RunMat
  • Pricing
RunMat
GitHub
GitHub
Sign inTry in browser
DesktopRuntimePhysicsCADAdministrationAgent

Get RunMat updates

New releases, guides and examples.

Product+
  • Runtime
  • GPU
  • Parallel
  • Compile
  • Interop and MEX
  • Simulation
  • MCAD
  • ECAD
  • Enterprise
  • AI integrations
  • Pricing
  • Download
Math+
  • Signal processing
  • Linear algebra
  • Statistics
  • Curve fitting
  • Differential equations
  • Control systems
  • Optimization
  • Finance
  • Data and tables
  • Plotting
Physics+
  • Simulation overview
  • Structural
  • Thermal
  • Electromagnetic
  • Joule heating
  • Thermal stress
  • Acoustics
  • Fluid flow
  • Conjugate heat transfer
  • Fluid–structure
  • What is FEA
Industries+
  • Aerospace
  • Automotive and transportation
  • Defense
  • Electronics and semiconductors
  • Energy
  • Industrial machinery
  • Life sciences
  • Quantitative finance
Learn+
  • Guides
  • Docs
  • Function reference
  • Language reference
  • Migration guide
  • Changelog
  • MATLAB Online
  • MATLAB AI agent
  • For students
Compare+
  • RunMat vs MATLAB
  • RunMat vs Ansys
  • RunMat vs SOLIDWORKS
  • RunMat vs Altium
  • RunMat vs Cadence
  • RunMat vs COMSOL
  • RunMat vs Python
  • RunMat vs Julia
  • RunMat vs Physics AI
  • RunMat vs Octave
  • Benchmarks
  • MATLAB alternatives
  • All comparisons
Company+
  • About
  • Blog
  • Contact
  • Source code↗
  • License
  • Privacy
  • Terms

Product

  • Runtime
  • GPU
  • Parallel
  • Compile
  • Interop and MEX
  • Simulation
  • MCAD
  • ECAD
  • Enterprise
  • AI integrations
  • Pricing
  • Download

Math

  • Signal processing
  • Linear algebra
  • Statistics
  • Curve fitting
  • Differential equations
  • Control systems
  • Optimization
  • Finance
  • Data and tables
  • Plotting

Physics

  • Simulation overview
  • Structural
  • Thermal
  • Electromagnetic
  • Joule heating
  • Thermal stress
  • Acoustics
  • Fluid flow
  • Conjugate heat transfer
  • Fluid–structure
  • What is FEA

Industries

  • Aerospace
  • Automotive and transportation
  • Defense
  • Electronics and semiconductors
  • Energy
  • Industrial machinery
  • Life sciences
  • Quantitative finance

Learn

  • Guides
  • Docs
  • Function reference
  • Language reference
  • Migration guide
  • Changelog
  • MATLAB Online
  • MATLAB AI agent
  • For students

Compare

  • RunMat vs MATLAB
  • RunMat vs Ansys
  • RunMat vs SOLIDWORKS
  • RunMat vs Altium
  • RunMat vs Cadence
  • RunMat vs COMSOL
  • RunMat vs Python
  • RunMat vs Julia
  • RunMat vs Physics AI
  • RunMat vs Octave
  • Benchmarks
  • MATLAB alternatives
  • All comparisons

Company

  • About
  • Blog
  • Contact
  • Source code↗
  • License
  • Privacy
  • Terms
GitHubXLinkedIn

© Dystr, Inc. · RunMat v0.7.1, Nov 17, 2026

RunMat® is a registered trademark of Dystr, Inc. MATLAB® is a registered trademark of The MathWorks, Inc. RunMat is not affiliated with, endorsed by, or sponsored by The MathWorks, Inc.

