the baseline

Inspect the bounded inputs, distribution assumptions, sample settings and seed. Establish the supplied run before changing one range or sampling control. Preserve the baseline laminate and loading so that differences between sample sets can be traced to the intended change.

Worked procedure

1. Follow the study’s laminate link and inspect the stack and resolved properties that it will use. Then read the model scope and identify the additional specimen, loading or calibration inputs belonging to this separate study. Record which values are supplied independently rather than assuming that every required quantity is inherited from the laminate.

2. Run the supplied study as a baseline and inspect the complete response curves, including their axes, units and parameter settings. Retain the numerical values or a clearly labelled capture before changing an input. Use the interpretation guidance below and the linked formulation to identify what each curve represents and which conclusions remain outside its scope.

3. Choose one editable parameter that belongs to this model and record its original and revised values. Keep the other inputs fixed, rerun the study, and compare the same output quantities over the same range. Explain the observed change using the linked formulation, including a discussion of whether the comparison stays within the model’s calibration and assumptions.

Review checkpoints

Do not interpret example calibration values as material allowables.

Record the assumptions and distinguish analytical verification from experimental validation.

Model limits

Seeded independent uniform sampling of shared modulus, ply-thickness and angle offsets. Recomputes laminate ABD for each sample. Bounds are assumptions, not measured distributions. Percentiles and sensitivity are exploratory, not reliability certification.

Interpret the comparison

Read percentiles and correlations together with the inputs that generated them. A repeatable seeded sample is useful for comparison, but it is not evidence that the chosen distributions represent manufacturing variability. A small linear correlation does not rule out nonlinear influence, and the study should not be labelled certified reliability.

How information passes between models

Micro → Laminates: Predicted ply stiffness, strength, density and expansion properties.

Materials → Micro: Constituent stiffness, strength, density and thermal / moisture properties.

Mechanical → Simulation: SIMULATION selects this case and its analysis model; the case owns its applicable cycle and input references.

Laminates → Mechanical: Ply angles and thicknesses, stiffness, mass and ply properties.

Models → Micro: Applied model assignment: Halpin–Tsai. Model parameters and formulation are used by Micro.

Models → Mechanical: Applied model assignment: Seeded laminate uncertainty. Model parameters and formulation are used by Mechanical.

Further reading and evidence

• Advanced analytical models: equations, calibration and limits
• Edit laminate materials, angles and thicknesses
• Run and review a model
• Connected inputs and result freshness

Review the recorded validation scope. Retain the original inputs and solver notices with the results. Representative teaching data are not design allowables.