Chapter Four · failure evidence
What Thermal & Heat Transfer Simulation got wrong, from 61 dissertations
The records evaluate computational and analytical models across diverse thermal engineering and heat transfer applications. Researchers repeatedly found that oversimplified thermal properties, lumped geometries, decoupled physics, and idealized boundary conditions led to significant numerical errors and simulation failures. These records come from PhD theses at 16 institutions, 2021 to 2026. Each links to its thesis. They were extracted by language models reading the full text, so treat each as a lead to read, not a verdict.
Static and uncalibrated transport property models fail under complex thermal conditions
Constant property assumptions and uncalibrated conductivity models introduce substantial drift during transient heating or non-equilibrium transport. These formulations fail to capture phenomena such as moisture interactions, microscopic carrier spreading, and phase change propagation rates.
Tried and failed
constant parameter coarse-grained effective temperature measurement applied to coarse-grained non-equilibrium molecular simulations. Reason: incorrectly predicted higher effective temperatures at lower cooling rates, violating thermodynamic theory
Tried and failed
empirical radial temperature gradient correlations applied to turbomachinery rotor disc thermal modeling. Reason: overestimated radial temperature drops leading to unphysically low bore temperatures
Novel turboshaft engines design and optimisation for rotorcraft applications. · Cranfield
Tried and failed
constant thermal properties in dynamic heat transfer applied to battery thermal equivalent circuit modeling. Outcome: did not generalise. Reason: cumulative error over long transient cycles caused model divergence from experimental data
Thermal-managed oriented, multi-domain battery modelling and experimental validation. · Cranfield
Tried and failed
apparent heat capacity method with elevated post-solidification capacity applied to multiphase freezing front thermal diffusion. Outcome: worse than baseline. Reason: artificially elevated heat capacity after solidification severely underpredicted propagation speed and overestimated liquid lifetimes
Europa's Surface and Shallow Water: Ice Shell Activity and Implications for Habitability · Georgia Tech
Considered and rejected
Considered and rejected: Rejected constant porosity assumption in heat transfer modeling because variable porosity accurately captures solid-dominant early heating followed by insulating pore air behavior
Contribution à la modélisation des phénomènes de transport et des modifications au cours du processus de cuisson · DSpace at the University of Batna 1
Tried and failed
uncalibrated default gas thermal conductivity models applied to numerical droplet combustion simulations. Outcome: worse than baseline. Reason: standard transport formulations overpredicted gas-phase heat conduction without an empirical 0.9 correction factor
Considered and rejected
Considered and rejected: Rejected the ConFD (Conduction Finite Difference) algorithm in EnergyPlus because it models temperature-dependent thermal conductivity without accounting for moisture effects
The Global Role and Impacts of Building Performance Diagnostics Under Climate Change Considerations · Georgia Tech
Tried and failed
microinstability transport models applied to tokamak edge plasma thermal conductivity. Outcome: did not generalise. Reason: models failed to quantitatively match experimentally inferred heat conductivity across edge operating regimes
Interpretation of Thermal Conductivity and Toroidal Momentum Transport in DIII-D Taking Into Account IOL and Pinch Velocity · Georgia Tech
Tried and failed
equilibrium molecular dynamics with machine learned potentials applied to interfacial thermal boundary conductance. Outcome: did not generalise. Reason: simulated boundary conductance failed to achieve quantitative agreement with experimental measurements
Lost to a baseline
At thicknesses below 150 nm, the proposed anisotropic/isotropic BTE models showed worse agreement for graphite in-plane thermal conductivity than Minnich's Fuchs-Sondheimer BTE model.
Lost to a baseline
Linearised Boltzmann (LB) momentum relaxation time approximation fails to capture hot-electron/Joule-heating distribution spreading at low impurity densities, overestimating drift velocity compared to full DG/MC solutions.
Electronic transport in graphene: 2D materials, composites, and 3D printed structures · University of Nottingham Repository
Lumped capacitance and low order network models omit spatial temperature gradients
Zero dimensional lumped capacitance and low order resistor-capacitor networks assume uniform temperatures that erase local thermal gradients. Consequently, these models miscalculate transient cooling rates, heat storage capacity, and peak temperatures across multi-element systems.