/
    • Grammar
    • Declarations
    • Procedures
    • Types, units and expressions
    • Files, modules and imports
    • Studies, analyses and state
    • Outputs, fields and observations
    • Operations and output queries
    • Operation reference
    • RunMat function integration
    • Diagnostics and tooling
    • Geometry features and AP242
    • Geometry, models and assignments
    • Constraints, loads and interfaces
    • Fluid materials
Docs>Physics>Language reference

Procedures

An analysis step is an analysis <name> { … } block in a study. It holds exactly one procedure block, <family> <procedure> { <fields> }, plus optional retain, accuracy, load_cases and initial_conditions members. Steps run in dependency order; after analysis.<step> orders a step after another, and a state binding (basis = analysis.modes.state.basis) makes one step consume another's result.

analysis modes {
    structural modal { modes = 12 }
    retain basis
}

analysis response after analysis.modes {
    structural modal_harmonic {
        basis = analysis.modes.state.basis
        frequencies = [50 Hz, 100 Hz, 150 Hz]
    }
}

Families and procedures

The table below is generated from the capability ledger, the same source as the compiler's table of family and procedure pairs. runmat check applies that table, so a combination the family does not run is reported at its line before anything is meshed.

FamilyRuns in this releasePlanned
Structural mechanicsstatic, modal, prestressed modal, buckling, transient, harmonic, prestressed harmonic, nonlinear, transient (modal superposition), harmonic (modal superposition), prestressed harmonic (modal superposition), response spectrum (modal superposition), random response (modal superposition)none
Heat transferstatic, transientnone
Electromagneticsstatic, harmonic, line parameterstransient, nonlinear
Acousticsharmonic, modal, transientstatic, nonlinear
Fluid flowstatic, transientharmonic, nonlinear
Thermomechanicalstatic, nonlineartransient, harmonic
Electrothermalstatic, transientharmonic, nonlinear
Conjugate heat transferstatic, transientharmonic, nonlinear
Fluid–structure interactionstatic, transientharmonic, nonlinear

Modal superposition reuses a retained modal basis; every other entry solves the full system directly.

What each procedure solves

With K\mathbf KK, M\mathbf MM and Cd\mathbf C_dCd​ the assembled stiffness, mass and damping matrices and f\mathbf ff the load vector:

ProcedureEquationIndependent domain
static (steady)Ku=f\mathbf K \mathbf u = \mathbf fKu=fnone
modal(K−ωi2M) ϕi=0(\mathbf K - \omega_i^2 \mathbf M)\,\boldsymbol\phi_i = \mathbf 0(K−ωi2​M)ϕi​=0, for the lowest modes pairsmode count
transientMu¨+Cdu˙+Ku=f(t)\mathbf M \ddot{\mathbf u} + \mathbf C_d \dot{\mathbf u} + \mathbf K \mathbf u = \mathbf f(t)Mu¨+Cd​u˙+Ku=f(t) (first order, CT˙+KT=q(t)\mathbf C \dot{\mathbf T} + \mathbf K \mathbf T = \mathbf q(t)CT˙+KT=q(t), for heat transfer)time
harmonic(K+iωCd−ω2M) u^=f^(ω)(\mathbf K + i\omega \mathbf C_d - \omega^2 \mathbf M)\,\hat{\mathbf u} = \hat{\mathbf f}(\omega)(K+iωCd​−ω2M)u^=f^(ω)frequency list
buckling(K+λiKσ) ϕi=0(\mathbf K + \lambda_i \mathbf K_\sigma)\,\boldsymbol\phi_i = \mathbf 0(K+λi​Kσ​)ϕi​=0, with Kσ\mathbf K_\sigmaKσ​ the geometric stiffness of a preloadmode count
nonlinearr(u,λ)=fint(u)−λf=0\mathbf r(\mathbf u, \lambda) = \mathbf f_\text{int}(\mathbf u) - \lambda \mathbf f = \mathbf 0r(u,λ)=fint​(u)−λf=0, solved by Newton iteration over the load factor λ\lambdaλload factor
modal-reduced formsthe same equations projected onto a retained modal basis, u≈Φq\mathbf u \approx \boldsymbol\Phi \mathbf qu≈Φqas above

Each family's physics reference states its weak form, element matrices, solver and recovery.

