The two critical aspects when it comes to verifying thermal analysis are the model discretisation and idealisation and the reliability of material and geometrical data. As a result, the mesh generation of 3D thermal models is critical.
Although existing tools to automatically generate the mesh exist, the classical approach to mesh thermal models is still widely used and mostly performed “manually” by the space thermal engineers' community.
It is very time consuming, prone to errors and dependent on the user.
A lot of methods to automatically generate the mesh have been implemented so far in different fields. However, there is still no standard method globally available for the whole space thermal engineers' community as such.
The regions that should be subjected to a mesh convergence exercise are the regions with significant temperature gradients and where the material choices are such that thermoelastic distortion is not reduced or eliminated by virtue of design.
Typically, components and units, with heat power dissipations, mounted on Printed Circuit Board and satellite structural panel or highly dissipating units within an instrument will cause temperature gradients over the sub-assembly and the adjacent structure. These highly dissipating components, and the adjacent structure, should be subjected to a mesh convergence exercise.
To identify the regions of the thermal models that require to be subjected to a mesh convergence exercise, an initial analysis run is performed by the thermal engineer, using fine meshes by default to highlight the temperature gradients within the same part, between parts and over time.
For instance, on an optical instrument, using a mesh size that is at least smaller than the size of the smallest interface of unit that is mounted on the optical instrument.
After the first analysis run, the thermal engineer increases the size of the mesh and re-run the model, etc. If the temperature change exceeds 0.1°C by default between the two runs, then the thermal engineer stops the iteration, so the mesh size is the one that does not allow a change of more than 0.1°C.
To make sure that a thermal model that has its mesh converged to sufficiently capture the temperature fields, the temperature change of thermal nodes between two different analysis runs of thermal model with different mesh size for all the analyses cases should not exceed 0.1°C by default, which would mean that the mesh has converged.
Surrogate models, trained on test data, can be used to evaluate mesh size iterations by mimicking the behaviour of the thermal simulation models as closely as possible which remove the need to run the thermal simulation models several times to meet the mesh convergence criteria.
Learn How to make a reliable thermal harness model for extreme space conditions ?
Benefits: Known approach.
Limitations: Requires running the analysis cases various times, built on user-defined basis, time-consuming and error-prone.
Standardising the thermal mesh generation process will reduce both time and computing, which affect the product development budget, as well as the risk.
We will standardise and fully automate this process and make it accessible to the entire space thermal engineers’ community via our Digital Engineer® Try it now