Tried and failed
lumped thermal capacitance modeling applied to inter-cell temperature distribution in battery modules. Reason: assumed uniform temperature per cell, losing spatial gradients needed to resolve inter-cell non-uniformities
Tried and failed
0D lumped capacitance thermal modeling applied to transient solid component heat transfer. Outcome: worse than baseline. Reason: Underestimated transient heat transfer and caused non-physical fast thermal decay compared to distributed 1D modeling
Aircraft engine transient performance modelling with heat soakage effects · Cranfield
Tried and failed
lumped capacitance thermal model applied to beam saturation temperature prediction. Reason: omitted spatial temperature gradients, leading to significant overprediction of saturation temperature
Fatigue crack growth and localisation of strain in zircaloy-4 · Imperial
Tried and failed
tuning layer thickness in RC network models applied to thermal dynamics of heavy building walls. Reason: increasing thermal resistance inherently coupled directly to capacitance in fixed-material assumptions
Comparison of Simplified Physics-Based Building Energy Model to an Advanced Neural Network for Automatic Fault Detection · Georgia Tech
Tried and failed
lumped thermal time constant compact models applied to transient thermal dynamics during power cycling. Reason: could not adequately capture both transient temperature rise and drop
Investigation of Transport Characteristics of Gallium Oxide FETs and Carbon Nanotube Network based FETs for Emerging Applications · Georgia Tech
Tried and failed
single-node RC lumped thermal model applied to building indoor temperature and energy prediction. Outcome: worse than baseline. Reason: failed to capture directional solar-air temperature variations and thermal capacitance dampening across varying insulation levels
CLIMATE—CARBON—EQUITY Making Sustainable Design Concepts Accessible for All · MIT
Tried and failed
lumped RC grey-box thermal modeling applied to building envelope thermal dynamics. Outcome: did not generalise. Reason: failed to capture dynamics without explicitly modeling solar absorption and re-emission in opaque components
Advanced control strategies for adaptive building envelope systems · IRIS - POLITO - prod
Lost to a baseline
The standard 5R1C building model underestimated shiftable electricity demand by 1.4 to 5.2 times compared to the 6R2C floor-heating model and IDA ICE detailed dynamic simulation
Integrating heat pump flexibility in building stock modeling to support decarbonization through sector coupled energy systems · DSpace-CRIS at TU Wien
Considered and rejected
Considered and rejected: Single-tau thermal model rejected in favor of the two-tau relaxation model for heat capacity analysis when poor thermal conductance existed between sample and platform.
Universal Aspects of Metamagnetic Transitions in Frustrated Ising Chains · EPFL
Analytical and empirical models fail to predict thermal fields in manufacturing melt pools
Closed-form analytical formulations and single energy density metrics neglect hydrodynamic instabilities, laser pulsing parameters, and droplet mass transfer. Without full numerical treatment or empirical corrections, these models miscalculate melt pool dimensions and resulting microstructures.
Considered and rejected
Considered and rejected: Weld thermal simulation methods rejected because they fail to reproduce real steep thermal gradients, fast heating rates, and actual stress/thermal fields of real welds
A study of coarse grain heat affected zone of accelerated cooled structural steels · Cranfield
Tried and failed
analytical thermal modeling without numerical iterations applied to melt track instability prediction in additive manufacturing. Outcome: did not generalise. Reason: simplified closed-form geometric melt pool calculations fail to capture complex hydrodynamic balling instabilities at specific process parameters
Prediction of Part Quality in Laser Powder Bed Fusion Metal Additive Manufacturing · Georgia Tech
Tried and failed
analytical thermal modeling applied to pulsed keyhole laser welding. Outcome: did not generalise. Reason: neglects temperature-dependent absorption, wavelength, pulse duration, and inter-pulse cooling effects
Application of fibre lasers in fabrications and processing of thin gauge alloys for engineering applications · Cranfield
Tried and failed
constant energy density process parameter scaling applied to metal additive manufacturing microstructure control. Outcome: did not generalise. Reason: energy density alone fails to capture thermal gradients and melt pool dynamics governing solidification morphology
Grain structure control of titanium alloys during wire + arc additive manufacturing · Cranfield
Tried and failed
1D lumped energy input parameter scaling applied to welding melt pool geometry prediction. Outcome: did not generalise. Reason: nominal energy ignores distinct contributions and distributions of arc heating versus droplet mass transfer
Fusion characteristics in P-GMAW of mild steel · Cranfield
Tried and failed
geometric feature regression without thermodynamic process variables applied to predicting mechanical properties in manufacturing. Outcome: no signal. Reason: melt pool dimensions alone lacked necessary thermal gradient information governing microstructure and strength
Tried and failed
closed-form analytical thermal modeling applied to melt pool dimension prediction. Outcome: did not generalise. Reason: asymmetric nonlinear sensitivity to input parameter underestimation versus linear sensitivity to overestimation
Prediction of Part Quality in Laser Powder Bed Fusion Metal Additive Manufacturing · Georgia Tech
Considered and rejected
Considered and rejected: Rejected relying on thermal conduction models alone without an empirical correction factor (F = 1.75) for predicting weld pool dilution in V-groove geometries.
