Thermal loads
Thermal loads enter the heat equation as volume and boundary integrals. Convection adds both a boundary operator and source:
The typed source checker validates the declarations on this page. Supported standalone steady
thermal studies can run through runmat run on a native host; the study's formulation,
boundaries and resource plan must pass admission. Authoring a boundary does not by itself make
its execution available. In particular, surface radiation remains explicitly rejected by the
canonical thermal procedure, as described below.
| Kind | Quantity/fields | Use |
|---|---|---|
heat_generation | Volumetric source in a solid region or explicit whole-model excitation | |
heat_flux | Signal-driven surface flux | |
temperature_field | K | Uniform imposed temperature input on a solid region |
ambient_temperature | K | Ambient-temperature excitation for a compatible formulation |
thermal.heat_flux | target, value | Constant natural surface flux |
thermal.convection | target, ambient_temperature, coefficient | Robin boundary |
thermal.radiation | target, ambient_temperature, emissivity | Authored surface-to-ambient gray radiation; thermal procedure execution remains rejected |
heat_generation targets a solid or, in explicit excitation form, whole_model.
temperature_field targets a solid. Heat flux, ambient temperature, and the natural boundary
blocks target faces. Every implemented thermal quantity here is scalar, so normal orientation—not
an authored direction vector—sets the sign of a surface flux.
load chip_loss {
heat_generation on geometry.module.region.die = 8000000 W/m^3
}
load cooled_surface {
thermal.convection {
target = geometry.module.region.finned_faces
ambient_temperature = 298.15 K
coefficient = 35 W/m^2/K
}
}Surface-to-ambient radiation admission
load radiating_surface {
thermal.radiation {
target = geometry.module.region.finned_faces
ambient_temperature = 298.15 K
emissivity = 0.8
}
}The authored contract uses nonnegative Kelvin and dimensionless emissivity in [0, 1].
The CPU face-form library integrates outward loss
epsilon * sigma * (T^4 - T_ambient^4) and its exact consistent tangent. This describes
exchange with a prescribed ambient, not enclosure/view-factor or participating-medium radiation.
The canonical thermal procedure currently rejects this boundary explicitly: nonlinear
steady/transient closure, resource admission, whole-solve heat accounting and execution/restart
qualification are not yet implemented. Authorability and local face evidence do not admit a solve.
See radiation qualification.
How do I apply volumetric heat generation?
load winding_loss {
heat_generation on geometry.motor.region.winding = 3500000 W/m^3
}This targets a solid region and preserves source intensity. Integrated power is ; compare it with the expected electrical, chemical, or mechanical loss. For temperature-dependent or spatially nonuniform loss, use a compatible signal/function or coupled physics state rather than one averaged constant.
How do I apply electronic, battery, or chemical heat generation?
Convert the governing loss model to a volumetric source only when its spatial support is known. For
uniform loss in a cell or die, use heat_generation on that exact solid region. For state-dependent
loss, generate the source from a typed RunMat function or transfer it from the electrical/chemical
procedure so units, parameter values, and integrated power remain in evidence.
How do I apply heat to a thin layer or surface heater?
Use surface heat_flux when thickness is intentionally collapsed. Use heat_generation in the
actual solid when through-thickness gradients or layer heat capacity matter. Converting a layer
source to surface flux requires with modeled thickness ; record that reduction.
How do I split total power among several components?
Create one region-specific generation load per component. If fractions are known, use and verify . Do not apply the full assembly power to each solid or divide by the total bounding-box volume.
How do I apply a heat flux to a surface?
The direct form creates a constant excitation:
load laser_flux {
heat_flux on geometry.part.region.illuminated_face = 25000 W/m^2
}The explicit natural-boundary form is:
load imposed_flux {
thermal.heat_flux {
target = geometry.part.region.hot_face
value = 25000 W/m^2
}
}Both carry power per area. Check the outward-normal sign convention and integrated power over the resolved area. A total wattage must be converted through known area or represented by a resultant- preserving contract; it is not numerically equal to heat flux.
How do I apply a known total heater power to a face?
For uniform distribution over resolved area , set and verify the integrated boundary power in evidence. If the heater footprint or spatial distribution matters, define that exact CAD region and use its field. A constant flux on a changing/deformed area needs an explicit reference- versus-current configuration convention.
