A thermal model for industrial automated tape laying was compared with measured tape temperatures during a dynamic process trajectory. At the tape surface facing the surroundings, the reported maximum absolute error was 13.52°C and the maximum relative error was 4.19%. Its root mean square error, or RMSE, a single measure of the temperature differences across the comparison, was 4.02°C; the normalized RMSE was 1.08%.
Those figures come from a calibrated model. Uncertain system parameters were iteratively adjusted within plausible physical ranges until simulated results closely aligned with experimental data, without changing the qualitative physical temperature profile. The reported performance was evaluated on the same dynamic trajectory and measurement point used for tuning, so independent validation under varying operating conditions remains outstanding.
A model for a moving, heated tape
The model couples advection, conduction, convection and radiation to describe heat transfer in the automated tape-laying process. It also uses analytical view factors for finite emitter and tape widths, assesses local mixed convection, and represents the tape’s opposing surfaces as two nodes.
For the time-dependent calculation, the researchers used a self-implemented two-stage Radau IIA implicit Runge-Kutta scheme of order 3 for stiff differential equations. The two-node evaluation reported opposing-surface temperature differences of up to 6 K. In the cited case, a lumped mean-temperature approximation had approximately 2% relative error around a mean temperature of 334 K.
Geometry and airflow matter
Radiation geometry produced a large modeling difference. For the reported geometry, a 1.5D simplification overestimated the real radiation exchange area by 159%. Flattening the curved emitter profile added another 29% overestimate. These were geometry-comparison results, not independent experimental uncertainty estimates.
The reported flow-regime analysis classified the conditions as mixed convection, meaning natural and forced convection were both represented. The Richardson number was about 6.23, with Reynolds and Rayleigh checks of about 658 and 2.19 × 10^6, respectively, within the stated laminar limits. A supplementary comparison found that a constant forced-convection model underestimated local heat losses and produced higher predicted tape temperatures than the local mixed-convection model.
Accuracy meets computing cost
The grid analysis showed different effects from refining the tape and heater grids. The baseline grid used 30 tape segments and 15 heater segments, and the simulation took 347 seconds. Refining the tape grid to 60 segments changed the measurement-point result by 0.17% NRMSE and 0.61 K RMSE. Refining the heater grid to 31 segments gave 1.86% NRMSE and 6.93 K RMSE, while increasing runtime by about 148% to 862 seconds.
What the test leaves unresolved
The reported validation used a dynamic trajectory and a defined measurement point, and the same trajectory and point were used for tuning. Independent testing under varying operating conditions remains to be done. Real-time execution was not targeted.
The authors also flag uncertainty in emissivity and convection coefficients because measurement equipment was limited. The radiation treatment uses diffuse-gray and lumped optical assumptions that may miss band-specific radiative effects.
Preprint status
The supplied document carries arXiv:2608.25470v1 and the date 26 August 2026. It states that all data supporting the findings are available within the article and supplementary materials. The work was supported by the LCM - K2 Center for Symbiotic Mechatronics within the Austrian COMET-K2 program.
Paper data and sources
Original title: Transient multimode heat transfer of an industrial automated tape laying process under rapidly changing conditions
Authors: Bernhard Rameder, Hubert Gattringer, Andreas Müller, Ronald Naderer
Journal/Repository: Composites Part A: Applied Science and Manufacturing (2026)
Status: Preprint, not yet peer-reviewed
First online: 2026-08-26
DOI: Not available
Original paper · Full text