Automotive & data work

Conditions to record before comparing simulation peak values

If you only put peak stress numbers side by side, differences in mesh, output interval, and evaluation location can look like design changes. Build a comparison condition table from synthetic cases and record the numerical difference together with the range you can actually judge.

Content checked 2026.09.19Example files included

Show contents

This translation was generated by AI. Check the code, units, and numbers against the original. Native-speaker review has not yet been completed for each language. 한국어

Who this is forEngineers who compile vehicle or component simulation results or confirm comparison conditions with the analyst

What you need

Besides the result images, obtain the model version, analysis settings, and result extraction conditions.
First define the physical quantity, unit, spatial region, and time window to compare.
Leave the supporting files and unresolved conditions so that the responsible expert in that analysis field can review the actual judgment.

01Attach the conditions next to the peak value

A peak value is a result found by a specific extraction method across the whole model or a specified region. If the part, the time, or the stress definition is missing, you cannot say the same target is being compared, even with the same units. The deliverable of this article is not a ranking of numbers but a condition record that shows whether a comparison is valid and what still needs checking.

02First define which peak value it is

The metric in the example is “the maximum von Mises stress saved in the root-ROI region between 0 and 20 ms.” Distinguish the maximum over the whole region from the maximum of the time history at a specific coordinate. The two values search different spatial ranges, so name each one and keep them in separate columns.

Quantity: record exactly which one, such as von Mises, principal stress, or a specific directional component.
Location: record where it is evaluated, such as integration points, element nodes, or averaged nodes, along with the averaging and extrapolation settings.
Region: record the part or set name, the selection criteria, and any excluded zones with the reason.
Time: record the start and end times, the saved times, and whether interpolation was used.
Coordinates: record the coordinate system name and the coordinates of the peak. When the mesh changes, do not identify the same physical location by node number alone.

03Order of recording before comparing

Record the file names, model revision, solver version, and analysis type of both results, and keep the original results.
Check that the material model, loads, constraints, contacts, and unit system match. If the purpose is a design change, state what was changed in a separate column.
Record the element type and order, local mesh size and quality, and how the region of interest was meshed. Do not treat two meshes as the same condition just because the total element count matches.
Record the solver time step and tolerance separately from the result output interval. Do not conclude that the calculation became more accurate just because the saved times became denser.
Align the evaluation location, averaging, coordinate system, and time window, then extract the peak value with its location and time.
If the conditions differ, mark the differences first. Have the responsible engineer separate items that can be aligned by re-extraction from items that require re-running the analysis.
Calculate the numerical difference, but record separately the range for which convergence, the physical model, and correlation with measurements have been reviewed.

04Synthetic comparison table for three cases

Condition	A	B	C
Geometry revision	R01	R01	R02
Local mesh size	4 mm	2 mm	4 mm
Result output interval	1 ms	0.2 ms	1 ms
Stress evaluation	Averaged nodal	Unaveraged element nodal	Averaged nodal
Region & time	root-ROI, 0~20 ms	root-ROI, 0~20 ms	root-ROI, 0~20 ms
Region maximum	214 MPa	248 MPa	205 MPa
Time of maximum	8 ms	8.4 ms	8 ms
Coordinates of maximum (mm)	(12, 3, 0)	(12.3, 3, 0)	(11.8, 3, 0)

A and B have the same geometry, but the mesh, output interval, and stress evaluation all differ at once. You can record the arithmetic fact that 248 MPa is higher than 214 MPa, but this table cannot separate which condition caused the increase. Nor does it support the explanation that the design got worse.

A and C are assumed to compare a geometry change, and the extraction conditions in the table are the same. However, there is no actual calculation history or mesh convergence data, so a physically valid design comparison has not been completed. Even with the same nominal mesh size, when the geometry changes you must recheck the local element layout and quality.

05What differences in mesh, time, and averaging mean

COMSOL's guidance on mesh refinement explains choosing a quantity of interest and comparing how results change on finer meshes. Local values can converge more slowly than integrated quantities, and near singularities at sharp corners local results may not converge even with refinement. When the peak keeps growing, do not pick the number from the smallest elements as the answer; review the model idealization and the evaluation method.

