I have over 500 hours of audio recordings of my server room. There's a lot of background noise from the fans, but mostly it's constant. Is there a way to automate an analysis of these audio files (mp3 each 30 min) and display changes? Like frequency or decibel changes?

Thanks,
Traigo

Yes, there is. Stream and decode the file, pipe the audio data, analyze it using a Fast Fourier Transform (FFT). After that, things get creative and complicated.

Here's an FFT tutorial:

Signal Processing Workshop

Here's a direct link to the above article's source TAR archive containing an FFT processor:

http://arachnoid.com/signal_processi...cessing.tar.gz

Yeah...I wa able to install and run the samples, but how to use this for my needs just seems to be over my head.

You would need to convert the MP3 file into time history data file. This creates an array of data that a FFT too can analyze. Your automated program will convert the file, read the data, perform FFT analysis and then finally compare the data.
As an fully automated process I might suggest python. You can convert the y axis to dBs.

I've never used these tools so do not know there limitations as to how many data points can be processed at one time. You may have to subdivide each file into something smaller.

http://www.integratedprocess.com/sigproc.html
http://people.csail.mit.edu/jrennie/python/numeric/
http://numpy.scipy.org/

Now what is your main objective? Hearing is very non-linear. You may want to weight the data.
http://en.wikipedia.org/wiki/A-weighting