Structural procedures

Procedure blockFieldsPublishes
structural staticnoneprimary (displacement state)
structural preloadnone; the step's loads are the preloadgeometric_stiffness
structural modalmodes (positive integer); optional residual_vectors (one to three global directions, [[0, 0, 1]]); optional solver = auto (default), factorization or iterativemodal_basis
structural transientstart_time, end_time, time_stepprimary
structural modal_transientbasis, start_time, end_time, time_stepprimary
structural harmonicfrequenciesprimary
structural modal_harmonicbasis, frequencies; optional static_correction = applied_loads or none (default)primary
structural prestressed_modalprestress, modes; optional residual_vectors and solvermodal_basis
structural bucklingprestress, modesbuckling modes
structural prestressed_harmonicprestress, frequencies; optional basis (modal reduction instead of a direct solve) and, with basis, static_correctionprimary
structural modal_response_spectrumbasis, spectra = [model.response_spectrum.<name>, …]primary
structural random_responsebasis, excitation = model.random_excitation.<name>, modal_damping_ratio (number), minimum_cumulative_mass_ratio (number)primary
structural nonlinearstart_load_factor, end_load_factor; optional kinematics = small_strain (default) or large_deformationprimary

Modal eigensolver

solver chooses how a modal extraction computes its modes:

  • auto (the default): the resource plan uses shift-invert block Lanczos on a sparse factorization when that plan fits the memory the run is admitted to use, and otherwise block LOBPCG with a multigrid preconditioner when that plan fits.
  • factorization: shift-invert block Lanczos only.
  • iterative: block LOBPCG only. It holds no factorization, so it fits models whose factor does not.

Both methods prove that no mode below the highest reported one was missed (a Sturm inertia count) and qualify each mode on the same accuracy criterion. When the plan for a required method does not fit the admitted memory, the run is refused before solving, and the error names the bytes the study needs and the bytes the budget offers. RunMat does not switch to the other method.

analysis modes {
    structural modal {
        modes = 6
        solver = iterative
    }
}

Modal truncation and residual vectors

A modal-reduced step solves in the space its basis spans: the retained modes, then any residual vectors, then any static correction. RunMat accepts the step when the reduced equations meet the step's relative accuracy tolerance. It then evaluates the full-order equilibrium of the expanded response and publishes the largest relative imbalance as the modal truncation measure (modal_truncation_balance_error in fea.results, with the frequency where it occurs). That measure reports how much response the reduced space leaves out. It is published for every modal-reduced sweep and is not compared with the algebraic tolerance. A direct solve is accepted on its full-order balance and reports it as max_frequency_balance_error.

Two enrichments shrink the truncation measure:

  • residual_vectors = [[0, 0, 1], …] on structural modal or structural prestressed_modal adds, for each global direction, the static response to a unit acceleration in that direction with the retained modes removed. Use it for base excitation and inertia loading.
  • static_correction = applied_loads on a modal-reduced frequency step adds the static response to the step's applied load (real and imaginary parts at the sweep's reference frequency), with the retained shapes removed. The reduced space then contains the exact quasi-static response, so the truncation measure falls sharply below the first retained mode. Use it for concentrated or surface loads.
analysis modes {
    structural modal { modes = 8 }
    retain basis
}

analysis response after analysis.modes {
    structural modal_harmonic {
        basis = analysis.modes.state.basis
        frequencies = [20, 40, 60] Hz
        static_correction = applied_loads
    }
}

structural nonlinear runs as a study's single step, and inside a procedure graph (after a preload, before a modal step) Not yet releasedStatus.

Thermal procedures

Procedure blockFieldsPublishes
thermal steadyoptional heat_source = analysis.<em-step>.state.<s>primary (temperature state)
thermal transientstart_time, end_time, time_step; optional heat_sourceprimary

heat_source heats the thermal step with the Joule loss density of an upstream electromagnetic harmonic step (induction heating).