Fusion characteristics in P-GMAW of mild steel · Cranfield
Idealized boundary conditions misrepresent convective cooling and transient thermal retention
Prescribing constant heat transfer coefficients, constant bulk temperatures, or uniform wall heat fluxes fails to replicate local fluid entrapment and entrainment physics. These simplified boundary specifications cause excessive cooling and underestimate thermal loads in transient or radiant environments.
Tried and failed
prescribing surface heat flux boundary condition applied to stable boundary layer simulations. Outcome: unstable. Reason: caused unphysical excessive cooling during the decay of inertial oscillations
Power output and wake effects of very large wind farms investigated by large-eddy simulations · Leibniz Universität Hannover Repository
Lost to a baseline
Simplified in-flight analytical cooling model was beaten by the detailed numerical heat transfer/forced-convection model developed with University of Warwick due to underestimating convective heat loss.
Drop-on-demand metal additive manufacturing from droplet formation to 3D structures · University of Nottingham Repository
Lost to a baseline
Mori's constant-thickness convective heat transfer model failed to correlate with measured cyclopentane hydrate growth rates across varying subcooling temperatures.
Investigations of clathrate hydrate formation using advanced optical diagnostic techniques · Imperial
Considered and rejected
Considered and rejected: Rejected the constant-volume axisymmetric box model for final thermal analysis due to lack of entrainment, cooling, and boundary physics.
The development of extremely high-grade ignimbrites: the effects of strain heating and implications of ash agglutination · Scholars' Bank
Tried and failed
heat balance with convective transfer assumptions applied to radiant cooling load calculation. Outcome: did not generalise. Reason: tabulated coefficients assume purely convective heat extraction, underestimating cooling requirements for radiant systems
Indoor and outdoor radiant cooling systems for buildings · UT Austin
Tried and failed
isothermal wall boundary condition applied to repetitively pulsed discharge thermal modeling. Outcome: did not generalise. Reason: overestimates conductive heat loss to walls over long multi-pulse sequences compared to single-pulse regimes
Hydrodynamic effect of nanosecond discharge pulses in air : high fidelity models and scaling laws · UT Austin
Considered and rejected
Considered and rejected: Rejected conventional regular convection boundary (RCB) with constant bulk temperature because it led to excessive cooling and failed to capture transient conductive heat retention
Bead shape control using multi-energy source (mes) for wire-based directed energy deposition (ded) process. · Cranfield
Tried and failed
constant heat transfer coefficient boundary condition applied to thermal modeling in complex geometries. Outcome: did not generalise. Reason: fails to capture local fluid entrapment and spatial thermal variations in non-cylindrical shapes
Design of gas adsorption systems with phase change materials aided by topology optimisation method · Imperial
Dimensionality reduction and homogenized conductivity assumptions distort spatial heat paths
Reducing three dimensional assemblies to two dimensional cross sections or using parallel homogenized conductivities misrepresents heat conduction paths. Treating heterogeneous regions and thermal bridges as uniform media introduces substantial bias into local temperature and heat transfer predictions.
Considered and rejected
Considered and rejected: Rejected simplified semi-infinite slab analytical model for thin substrate layups due to failure to capture significant conductive bulk heating (requires finite slab switching).
Heat transfer modelling, optimisation and monitoring of flashlamp- assisted automated tape placement. · Cranfield
Considered and rejected
Considered and rejected: Rejected homogeneous substrate thermal modeling (standard HotSpot/3D-ICE) for 2.5D chiplets because treating air gaps as silicon mispredicts inter-chiplet heat propagation.
