Draft. This manual is new and still growing. If something here does not match what you see in the software, the software is right — tell us and we will fix the page.

Structural, thermal and the other solid families

Structural, thermal, electromagnetics, acoustics, explicit dynamics and shape optimisation are validated by a different set of rules from the fluids families, and they meet the meshers and solvers on a different path. Nothing on this page appears on a fluids case, and nothing on the fluids pages appears here.

Every entry is a Problems entry unless it says otherwise. They land on the step that shows the thing they are about — Mesh, Material, Boundaries, Loads, Restraints, Contacts or Models — so the row's step name tells you where to go.

Back to Troubleshooting.

The two that cost the most time

Both describe a model that is not wrong so much as not determined: there are infinitely many answers, so the matrix is singular. The solver reports it a long way from the cause, which is why the application says it by name first.

"Nothing is restrained. A model with no restraint is free to move, so there is no unique solution — the solver reports this as a pivot error far from its cause."

Structural, against Restraints. A body under load with nothing holding it is not a hard problem, it is an unbounded one — it accelerates.

The harmonic response analysis is the exception and does not raise this: a free body being shaken is a well-posed problem, because the rigid-body modes sit at zero frequency and a sweep is never there.

"Every boundary is a heat flux, so nothing fixes what the temperature actually is — only its gradient. There is no unique solution, and the solver reports this as a pivot error far from its cause. Hold one face at a temperature, or give one a convection condition."

Thermal, against Boundaries. The thermal counterpart, and it fails in the same way. A fixed temperature or a convection condition will pin it; convection counts because tying the surface to an ambient is itself a reference.

If nothing at all has been applied, a different message fires instead: "Nothing is applied to any face. An unstated face is insulated, so this describes a part with no way for heat to enter or leave."

Units typed into the wrong field

Both of these are cases that solve and give you a wrong answer. They are the highest-value entries on this page for exactly that reason.

"The temperature on 'inner' is below absolute zero. Temperatures here are in kelvin — 20 °C is 293.15 K."

Thermal, against Boundaries. Almost always °C typed into a field that wants kelvin. It is named explicitly because -20 looks like an ordinary cold boundary until you notice the unit.

"A relative permittivity below 1 is a material emptier than vacuum, which is almost always an absolute permittivity in F/m typed into a field that wants a ratio. The solver applies ε₀ = 8.854e-12 F/m itself; this number is the multiplier on it."

Electromagnetics, against Material, on an electrostatic analysis. This one converges. On a purely fixed-potential case even the potential field comes out unchanged — it is the energy, the flux and any charge-driven case that come out nonsense.

Meshes

"No mesh. Generate or import one before solving."

Against Mesh. On an acoustic radiation analysis the wording is specific about which mesh: "No mesh. Generate or import the vibrating part before solving."

"'<file>' is not a mesh format this can read. Expected .med, .mail, .msh, .unv or .inp."

Against Mesh, naming the file. Those five extensions are the whole list.

"An Abaqus mesh (.inp) can be solved by CalculiX but not by code_aster. Mesh to .med or .msh to use either."

Against Mesh. Not an error in the file — a property of the case that means only one of the two backends can run it. It is said before a solve starts so it is a choice rather than a failure.

Materials

Each of these fires only when the analysis actually reads the number, so a material with gaps in it is a fine choice until something needs the missing figure.

"Young's modulus must be greater than zero."

Structural, against Material. Its companion is "Poisson's ratio must be between -1 and 0.5. At 0.5 the material is incompressible and the solve cannot converge."

"Thermal conductivity must be greater than zero. Steel is about 50 W/m/K, aluminium 200, and a plastic 0.2."

Thermal, against Material. A transient thermal analysis needs two more: "A transient analysis needs a density: how fast the part heats up is a question about its mass." and "A transient analysis needs a specific heat capacity. Steel is about 460 J/kg/K, aluminium 900, water 4180."

"Electrical conductivity must be greater than zero. Copper is 5.96e7 S/m, stainless about 1.4e6, and a polymer is an insulator rather than a poor conductor — it does not belong in a conduction model at all."

Electromagnetics, against Material, on a current-conduction analysis only.

"The speed of sound must be greater than zero. Air is 343 m/s and water 1481 - and every wavelength and resonance in the answer scales with it."

Acoustics, against Material. Its companion is "The density must be greater than zero. It sets the impedance, which is what turns a vibrating surface into a pressure."

"This material has no viscosity, thermal conductivity, specific heat or specific heat ratio stated. The thermoviscous solver needs all four - pick a material with fluid properties, or switch to the lossless harmonic analysis."

Acoustics, against Material, on the thermoviscous analysis only. Harmonic and radiation never read those four.

Shells and beams

A shell is a surface and a beam is a line. Neither carries the information that makes it a structure, so it has to be stated.

"The shell has no thickness. A sheet body is a surface; the thickness is what makes it a structure."

And "The beam has no cross-section. A wire is a line; the section is what makes it a member."

Against Models, and it names the model type you are on. Three related rules say what else does not cross over: "A temperature field cannot be applied to a shell or beam: the thermal study is solved on a solid mesh and the two do not share nodes.", "A flow load maps onto a solid's faces; it is not built for a shell or beam.", and "Connectors are not built for a shell or beam mesh."

"A beam has no surface for a pressure to act on. Use a force: per metre on the member, or in newtons at a vertex."

Against Loads. The shell version names the group instead: "The pressure on 'edge_1' is on an edge or a vertex. A pressure acts on the shell's face; use a force there instead."

The job failed

These are Log lines, and they also become the run's status on the run pane.

"The solver exited 1 but wrote no result. The log is above."

The message names nothing because there is nothing it can name — the process ended and left no result table to read. The reason is in the Log above it, in the solver's own output.

If the solver's own last words are a pivot or factorisation failure, look at the two determinacy problems at the top of this page — those are the cases the application tries to catch before a solve for exactly this reason. Its siblings are worded the same way for the other backends: "Elmer exited 1 and wrote no result. The log is above.", "The OpenRadioss starter exited 1 and wrote no restart, so the deck did not pass its checks. The log is above.", and so on.