Electromagnetic procedures

Procedure blockFields
electromagnetic staticnone; the model's electromagnetic { equation = … } selects electrostatics, DC conduction or magnetostatics
electromagnetic harmonicfrequencies
electromagnetic transient Not yet releasedStatusstart_time, end_time, time_step
electromagnetic nonlinear Not yet releasedStatusstart_load_factor, end_load_factor
electromagnetic line_parametersnone; the cross_section geometry carries the stackup and traces (trace impedance)

Acoustic procedures

Procedure blockFields
acoustic harmonicfrequencies
acoustic modalmodes
acoustic transientstart_time, end_time, time_step
acoustic static Not yet releasedStatusnone
acoustic nonlinear Not yet releasedStatusstart_load_factor, end_load_factor

Fluid procedures

Procedure blockFields
fluid steadynone
fluid transientstart_time, end_time, time_step
fluid harmonic Not yet releasedStatusfrequencies
fluid nonlinear Not yet releasedStatusstart_load_factor, end_load_factor

Coupled procedures

Procedure blockFields
thermomechanical preloadtemperature = analysis.<thermal-step>.state.<s>; publishes geometric_stiffness for prestressed modal, buckling and harmonic steps
thermomechanical steadynone
thermomechanical nonlinearstart_load_factor, end_load_factor
thermomechanical transient Not yet releasedStatusstart_time, end_time, time_step
thermomechanical harmonic Not yet releasedStatusfrequencies
electrothermal steady, electrothermal transientnone; time fields for transient
electrothermal harmonic Not yet releasedStatusfrequencies
electrothermal nonlinear Not yet releasedStatusstart_load_factor, end_load_factor
conjugate_heat_transfer steady, conjugate_heat_transfer transientnone; time fields for transient
conjugate_heat_transfer harmonic Not yet releasedStatusfrequencies
conjugate_heat_transfer nonlinear Not yet releasedStatusstart_load_factor, end_load_factor
fluid_structure_interaction steady, fluid_structure_interaction transientnone; time fields for transient
fluid_structure_interaction harmonic Not yet releasedStatusfrequencies
fluid_structure_interaction nonlinear Not yet releasedStatusstart_load_factor, end_load_factor

static and steady are the same procedure; use whichever reads naturally for the family.

Domains

No procedure infers its domain from a signal, an output or another step; each states it:

FieldTypeRule
start_time, end_timetime0≤tstart<tend0 \le t_\text{start} < t_\text{end}0≤tstart​<tend​.
time_steptimePositive, and divides tend−tstartt_\text{end} - t_\text{start}tend​−tstart​ exactly.
frequencieslist of frequenciesFinite, nonnegative, nonempty and strictly increasing, such as [10 Hz, 20 Hz, 50 Hz].
modespositive integerNumber of eigenpairs, from the lowest.
start_load_factor, end_load_factornumbersFinite and increasing; the load is scaled by the factor.

State flow

Each step publishes typed states: a displacement or temperature state (primary), a modal basis (modal_basis), a preload's geometric stiffness (geometric_stiffness). retain <name> publishes a step's state under a chosen name, so retain basis after structural modal publishes analysis.<step>.state.basis.

A consuming step names the state it needs:

analysis preload { structural preload {} }

analysis modes after analysis.preload {
    structural prestressed_modal {
        prestress = analysis.preload.state.geometric_stiffness
        modes = 6
    }
    retain basis
}

analysis response after analysis.modes {
    structural prestressed_harmonic {
        prestress = analysis.preload.state.geometric_stiffness
        basis = analysis.modes.state.basis
        frequencies = [60 Hz]
    }
}

The checker verifies each binding: the producer exists, the state kind matches, the producer runs on the same mesh and solution space, and the state is complete and persisted. after orders steps; it does not replace a state binding. Every state binding adds its own dependency, so a step that consumes two producers waits for both.

Members of every analysis step

MemberFormMeaning
retain <name>statementPublishes the step's state under <name>.
load_cases = [load_case.<a>, …]listThe load cases this step applies. [] applies none; omitting it applies the model's load cases. Supports and prescribed values always apply.
initial_conditions = [model.initial.<a>, …]listInitial state for a step that starts from one, such as a transient step. See initial conditions.
accuracy { algebraic_measure = <measure>; algebraic_relative_tolerance = <number> }blockThe step's mandatory algebraic accuracy check. Both members are optional, and at least one is required. algebraic_relative_tolerance is strictly between 0 and 1. It sets solver stopping and the check on the computed solution; it does not bound discretization error. See algebraic accuracy.
analysis heat_up {
    thermal transient {
        start_time = 0 s
        end_time = 60 s
        time_step = 0.5 s
    }
    load_cases = [load_case.heater_on]
    initial_conditions = [model.initial.ambient]
    accuracy { algebraic_relative_tolerance = 1e-8 }
}

