Map temperature for thermal structural analysis in MSC Patran/Nastran

[D.Addari]

Hi,

I have a model built in APEX and converted in Patran comprising of 40+ components, and I need to perform a thermal-structural analysis on it in Nastran.
The temperature of the elements are calculated using another software (ESATAN-TMS), and the results are provided as nodal temperatures (each thermal node in ESATAN corresponds to the centre of a structural element in NASTRAN). From the Nastran handbook, I see I can apply temperature loads to elements using TEMPF card, meaning I am able to convert the thermal node temperature to a load temperature on the structural element at a specific instant for all the components in the FEM.

The problem is that I can only define one TEMPERATURE(INITIAL) and TEMPERATURE(LOAD) sets in a NASTRAN executive file, whereas I need 40+ INITIAL and LOAD subsets.
In short I need to:
- taken the lower temperature for each component from the worst cold case scenario from the thermal analysis, assign them as initial conditions to the structural FEM
- taken the highest temperature for each component from the worst hot case scenario from the thermal analysis, assign them as the load (end) conditions to the structural FEM

Any help is extremely appreciated.
Thanks,
Dan

 

[dmapguru]

I don't think TEMPF is going to be of any use to you as it is specific to P-elements, not H-elements. If you prepared your data in Apex and Patran, I doubt you are using P-elements.

If you have a temperature field which is defined by a point at the centroid of every element, it would be a reasonable assumption that the GRID points of these elements all have the same temperature as the centroid (i.e. each element has a uniform temperature field). If you then assign this temperature to the GRID points on a per element basis, you end up with many GRID points where a temperature is defined multiple times depending on how many elements are attached to the GRID point. In this case, Nastran will average the temperatures of the multiple definitions.

In order to get the temperatures at the GRID points, you will need either to use Patran to create a temperature field based on the centroidal temperatures, and then extrapolate (map) the temperatures to the GRID points (this can take Patran a long time for a large model), or you could write a program in something like Python that will associate your centroidal temperatures to each element, and then loop over all the elements in the model and assign the element centroidal temperature to all GRID points connected to that element.

The bulk data entries you need to write are TEMP entries (GRID point temperatures), and possibly a TEMPD entry for any GRID points where the temperature is not defined explicitly.

Finally, if I understood your problem correctly, you have a worst case initial temperature and worst case final temperature for each component. In this case, why not write each component's temperatures to TEMP entries where each component has their own SID, then use the LOAD entry to combine all the TEMP SIDs. In this way, you will have your single TEMP(INIT) set and single TEMP(LOAD) set that corresponds to the worst case initial and final temperature distributions.

DG

 

[D.Addari]

Thanks DG for the response.

In the end, I used the TEMP card, one for the initial condition and one for the final condition. The cards are then called by TEMPERATURE(INITIAL) and TEMPERATURE(LOAD).
I took the nodal temperature from the thermal software, and assigned it to the grid points. Each nodal temperature corresponds to a set of grid points (based on the mesh of the thermal and structural models). This way I was able to assign the correct temperature to each component. It took a long time but the model is working correctly.

DA