This is a bit of an open-ended question and I'm sure if you ask a dozen different engineers you'll get a dozen different answers with this prompt. The first thing that comes to mind though is if the temperature and the RPM are correlated (i.e. they vary together whether trending in the same or opposite directions), could you come up with a formula to describe the range which you expect them using an analytic and create an event frame if it goes out of that range?
If that's too prescriptive and you can't develop a specific function for their correlation, maybe something as simple as the difference between the slopes of the trends over the previous 5, 10, 15 minutes can exceed a certain amount.
Just some thoughts.