Algebraic accuracy

Every analysis step checks that its computed solution satisfies the assembled equations. algebraic_measure selects the measure the check uses:

  • backward_error: the normwise backward error in the ∞-norm. For a linear system it is ‖b − A x‖ / (‖A‖ ‖x‖ + ‖b‖). For a mode with eigenvalue λ and shape x it is ‖K x − λ M x‖ / ((‖K‖ + |λ| ‖M‖) ‖x‖). It is the smallest relative change to the operators (and load) for which the computed answer is exact. It does not grow with mesh refinement or conditioning, so one tolerance means the same thing on a coarse and a fine model. The default tolerance is 1e-12. Double-precision direct solvers and converged iterative solvers reach about 1e-16 in this measure, and evaluating A x itself is uncertain by up to a few 1e-14 for quadratic solid elements.
  • relative_residual: the Euclidean residual relative to the terms that form it, ‖b − A x‖ / ‖b‖, or ‖K x − λ M x‖ / (‖K x‖ + |λ| ‖M x‖) for a mode. The default tolerance is 1e-8. In double precision this measure cannot go below about 1e-16 × the condition number, so on a fine mesh it can reject a correct solution.

Structural modal and prestressed modal steps use backward_error by default. Every mode publishes both numbers: modal.backward_error and modal.residual_norm (the relative residual). The step's measure decides which one is checked. The other procedures check relative_residual in this release, and setting algebraic_measure = backward_error on them is an error.

ProcedureAlgebraic check in this releasePublished
structural modal, structural prestressed_modalbackward error of each mode, 1e-12modal.backward_error and modal.residual_norm per mode; diagnostics FEA_MODAL_ACCURACY and FEA_MODAL_CONVERGENCE
structural static, structural harmonic, structural transient, structural prestressed_harmonicrelative residual of the solved equations, 1e-8structural.residual_norm
structural nonlinearrelative residual of every Newton correction solve, 1e-8the largest correction residual in the numerical receipt
thermal steady, transient and nonlinear stepsrelative residual of every accepted linear solve, 1e-8the largest linear residual in the numerical receipt
electrical static stepsrelative residual of the lifted linear system, 1e-8linear_relative_residual in the numerical receipt
magnetic static stepsrelative physical and gauge residuals, 1e-8the numerical receipt
structural bucklingrelative residual of each modethe buckling modes' backward errors are computed but not checked

The engineering checks are separate from this table and keep their own criteria: mesh convergence, modal mass capture, energy balance, Newton force balance and the fluid scaled residuals.

The backward-error check for every algebraic solve is Not yet releasedStatus. Static, thermal, electrostatic, magnetostatic and harmonic solves, each transient step and each Newton linearization will check η = ‖b − A x‖ / (‖A‖ ‖x‖ + ‖b‖) at 1e-12 by default and publish it beside the relative residual. algebraic_measure = relative_residual will keep the current check.

Setting algebraic_measure resets the tolerance to that measure's default, and algebraic_relative_tolerance then replaces it:

analysis modes {
    structural modal { modes = 6 }
    accuracy {
        algebraic_measure = relative_residual
        algebraic_relative_tolerance = 1e-6
    }
}

Response spectrum and random response

    analysis modes {
        structural modal { modes = 30 }
        retain basis
    }
    analysis response_spectrum after analysis.modes {
        structural modal_response_spectrum {
            basis = analysis.modes.state.basis
            spectra = [model.response_spectrum.qualification_x]
        }
    }

The spectrum is a model declaration (response_spectrum <name> { … }), and a random excitation a random_excitation <name> { … } power spectral density; both are referenced from the step. The full studies are modal-response-spectrum.fea and modal-random-response.fea.

On this page
  • Families and procedures
  • What each procedure solves
  • Structural procedures
  • Modal eigensolver
  • Modal truncation and residual vectors
  • Thermal procedures
  • Electromagnetic procedures
  • Acoustic procedures
  • Fluid procedures
  • Coupled procedures
  • Domains
  • State flow
  • Members of every analysis step
  • Algebraic accuracy
  • Response spectrum and random response