How do I model a laser or moving heat source?
A constant heat_flux models a stationary uniform spot only. A moving Gaussian source needs beam
power, absorptivity, radius, path, coordinate frame, and time dependence in a typed spatial function.
The current .fea signal domain is time, frequency, study parameter, or constant; it does not expose
the coordinates of each target point to a source function. A moving spatial source is therefore not
first class. Do not approximate it by turning an entire face on and off unless that is the intended
reduced model and its deposited-power integral is verified.
Why is my heat-flux sign reversed?
Flux sign follows the resolved outward normal and formulation convention. Preview the CAD face, check its normal, and compare integrated boundary power with the expected input. Reversing a face selector or negating a value are not interchangeable if several bodies share an interface.
How do I prescribe a temperature field as a load?
load operating_temperature {
temperature_field on geometry.bracket.region.body = 373.15 K
}temperature_field is useful when temperature drives thermomechanical response or supplies an
initial/known field without solving heat transfer in that step. It targets solver nodes/regions and
conflicting overlapping fields are rejected. Use thermal.temperature when the thermal equation is
being solved and the temperature is an essential boundary condition.
How do I provide ambient temperature as a varying excitation?
signal ambient_cycle : temperature over time {
samples = [
{ at = 0 s, value = 293.15 K },
{ at = 600 s, value = 323.15 K },
]
interpolation = linear
extrapolation = reject
}
excitation chamber_temperature {
ambient_temperature {
target = geometry.part.region.exterior_faces
signal = signal.ambient_cycle
}
}Ambient temperature needs a compatible convection/radiation formulation to transfer heat. It is not by itself the same as prescribing the part surface to that temperature.
How do I model convection to air or coolant?
load forced_convection {
thermal.convection {
target = geometry.part.region.cooled_faces
ambient_temperature = 298.15 K
coefficient = 80 W/m^2/K
}
}This applies with constant and . Split regions when different flow zones need different coefficients. Do not use convection as an undocumented surrogate for radiation or contact resistance.
How do I choose a convection coefficient?
Use a correlation, test result, or coupled flow solution valid for the geometry, fluid, regime, and temperature range. Record whether is local or area-averaged. A single coefficient on all faces can hide stagnant zones and fins with very different flow. Run a sensitivity range when is an uncertain engineering input.
How do I model natural versus forced convection?
The thermal.convection syntax is the same once and are known; their derivation is
different. For natural convection, depends on orientation, gravity, geometry, and temperature
difference. For forced convection, it depends on flow speed and regime. Solve conjugate heat
transfer when the fluid field and local coefficient must be predicted rather than prescribed.
How do I apply different ambient temperatures to different faces?
Split the CAD into governed regions and declare one convection load per environment. Do not select a parent region and then try to override a subset with a second load: overlapping Robin operators can double-count heat transfer unless the formulation explicitly defines precedence.
How do I create a time-varying power or heat-flux cycle?
Use a typed heat_generation or heat_flux signal over time, then apply it with a matching
excitation. Interpolation and extrapolation are mandatory because the solver may sample between
authored points. Signal/function source identity is retained in the result and Desktop preview.
signal power_cycle : heat_generation_density over time {
samples = [
{ at = 0 s, value = 0 W/m^3 },
{ at = 30 s, value = 2000000 W/m^3 },
{ at = 90 s, value = 2000000 W/m^3 },
{ at = 120 s, value = 0 W/m^3 },
]
interpolation = linear
extrapolation = reject
}
excitation cycling_loss {
heat_generation {
target = geometry.module.region.die
signal = signal.power_cycle
}
}How do I apply a measured transient surface heat flux?
Use heat_flux rather than heat_generation when the measurement is power per unit area. The
signal is scalar; its sign is resolved against the selected face's outward normal:
signal measured_flux : heat_flux over time {
samples = [
{ at = 0 s, value = 0 W/m^2 },
{ at = 2 s, value = 12000 W/m^2 },
{ at = 8 s, value = 4000 W/m^2 },
{ at = 10 s, value = 0 W/m^2 },
]
interpolation = linear
extrapolation = reject
}
excitation measured_surface_heating {
heat_flux {
target = geometry.calorimeter.region.heater_footprint
signal = signal.measured_flux
}
}Verify the delivered history . If the measurement is total watts, divide by the admitted footprint area only for a justified uniform-flux approximation.
