Hi Everyone,
In ls-prepost there is the option when plotting XY data to apply a filter to the data, my question is if there are any general recommendations for when to apply a filter and how to select the filtering method? Coming from a physical model one might apply a filter to remove noise from the signal due to the instrument such as a force transducer, but I believe that wouldn't be a problem in a numerical model so when should filters be applied to data?
Cheer!
Hey @pjay,
Great question. To start, yes you can filter time history data and export it with the filtering. In the plot window, go to the Filter tab, then click the dropdown by the Filter and select the filter system you would like to use (SAE is common), then select the time unit that your model is using, and select the filter frequency next to C/s(Hz). Then, make sure the curve in the bottom left box that you want to filter is selected and click Apply. You can then export this by clicking the Save tab, select Output Type (csv is common), browse for the path to save to and enter a filename (the relative extension will be written according to your output type selection). Then click Save.
Filtering is a topic of conversation within computational modeling. It's generally a good idea to replicate filtering practices that are used in similar real-world applications. This means that your filtering will be dependent on the field that you are working in. Again, SAE would be a good place to search for these guidelines but it may take a little time to find something.
Additional filtering can be okay with simulation results since simulations are numerical and can have more noise than their physical counterparts, depending on modeling techniques. Just make sure to always report the filtering method and frequency that you use because then people can make their own assessment of whether they believe your results or not. This adds a little bit of built-in accountability for the people creating the simulations and reporting on results.
This NASA paper is a good read on the subject:
https://ntrs.nasa.gov/citations/20020085101
In the end, I generally stay away from filtering below SAE 600 unless there is a good reason. You should be okay with SAE 1000 or 600 as long as you report it.
@negativevolume, thank you for the detailed response and the NASA paper is a great paper on the topic!