System-aware Architectural Co-design to Tackle the Memory Wall · EPFL
Tried and failed
2D cross-section modeling of periodic discrete conductors applied to heat transfer across insulated assemblies. Reason: Smearing 3D discrete thermal bridges into 2D either overestimates or underestimates transverse heat conduction
Computational study of failure of curtain wall facade systems exposed to fire · Imperial
Considered and rejected
Considered and rejected: Homogenized parallel thermal conductivity for multi-layered formations, because it severely underestimates bottomhole interval heat transfer improvements (predicted 2.33% vs 11.25%).
Considered and rejected
Considered and rejected: Rejected standard area-weighted average equation (parallel path formula) for standalone compliance analysis because it assumes fully effective thermal heat transfer across the entire component and overestimates R-value reduction
Considered and rejected
Considered and rejected: Rejected relying solely on 2D thermal simulations (e.g., THERM) for complex assemblies in favor of 3D thermal modeling coupled with whole-building simulation
Tried and failed
effective medium continuum model applied to nanofluid effective thermal conductivity. Outcome: worse than baseline. Reason: neglects nanoscale interfacial thermal resistance and non-local transport mechanisms present in atomistic mixtures
Non-local methods in transport phenomena · UT Austin
Lost to a baseline
Analytic Carslaw-Jaeger cylindrical thermal conductivity solution was systematically biased 10–20% higher than numerical modeling by Grott et al. (2021) due to low mole aspect ratio (<15).
Geodynamic Modeling of Mars Constrained by InSight · Virginia Tech
Decoupling thermal transport from fluid and mechanical interactions produces inaccurate predictions
Simulating heat transfer independently of structural deformations, electro-thermal feedback, or conjugate fluid boundaries leads to severe predictive errors. Neglecting these coupled interactions distorts internal temperature distributions and overestimates structural stresses.
Tried and failed
Physics-informed neural networks for conjugate heat transfer applied to indirect liquid cooling thermal modeling. Outcome: did not generalise. Reason: Inconsistent temperature drop trends across varying flow rates between PINN, CFD, and experimental data
Experimental Validation of Indirect Liquid Cooling Computational Thermal Modeling · MIT
Tried and failed
isothermal DC simulation applied to RF transistor breakdown voltage prediction. Outcome: did not generalise. Reason: Isothermal assumptions ignore electro-thermal feedback and thermally assisted breakdown mechanisms.
Advancing SiGe HBT Performance: A Study on RF-SOA Prediction, BEOL Engineering, and Layout-Driven Reliability Optimization · Georgia Tech
Tried and failed
decoupled sub-element simulation ignoring cross-element conduction applied to reactor core thermal-hydraulic modeling. Reason: neglecting inter-element thermal conduction distorts local temperature distributions and underpredicts coolant heating
Coarse-mesh-based Reduced-order Package for Multiphysics Simulation of Nuclear Thermal Propulsion Reactor Core · Georgia Tech
Tried and failed
fully decoupled multiphysics finite element simulation applied to thermo-active pile foundations in clay. Outcome: worse than baseline. Reason: ignoring transient consolidation and heat transfer interactions severely overestimated thermally-induced structural stresses
Tried and failed
omitting conjugate heat transfer coupling applied to particle thermal degradation simulation. Reason: removes thermal boundary layer causing unrealistically fast conversion
Multi-scale simulations of biomass particle pyrolysis using lattice Boltzmann method · Iowa State
Tried and failed
linear temperature profile approximation applied to counter-flow heat exchanger modeling. Outcome: did not generalise. Reason: unbalanced fluid heat capacitance causes nonlinear thermal profiles leading to large mean fluid temperature errors
ON THE SUITABILITY OF PRINTED CIRCUIT HEAT EXCHANGERS FOR HIGH TEMPERATURE NUCLEAR SERVICE · Georgia Tech
Tried and failed
uncoupled thermal-displacement finite element modeling applied to porous media consolidation simulation. Outcome: worse than baseline. Reason: Inability to track pore pressure dissipation and changing void ratio during consolidation
Numerical modelling of the electro-kinetic effect on pipe-soil interaction. · Cranfield
Numerical solvers experience convergence failures or excessive computational overhead
Explicit time stepping and extreme operating conditions cause severe numerical instabilities when stability criteria or physical voltage limits are breached. High fidelity finite element models can also prove computationally prohibitive for real-time simulation and iterative design tasks.