How do I convert total component power into heat generation?
If a component dissipates total power uniformly in resolved volume , use . Prefer a resultant-preserving power contract when available; otherwise compute the intensity from the admitted CAD volume and record that dependency. Do not divide by a nominal volume that differs from the meshed region. Verify in result evidence.
How do I model an insulated or adiabatic wall?
For the standard conduction weak form, an omitted natural boundary has zero normal heat flux. Keep the face out of heat-flux, convection, and interface declarations and document the intended adiabatic assumption in the study. Do not add a zero-temperature constraint; that is an isothermal sink, not insulation.
How do I model thermal radiation?
The current closed load set does not include a surface-radiation law . Do not approximate radiation with a constant convection coefficient unless that linearization and its operating temperature are part of the model intent. An admitted radiation boundary needs emissivity, surrounding/radiation temperature, view-factor or enclosure semantics, nonlinear tangent behavior, and energy-balance evidence.
How do I linearize radiation near an operating temperature?
If a documented reduced model is acceptable, derive about stated mean absolute temperature and use convection with and the radiation-environment temperature. Record the linearization temperature and run a sensitivity check. This approximation is not a first-class radiation solve and becomes inaccurate over wide temperature ranges.
How do I model heat transfer across a contact interface?
Use conjugate_heat_transfer.perfect, .conductance, or .resistance between the two named
surfaces. A heat flux applies a known boundary source; it does not conserve equal-and-opposite energy
between independently solved bodies. See contact and multiphysics interfaces.
How do I model heat-sink contact or thermal interface material?
Use a conductance/resistance interface when the layer is reduced to a boundary law. Model the layer as a solid with conductivity, density, and heat capacity when its thickness gradients or transient storage matter. Do not represent contact resistance by lowering bulk conductivity in unrelated parts.
How do I model Joule heating from an electrical solve?
Transfer the retained electromagnetic loss field through an electrothermal procedure. The transfer
must conserve integrated electrical loss and record topology mapping. Use manually authored
heat_generation only for a deliberate one-way reduced model with the source calculation retained
as evidence.
How do I model plastic-work, friction, or damping heat?
These are coupled dissipation sources. Use a procedure/interface that transfers the correct fraction of mechanical work into thermal power. A constant guessed heat flux is acceptable only as an explicit engineering approximation. State partition factors, contact side, and time dependence.
Why does the thermal energy balance not close?
Check volume and surface units, source-region volume, convection area, boundary normal signs, interface double counting, transient stored energy, and whether imported coupled losses preserve their resultant. In steady state, total source power should equal total outward boundary/interface power within tolerance. In transient analysis, include the rate of stored thermal energy.
What should I inspect after applying thermal loads?
Inspect resolved volumes/areas, integrated power by source, convection/radiation sink power, contact transfer on both sides, peak and mean temperatures, and global energy residual. For signal-driven loads, verify time-domain coverage and interpolation at every solver step.
For varying heat input, declare a heat_generation or heat_flux signal over time and apply it
with an excitation of the same kind. Surface flux and volume generation are deliberately different
dimensions and target roles. A temperature constraint fixes the primary field; convection and flux
do not.
Check integrated power, exterior area, sign convention, temperature units, and energy balance.
How do I model a chip, PCB, or battery heat source in RunMat?
Create a solid region for each dissipating component, assign its thermal material, and apply
heat_generation using power divided by that region's admitted volume. Keep sources separate when
their powers or duty cycles differ:
load processor_loss {
heat_generation on geometry.board.region.processor_die = 18000000 W/m^3
}
load regulator_loss {
heat_generation on geometry.board.region.regulator_die = 6000000 W/m^3
}
load_case full_power {
include load.processor_loss
include load.regulator_loss
}Verify the integrated watts for each component. Applying total board power to a bounding-box region changes source density and spreads heat through material that does not generate it.
How do I model a prescribed heater power and convection together?
Declare the heater and cooling boundaries as separate loads and include them in the same case. The heater adds energy; convection removes or adds energy according to the current wall temperature:
load cartridge_heater {
heat_generation on geometry.assembly.region.heater = 4000000 W/m^3
}
load ambient_cooling {
thermal.convection {
target = geometry.assembly.region.exposed_faces
ambient_temperature = 293.15 K
coefficient = 12 W/m^2/K
}
}Do not subtract an estimated convective wattage from the heater input; the Robin boundary computes heat exchange from the solved wall temperature.
