Structural mechanics validation
Each check compares a RunMat result with an independent reference: a closed-form solution, a manufactured solution or a published benchmark. The tests listed under a check run on every change and in the evidence report; receipts record a public study run with its build, commit and mesh. The validation methodology explains how checks are built and how tolerances are set, and Status summarizes every family.
Static
Cantilever bending against beam theory
Problem. A 1 m steel cantilever (E = 200 GPa, ν = 0.3) with a 100 mm × 100 mm section is fully fixed at x = 0 and carries a 1,000 N downward traction resultant on the free end face. It is the cantilever of the linear static and mesh refinement guides, run as a public study on Tet10 meshes at 150, 75, 50 and 25 mm and Tet4 meshes at 150 and 50 mm.
Reference. Euler–Bernoulli beam theory: tip deflection = 0.2000 mm and top-surface bending stress with , 5.40 MPa at x = 100 mm. The support reaction equals the applied 1,000 N. Timoshenko theory adds the shear deflection 0.0016 mm (0.8 %), so a converged solid model lies between the two theories. (J. M. Gere and B. J. Goodno, Mechanics of Materials, 8th ed., Cengage Learning, 2013.)
Compared. Largest displacement on the loaded end face; nodal von Mises stress on the top surface at x = 100 mm, one section depth from the clamp; support reaction. The largest element stress at the clamped edge is recorded to show the singularity of the fully fixed face.
Tolerance. None set; the receipts record each value. Tet10 tip deflection is expected between the Euler–Bernoulli and Timoshenko values.
Result. Run receipts dated 2026-10-29. Tet10 tip deflection converges to 0.2007 mm, 0.34 % above Euler–Bernoulli theory and below the Timoshenko value. The stress at x = 100 mm settles at 5.41 MPa, 0.2 % above beam theory. The largest element stress at the clamped edge does not settle (5.39, 5.45, 5.27, 5.76 MPa), as expected at a singularity. The reaction balances 1,000 N on every mesh. Tet4 locks in bending: −50 % at 150 mm and −19 % at 50 mm.
| Quantity | Mesh | RunMat | Reference | Difference |
|---|---|---|---|---|
| Tip deflection: largest displacement magnitude over the loaded end face | Tet10, 150 mm, 2,727 nodes | 0.200034 mm | 0.2 mm | +0.02 % |
| Tet10, 75 mm, 2,758 nodes | 0.200295 mm | 0.2 mm | +0.15 % | |
| Tet10, 50 mm, 4,679 nodes | 0.200474 mm | 0.2 mm | +0.24 % | |
| Tet10, 25 mm, 32,075 nodes | 0.200678 mm | 0.2 mm | +0.34 % | |
| Tet4, 150 mm, 431 nodes | 0.10069 mm | 0.2 mm | −49.66 % | |
| Tet4, 50 mm, 729 nodes | 0.162496 mm | 0.2 mm | −18.75 % | |
| Nodal von Mises stress on the top surface at x = 100 mm, y = 50 mm | Tet10, 150 mm, 2,727 nodes | 5.43303 MPa | 5.4 MPa | +0.61 % |
| Tet10, 75 mm, 2,758 nodes | 5.42193 MPa | 5.4 MPa | +0.41 % | |
| Tet10, 50 mm, 4,679 nodes | 5.41287 MPa | 5.4 MPa | +0.24 % | |
| Tet10, 25 mm, 32,075 nodes | 5.41221 MPa | 5.4 MPa | +0.23 % | |
| Tet4, 150 mm, 431 nodes | 1.92644 MPa | 5.4 MPa | −64.33 % | |
| Tet4, 50 mm, 729 nodes | 3.51563 MPa | 5.4 MPa | −34.90 % | |
| Largest element von Mises stress in the beam (at the clamped edge; does not converge) | Tet10, 150 mm, 2,727 nodes | 5.39247 MPa | none | |
| Tet10, 75 mm, 2,758 nodes | 5.45246 MPa | none | ||
| Tet10, 50 mm, 4,679 nodes | 5.26615 MPa | none | ||
| Tet10, 25 mm, 32,075 nodes | 5.76473 MPa | none | ||
| Tet4, 150 mm, 431 nodes | 24.80245 MPa | none | ||
| Tet4, 50 mm, 729 nodes | 9.19027 MPa | none | ||
| Resultant reaction force at the fixed end, z component | Tet10, 150 mm, 2,727 nodes | 1000 N | 1000 N | +0.00 % |
| Tet10, 75 mm, 2,758 nodes | 1000 N | 1000 N | +0.00 % | |
| Tet10, 50 mm, 4,679 nodes | 1000 N | 1000 N | +0.00 % | |
| Tet10, 25 mm, 32,075 nodes | 1000 N | 1000 N | +0.00 % | |
| Tet4, 150 mm, 431 nodes | 1000 N | 1000 N | +0.00 % | |
| Tet4, 50 mm, 729 nodes | 1000 N | 1000 N | +0.00 % |
Tet4 patch test under an affine displacement field
Problem. Connected patches of linear tetrahedra (Tet4) under an affine displacement field, the standard patch test for solid elements.
