Learning Forum

[Laure]

Hello Ansys community,

I need to calibrate a 2D transient temperature model to match observed time-series temperatures. Recorded temperatures correspond to the diurnal temperature variation in a solid body at a specific depth (attached is a cross-plot example). I will use variables such as thermal and mass properties to obtain the match. Wonder if I can define as the objective/error function a time series of observed temperature minus simulated temperature, or any other approach that allows me to optimise the calibration process.   

Thanks in advance.

 

[Chandra Sekaran]

Wonder if I can define as the objective/error function a time series of observed temperature minus simulated temperature, or any other approach that allows me to optimise the calibration process.   

Perhaps you can use RMSE as explained at https://c3.ai/glossary/data-science/root-mean-square-error-rmse/#:~:text=To%20compute%20RMSE%2C%20calculate%20the,square%20root%20of%20that%20mean

[Laure]

Thank you Chandra,
The point is that I do not know where to parametrize temperature time series in Mechanical. The options that I can see are all scalars (min, max, average). Any idea??
Thanks in advance.

 

 

[Chandra Sekaran]

You can use Mechanical scripting interface to retrieve this data and then do the RMS error calculation. Or may be better to do with APDL commands in Post26. In Post26 you can use NSOL command to store temperature time history at a node. You can retrieve this data into an array parameter using VGET command. May be others on the forum have better idea of how to do within Mechanical.

[Laure]

Thanks Chandra, I haven't used the command line approach in Ansys. Do you know if there is any documentation (e.g., tutorials) for beginners??