Tried and failed
standard compact modeling at true cryogenic temperatures applied to heterojunction bipolar transistor simulation. Outcome: did not converge. Reason: extremely small thermal voltage caused severe slope error and simulator convergence failure
Using SiGe HBTs for quantum science at deep cryogenic temperatures · Georgia Tech
Considered and rejected
Considered and rejected: Rejected FEA-based thermal simulation because it is computationally prohibitive and cannot update road temperatures in real-time during interactive parameter tuning.
A visual process simulation for novice users of desktop 3D printers to reduce part defects · Iowa State
Tried and failed
explicit finite difference transient thermal modeling applied to building envelope thermal performance estimation. Outcome: unstable. Reason: numerical instability when Fourier numbers dropped below stability thresholds, requiring severe time-step throttling
Tried and failed
axial marching subchannel solver with crossflow applied to low-flow natural circulation thermal hydraulics. Outcome: unstable. Reason: low flow rates, steep radial power gradients, and small subchannel geometry caused numerical instabilities
A Whole-Core Thermal Hydraulic Model for Pin-Fueled Fluoride-Salt-Cooled Reactors · Georgia Tech
Considered and rejected
Considered and rejected: Rejected internal energy/enthalpy formulations across a temperature interval for thermal spray, using temperature with exact latent heat tracking at constant melting temperature.
Particle-based approach to model fuel droplet impact and thermal spray coating processes · Iowa State
Left open by the authors
Problems the authors named and did not get to.
Left open
Simulate and analyze post-quench slow voltage and temperature drift duration across varied deposited energy, cooling conditions, and exponential current dumps. Blocker: Requires experimental validation data from the SULTAN test facility or proprietary multi-physics setup parameters.
Left open
Experimentally validate the 1D thermal resistance model using physical iPEBB substrate walls and power electronics components. Blocker: Requires physical hardware and experimental thermal measurement apparatus for iPEBB substrate walls and components
First Order Thermal Model of the Navy integrated Power Electronics Building Block · MIT
Left open
Model dynamic thermal management system interactions and temperature feedback for power electronics and PEM fuel cells in the MATLAB/Simulink propulsion framework. Blocker: Lacks specific governing equations, thermal architecture specifications, and empirical validation parameters for the proprietary aircraft model.
Modelling and evaluation of electrified regional aircraft propulsion systems · Cranfield
Left open
Perform 3D finite element thermal analysis of the integrated Power Electronics Building Block to validate analytical thermal resistance models. Blocker: None
First Order Thermal Model of the Navy integrated Power Electronics Building Block · MIT
Left open
Model flash temperatures and ambient humidity using micro-scale material properties and validate experimentally under loads >20 N and temperatures >100 °C. Blocker: Requires a physical tribological test rig (pin-on-disc setup with heating and loading controls) for experimental validation.
Left open
Conduct physical experiments to validate the thermal simulation results and directly measure the thermal conductivity of Ferrite N49. Blocker: Requires physical experimental apparatus, converter hardware, liquid cooling test bench, and material measurement tools.
Thermal Analysis of Indirect Liquid Cooling for the Navy integrated Power Electronics Building Block · MIT
Left open
Validate the InVEST urban cooling model against micro-scale urban canopy energy balance models and computational fluid dynamics simulations. Blocker: None
Left open
Convert photothermal cantilever simulations from 2D to 3D and directly validate temperature profiles using thermocouples or temperature-sensitive samples. Blocker: Requires a physical lab with microfabricated cantilevers, thermocouples, or temperature-sensitive samples for experimental validation.
Ultra small biomorphic cantilevers for high-speed atomic force microscopy · EPFL
Left open
Optimize the hardware footprint of converter submodules considering thermal stress, safety, EMI criteria, and assembly constraints. Blocker: Requires physical hardware design, fabrication, and physical testing apparatus for thermal and EMI validation
Modeling, Control, and Fault-tolerant Operation of the Isolated Modular Multilevel DC-DC Converter for MVDC Applications · Georgia Tech
Left open
Model and evaluate thermal cooling configurations for the 800W transformer coils using CFD simulation. Blocker: Undefined physical CAD packaging and system integration constraints for the transformer coils
Thermal Analysis of Indirect Liquid Cooling for the Navy integrated Power Electronics Building Block · MIT
Checking a claim in this area?
We can run the same search on any method or claim. If nothing turns up, we will say so, and that proves nothing on its own.