How do I choose between temperature, heat flux, and convection boundary conditions?
Use thermal.temperature when a controller or reservoir truly fixes surface temperature. Use
thermal.heat_flux when surface power density is known independently of surface temperature. Use
thermal.convection when heat flow is proportional to the difference between unknown wall
temperature and known ambient. These produce different thermal resistances and cannot be selected
only to obtain a desired peak temperature.
How do I check a thermal load mesh refinement study?
Across meshes, compare admitted source volume or area, integrated source power, boundary heat flow, stored-energy change, and temperature norms. Thin layers and sharp source footprints can require local refinement even when total energy closes. A stable peak temperature with a drifting energy balance is not convergence; neither is energy closure with an unresolved local gradient at a material or contact interface.
How do I model a thermal soak followed by cooldown?
Use an ambient-temperature signal for the chamber or fluid history and a convection boundary for the heat-transfer law. Do not prescribe the part temperature unless the test fixture actually controls the part surface to that temperature.
signal chamber_cycle : temperature over time {
samples = [
{ at = 0 s, value = 293.15 K },
{ at = 1800 s, value = 393.15 K },
{ at = 5400 s, value = 393.15 K },
{ at = 7200 s, value = 293.15 K },
]
interpolation = linear
extrapolation = hold_endpoint
}
excitation chamber_air {
ambient_temperature {
target = geometry.part.region.exposed_faces
signal = signal.chamber_cycle
}
}
The selected transient formulation must admit ambient_temperature as the varying film ambient and
must provide the corresponding heat-transfer coefficient. The direct thermal.convection boundary
currently accepts a constant ambient_temperature; do not write a signal reference there and assume
it was consumed. The thermal time constant depends on heat capacity, conduction, exposed area, and
film coefficient. Check energy stored during heat-up, energy rejected during cooldown, and time-step
convergence at the ramps.
How do I model heat from a chip, winding, battery cell, or resistor?
Start from the known power definition. Use heat_generation on the actual dissipating volume when
loss is volumetric, thermal.heat_flux for a thin resolved heater, or an electrothermal procedure
when current distribution must determine . If a component dissipates
total power uniformly in selected volume , the source density is .
Do not apply as though it were : watts and watts per cubic meter are different typed quantities. After projection, verify that and that the source did not spread into non-dissipating encapsulant, substrate, or coolant.
How do I model a prescribed temperature map from test or another solver?
Use temperature_field only when the temperature field is an input to a consuming procedure, such
as thermal-strain loading. A steady or transient thermal solve normally solves temperature; fixing
the same field as a load would remove that unknown. Bind imported data to coordinates, units,
interpolation, out-of-domain behavior, and source identity, then retain projection error and
coverage.
For thermomechanical work, map the thermal result through the typed state edge so the structural procedure can prove which mesh, time/frequency case, and temperature reference produced thermal strain. Avoid exporting anonymous nodal CSV data and reattaching it by row order.
The current .fea grammar does not yet accept an imported topology field for temperature_field.
It accepts a uniform scalar signal on a solid region. The mapping contract described above is the
required future boundary, not syntax that can be authored today.
How do I represent a heater controlled to constant power or constant temperature?
Constant power is a source (heat_generation or thermal.heat_flux). Constant temperature is an
essential constraint (thermal.temperature). A thermostatic controller is neither: it is a
feedback law that switches or modulates power from measured temperature. Model that controller with
an admitted function/coupling contract; do not alternate between power and fixed temperature to
force the desired curve.
For every control representation, report commanded power, delivered power, sensor temperature, switching or saturation state, and global energy balance.
How do I verify a complete steady thermal model?
Compute each contribution in watts rather than comparing temperature contours:
At steady state, signed source and sink power must close, including every interface transfer. Then refine the mesh and compare integrated powers, mean temperature, and the temperature at engineering probe locations. A peak at a point source or sharp material corner may not converge; qualify it by distance or averaged region. For transient work, add the stored-energy rate and repeat with a smaller time step.
Further reading: thermal constraints, heat-transfer studies, and electrothermal analysis.