Reference. The exact uniform strain and stress of the affine field (strain 0.1, 0.2 and 0.3 in x, y and z; stress 32, 40 and 48 in the fixture's units), its strain energy, and nodal forces equal to the independent boundary tractions.
Compared. Recovered strain and stress at elements and integration points, strain energy, internal nodal forces.
Tolerance. Absolute 1e-12 on strain and 1e-11 on stress and energy; relative 1e-12 on nodal forces and energy.
Result. Passed (2 tests) in the evidence run of 2026-11-08.
Tests: connected_tet4_affine_patch_balances_independent_boundary_tractions, connected_tet4_affine_patch_recovers_analytic_strain_stress_and_energy.
Tet10 nodal stress recovery under a linear stress field
Problem. Two quadratic tetrahedra (Tet10) sharing a face carry a linear stress field; the two elements belong to different regions.
Reference. The linear stress field evaluated exactly at every corner and midside node, with no averaging across the region interface node.
Compared. Recovered nodal stress components.
Tolerance. Absolute error at most 1e-12 × (1 + |σ|) per component.
Result. Passed (1 test) in the evidence run of 2026-11-08.
Tests: tet10_nodal_stress_is_exact_for_a_linear_field_and_keeps_regions_apart.
Shell membrane patch under biaxial plane stress
Problem. A one-layer shell patch in uniform biaxial membrane stress.
Reference. Plane-stress membrane resultants , , , no bending or shear resultants, and von Mises stress .
Compared. Membrane force resultants, bending and shear resultants, von Mises stress.
Tolerance. Relative 1e-11 on resultants; bending and shear resultants below 1e-12.
Result. Passed (1 test) in the evidence run of 2026-11-08.
Tests: one_layer_membrane_patch_matches_the_independent_plane_stress_reference.
Simply supported plate under sinusoidal load
Problem. A simply supported plate meshed with triangular shell elements carries a sinusoidal transverse pressure; the mesh is refined.
Reference. Navier's series solution for the center deflection of a simply supported plate. (S. P. Timoshenko and S. Woinowsky-Krieger, Theory of Plates and Shells, 2nd ed., McGraw-Hill, 1959.)
Compared. Center transverse displacement.
Tolerance. The center displacement approaches the Navier value with refinement; the finest-mesh error is below 8 %.
Result. Passed (1 test) in the evidence run of 2026-11-08.
Tests: simply_supported_sinusoidal_plate_converges_to_the_navier_solution.
Modal
Fundamental frequency of an axial rod
Problem. A solid slab rod, fixed at one end and free at the other, in axial vibration on Tet4 meshes (2, 4 and 8 cells) and Tet10 meshes (1, 2 and 4 cells). The retained one-mode response is also compared with a direct harmonic solve.
Reference. Fixed–free rod fundamental frequency with .
Compared. First natural frequency; modal-truncation response against the direct harmonic response.
Tolerance. Frequency error falls by more than 3× (Tet4) or 6× (Tet10) per refinement; finest-mesh error below 2e-3 (Tet4) and 2e-4 (Tet10).
Result. Passed (2 tests) in the evidence run of 2026-11-08.
Tests: solid_rod_modal_frequency_and_retained_response_have_independent_evidence, solid_rod_tet10_modal_frequency_and_retained_response_have_independent_evidence.
Harmonic
Axial rod under harmonic end traction
Problem. A Tet4 slab rod, fixed at one end, carries a harmonic axial traction on the free end at three frequencies on both sides of the first resonance; 2, 4 and 8 cells.
Reference. The continuum complex displacement of the driven rod and the support reaction .
Compared. Complex displacement field and support reaction; mean input power of the undamped response.
Tolerance. Displacement and reaction errors fall by more than 2.5× per refinement and are below 2e-2 on the finest mesh; mean input power below 1e-9.
