I have a sampled temporal signal, result of a transient fluid simulation. At the beginning the flow is being established, so the first few seconds of signal should be ignored (question : is there a mathematical way to determine how much time at the beginning of the signal to truncate ?).
I would like to assess how well the signal is converging towards a signal oscillating around a fixed average, and determine what is an appropriate averaging window size.
I also would like to have a way of judging the 'confidence' in the obtained average (i.e. if I continue recording the signal, what confidence do I have that I will be within a certain tolerance of the same average using the same or larger window size ?)
Doing an FFT of the signal I see large magnitudes at the lowest frequencies. Does this mean my sample size is not large enough ? (i.e the lowest frequencies that can be detected are still too high ? )