The result output interval and the time step the solver uses internally can be different. COMSOL's time-dependent examples also distinguish output times from adaptive internal steps. A peak taken from sparse output times can miss a peak between outputs, and reducing the output interval alone does not guarantee accurate time integration.

Averaged nodal stress and unaveraged element nodal stress are different post-processing results. Ansys's official documentation distinguishes the values of elements connected to a node from the averaged display. Programs can differ in the order of extrapolation, averaging, and invariant calculation, so check the settings and definitions in the version you actually use.

06Write the arithmetic difference and the engineering conclusion separately

B와 A의 영역 최대값 차이 = 248 − 214 = +34 MPa
A를 기준으로 한 상대 차이 = (248 − 214) ÷ 214 × 100 ≈ +15.9 %

C와 A의 영역 최대값 차이 = 205 − 214 = −9 MPa
A를 기준으로 한 상대 차이 = (205 − 214) ÷ 214 × 100 ≈ −4.2 %

Example record sentence: “The saved-result peak for synthetic case B is 34 MPa higher than A. Because mesh, output interval, and evaluation method differ, a causal comparison is withheld. C is 9 MPa lower than A, but this example has no evidence of convergence or validation against a real model.” If the reference value for a relative difference is 0, do not calculate a percentage; record the absolute difference and why it cannot be calculated.

If the coordinates of the region maximum moved, also look at the value at a specific coordinate as needed. At the fixed coordinate (10, 3, 0) mm in the synthetic data, the time-history maximum is 210 MPa for A and 199 MPa for C. This is a different metric from the region maxima of 214 MPa and 205 MPa. Whether that coordinate refers to the same material location after the geometry change must be checked separately in the real model.

07Check before handing over the comparison table

Check that each value in synthetic-cases.json matches the table. Do not remove the synthetic label from the data description.
Do not treat blanks in the condition table as the same or as defaults; mark them as unconfirmed.
Do not keep only the peak number; keep the region, location, time, averaging, and units with it.
When comparing mesh changes, use the same physical region and agree with an expert on how to assess convergence of the quantity of interest.
For dynamic problems, check not only the saved times but also related settings such as internal steps, tolerances, and numerical damping.
If you exclude a suspected singularity zone, record the reason and the evaluation method. Do not exclude areas arbitrarily to get a lower number.
Review the numerical differences in the report separately from judgments about cause, safety, or performance.

08Common errors and limits of this example

Common error	Record to keep
Judging it improved because the color is less red	Same legend range, actual values, and extraction conditions
Comparing the same node number across different meshes	Coordinate system, physical location, and region correspondence
Recording the output interval as the internal solver step	Record the two values as separate items
Comparing averaged and unaveraged values as the same metric	Evaluation location and averaging settings
Writing a safety conclusion from a single peak value	Expert review of the model, criteria, and validation in that field

No actual finite element analysis was performed for this material, and mesh convergence, time convergence, material models, contact, and correlation with tests were not verified. Fatigue, crash, welding, composites, and similar fields use different metrics and evaluation methods, so they need review by an expert in that field. The differences calculated here only check the arithmetic relationships between synthetic numbers.

README.txt: scope of use of the synthetic data
synthetic-cases.json: synthetic cases with conditions and peak values
comparison-checklist.txt: condition table to fill in before reviewing real results
expected-differences.json: arithmetic check results for the synthetic peak values

References were checked at official sources on September 19, 2026. Product-specific screen operations are not covered; check the result definitions for the version of the analysis software you use.

Execution and verification record

Windows local Python standard library Fraction; official sources checked 2026-09-19

Checked the condition keys, units, time windows, and peak value records of the 3 synthetic cases
Checked B−A = 34 MPa, C−A = −9 MPa, and the relative differences with A as the reference
Checked that the peak times fall within the synthetic time window and that the example values match the text

Verification limits

Did not use a real vehicle, component, or solver model
Did not verify mesh or time convergence or the validity of the physical model
Does not provide judgments on safety, durability, regulatory compliance, or design approval

Site-wide writing and verification principles

Example files to run yourself

Includes code, input data, and instructions. Extract the ZIP and read README.txt first.

Download example ZIP

Example code, filenames, and input keys remain unchanged. Refer to the commands and checking steps in the translated article as well.

References

The explanations and examples were written for this site. See the official sources below for the related behavior and concepts.