Result. Passed (1 test) in the evidence run of 2026-11-08.
Tests: solid_rod_harmonic_face_traction_converges_to_independent_continuum.
Transient
Undamped free vibration under average-acceleration integration
Problem. An undamped oscillator in free vibration, integrated with the Newmark average-acceleration rule, both directly and in modal coordinates, with the time step halved twice.
Reference. The exact harmonic free response.
Compared. Displacement and velocity at the final time.
Tolerance. Error ratio above 3.8 per halving of the time step (second order) on both paths; finest direct Newmark error below 2e-3; modal relative residuals at most 2e-15.
Result. Passed (2 tests) in the evidence run of 2026-11-08.
Tests: average_acceleration_modal_integration_is_second_order, undamped_free_vibration_converges_at_second_order.
Rod under affine forcing in time
Problem. A Tet4 and a Tet10 rod under a physical forcing that is affine in space, integrated with 20, 40 and 80 time intervals, with and without damping.
Reference. The independent continuum response of the rod and its work–energy balance.
Compared. Displacement, velocity and acceleration fields; work–energy balance.
Tolerance. Spatially exact quantities below 1e-12; time-discretization errors fall by more than 3× per halving and are below 2e-4 at 80 intervals.
Result. Passed (1 test) in the evidence run of 2026-11-08.
Tests: affine_physical_forcing_qualifies_tet4_tet10_temporal_fields_and_work_energy.
Buckling
Fixed–free column against Euler's buckling load
Problem. A solid column, 1 m long with a 100 × 50 mm section, fixed at the base and free at the top, carries a 250 kN axial preload; Tet10 mesh.
Reference. Euler's critical load for a fixed–free column, . (S. P. Timoshenko and J. M. Gere, Theory of Elastic Stability, 2nd ed., McGraw-Hill, 1961.)
Compared. First buckling load factor; eigenpair residuals.
Tolerance. First load factor within 1 % of /250 kN; eigenpairs converged to a relative tolerance of 1e-8.
Result. Passed (1 test) in the evidence run of 2026-11-08.
Tests: fixed_free_column_buckles_at_the_euler_load_with_converged_eigenpairs.
Response spectrum
Modal combination rules
Problem. Two signed modal responses with set frequencies and damping are combined by SRSS, absolute sum and CQC.
Reference. The SRSS, absolute-sum and CQC formulas evaluated independently, including an independently computed CQC correlation coefficient. (A. Der Kiureghian, "A response spectrum method for random vibration analysis of MDF systems," Earthquake Engineering & Structural Dynamics 9(5), 419–435, 1981.)
Compared. Combined peak response.
Tolerance. Relative 2e-13.
Result. Passed (1 test) in the evidence run of 2026-11-08.
Tests: modal_combination_matches_independent_quadratic_form_references.
Nonlinear (plasticity)
Uniaxial bar beyond yield
Problem. A bar in uniaxial tension loaded past yield, with J2 plasticity and linear hardening, solved by Newton iteration on Tet4 and Tet10 meshes and restarted cold from its persisted state.
Reference. The closed-form uniaxial elastoplastic stress at the final load (5.6 in the fixture's units). (J. C. Simo and T. J. R. Hughes, Computational Inelasticity, Springer, 1998.)
Compared. Axial stress at every integration point; restart state.
Tolerance. Axial stress within 2e-10 of the closed form.
Result. Passed (1 test) in the evidence run of 2026-11-08.
Tests: physical_newton_matches_independent_uniaxial_plastic_bar_and_cold_restart_both_orders.
Manufactured fully yielded cube
Problem. A cube under body force and surface traction yields completely; J2 plasticity, Tet4 and Tet10 meshes, applied in one increment and in eight.
Reference. A manufactured polynomial J2 solution: quadratic terminal strain with an exact radial return on its proportional path.
Compared. Displacement, stress, hardening variable, plastic history, energies and reaction.
Tolerance. Tet10 in one increment: displacement, stress, hardening variable and energy errors below 1e-7 under the Keast15, Duffy5 and Duffy8 rules. In eight increments the yield front crosses elements mid-step; that path error falls at order 1.5 or better with refinement. Tet4: every quantity refines.
Result. Passed (3 tests) in the evidence run of 2026-11-08.
Tests: fresh_tet10_integration_trajectories_meet_original_eight_step_exactness, manufactured_fully_yielded_cube_refines_and_checks_load_step_sensitivity, manufactured_fully_yielded_tet10_cube_is_exact_on_both_load_schedules.