Before an FEA plot signs off a change, read the mesh report

Wed, 04 Nov 2026 · 6 min read · Written with an AI assistant

SOLIDWORKS Simulation 2027 counts four kinds of bad element after every mesh. Here is what that report catches, what it cannot tell you, and what else to check before an FEA result moves a design change on.

A stress plot that looks plausible is not evidence that a design change is safe. If the mesh under it contains inverted or badly stretched elements, the colours can be confident and wrong.

That matters now because SOLIDWORKS 2027 is arriving: reseller GoEngineer says it is expected to release in Fall 2026, with a virtual launch event on 4 November and another session on 5 November. Its Simulation module adds an automatic mesh quality report and changes how bolt preload works in nonlinear studies. One caveat up front. Every 2027 feature described here comes from Dassault’s What’s New 2027 document, which is marked PRERELEASE. Treat the details as described before release; the shipping version may differ.

The plot is only as good as the mesh

When a changed part is accepted on the strength of an FEA result, the mesh is part of the evidence. Until now the built-in check was narrow. According to Dassault, the mesh quality tool used to detect only distorted elements with negative Jacobian ratios, and in 2027 also detects elements with high aspect ratios, zero areas or zero volumes.

Those four problems are worth knowing by mechanism:

Negative Jacobian. The element’s mapping has folded inside out. Results there are meaningless, and the solver may not run at all.
High aspect ratio. Long, thin elements represent stress gradients poorly, especially across their short dimension.
Zero area. A degenerate face, usually a symptom of collapsed or sliver geometry.
Zero volume. A flattened element that contributes no stiffness.

Treat a clean report as the floor, not the ceiling. It says the elements are well formed. It says nothing about whether the mesh is fine enough where the stress matters. That still needs the usual discipline: refine locally at the region of interest, re-solve, and see whether the peak value settles. I could not find published numeric limits for aspect ratio or Jacobian in the sources behind this piece, so I will not quote any. Use your company’s own criteria and write them down.

Using the warning and the report properly

There are two mechanisms. First, the Mesh PropertyManager has an option, “Issue warning for distorted elements”, which shows a warning when poor-quality elements are found; you can carry on with the current mesh, or cancel to check for geometry irregularities and change the mesh settings. Second, after meshing, Simulation automatically shows a report counting each of the four element problems, and the check can be re-run by right-clicking Mesh in the study tree and choosing “Check Mesh Quality”. Dassault’s guidance is to refine the mesh where distorted elements are found.

My view: turn the warning on for any study that will back a design decision, and re-run the check after every geometry change, however small. A change request rarely touches the region the original analyst worried about. It touches somewhere else, and that is where the new bad elements appear.

There is a second reason to re-check after upgrading. The Standard, Curvature-based and Blended curvature-based meshers now use the Converged Volume Mesher (CVM) as the default volume filler for solid meshing. If an old model is re-meshed in 2027, assume it is a different mesh from the one that produced the archived result.

A worked example. A machined bracket has a bolt boss next to a fillet. A change request increases the boss diameter for a larger fastener. The new boss edge now stops just short of the fillet tangent line, leaving a narrow sliver face. The engineer re-meshes, the warning fires, and the report counts high aspect ratio elements.

Carrying on would produce a stress plot with a hotspot right at the sliver. It might be real. It is more likely an artefact of stretched elements sitting on an accidental face. The better path is to cancel, inspect the geometry, and remove the sliver at source: either move the boss so it blends cleanly into the fillet or clears it properly. Then apply local mesh control around the boss, re-mesh, and run Check Mesh Quality again. Only once the report is clean, or the remaining flagged elements are shown to be away from the area of interest, is the peak stress worth comparing with the pre-change result. The geometry fix is also a design improvement, which is a useful side effect of taking the warning seriously.

Upgrading old studies without surprises

Bolt preload in nonlinear studies can now be time-based, with two methods: “Gradual” tightens bolts to the preload between set start and end times, and “Immediate” applies the preload before the nonlinear analysis starts. When an existing nonlinear study is opened, bolt connectors use the Immediate method so the original study behaves as before.

Immediate is the right default for most joint checks. Gradual earns its place when the order of events affects the answer: flanges whose contact closes as bolts are tightened, or a study where you want preload established within a defined time window before external load builds up. That is an engineering judgement, not a software setting to flip for curiosity.

If you do switch, record it. A result solved with Gradual preload is not a like-for-like comparison with the archived Immediate result, even if nothing else changed. Put the method and the time window in the change record next to the result it produced.

Results to read before sign-off

The useful question before approving is whether loads go where the design intent says they should. Several 2027 additions make that easier to answer with numbers rather than colours:

Bonded interfaces. For Surface to Surface bonding, you can plot “BN: Bonding Normal stress” and “BT: Bonding Tangential stress”; vectors along the positive normal mean tension, the opposite direction compression. If a bonded joint meant to carry shear shows mostly tension, the model or the design needs another look.
Beam shear. “List Beam Forces” now has columns for shear stress in both local directions, which earlier versions only showed in plots.
Springs. The General Spring connector now works in nonlinear static and dynamic studies, takes up to six stiffness parameters in a local coordinate system, and reports shear, axial force, bending moment and torque via “List Connector Forces”.
Thermal totals. The heat power summary now correctly includes heat transfer between contacting or bonded bodies, with internal flow balanced automatically in whole-model totals. Expect totals from older versions not to match; check the energy balance before comparing.
Summation sensors. “Summation of Selected Entities” adds up quantities over selected entities, replacing one-by-one extraction and manual sums. Total reaction across a bolt pattern is a good candidate.

Also check the interactions themselves. Right-clicking one and choosing “Details” opens an Interaction Details dialog showing everything about it. Note that Simulation Standard, Professional and Premium are bought separately, so confirm which study types your licence covers before planning around a feature.

A pre-release checklist, and what it means for tooling

For any change that leans on FEA, attach the following to the change record, building on the trail described in our earlier ECR, ECO and ECN piece:

Mesh quality report: clean, or each flagged element explained and shown to be away from the area of interest.
Evidence of local refinement at the governing location, with the peak value at each mesh.
Interaction Details reviewed for every contact and bond that carries load.
Bolt preload method and timing, stated explicitly.
Key results listed as numbers: connector forces, beam shear, interface stresses, summed reactions.
A note that the study was re-meshed in 2027 if it was upgraded.

Most of that list is mechanical. Counts, flags, method settings and listed values are exactly what tooling should collect and attach without anyone pasting screenshots into a spreadsheet. What tooling cannot do is decide whether a hotspot is physics or artefact, or whether Gradual preload reflects how the joint is actually assembled. That stays with the engineer, and the record should show who decided.

The 2027 mesh report is a welcome step because it turns a quiet failure into a visible count. It does not make a study correct. Read it, act on it, and put it in the change record alongside the numbers that justified the decision. If you are testing the pre-release features and see behaviour that differs from the documentation, we would like to hear about it; follow the journal for more on simulation in the design loop.