So, as a former statistician, software guy (math/numerical analysis was my focus in grad school), and recording dabbler for some time now, I'm going to try to be patient, but you must understand that what you're doing doesn't make a ton of sense from almost any technical viewpoint, at least as simplistically as you describe it.
In any case, this is not a "home recording" kind of thing you're asking about, so you might pose your question on a scientific research forum somewhere, just to get some other opinions while we're trying to imagine what's really going on.
As [MENTION=89697]ecc83[/MENTION] (Dave) says, quantizing sound is done with calibrated tools. You can do this anywhere, anytime, but if you're going to do something scientific, you need a controlled environment, so you need to worry about every bit of sound that's being captured, and that means a contained space for your engine to be running in, with good sound isolation from external noises. And a good exhaust fan, unless this is an electrical motor and not internal combustion?
Now, whether you need a calibrated microphone I think is a good question. If you're trying to establish a baseline, then it's important to know whether it's going to include ultrasonic and subsonic measurement data, because those kinds of microphones are not something most home recorders are interested in, and few of us even have calibrated measurement microphones or sound level meters, which you'd need to correlate the physical volume level (dB) at the microphone against the waveform captured digitally. (I'm leaping at the conclusion you'll want this data in a computer - something you've not explicitly stated.)
Once the sound(wave) is captured in the PC, you can do spectrum analysis on that. If the microphone is not flat (and practically none are), you'd have to bias it against the microphone's response that you recorded with a flat input (white noise or sweep). Keep in mind the room where the motor is located will add a great deal and "unflatten" anything recorded unless it's is essentially an anechoic chamber.
If all you want to do is compare scenario A vs B, eg., the motor at idle vs at 2x idle RPM, then you may not need to calibrate the microphone, but you still need the room to not be contributing additional data to the recording, which it will do if it doesn't absorb all frequencies equally.
Any test case at t0 vs t1 would need to use the same microphone, always tested to insure it still recorded sounds the same way, whether you were validating its flatness or simply that it produced the same output for the same input.
I still think if you are just measuring volume at different frequencies under different load conditions (e.g.), any good microphone can provide enough information to provide good data. The space where you are measuring will be the problem.
P.S. A USB microphone or a microphone via a USB audio interface will put data into a computer. Audio software can generate a statistical analysis of a section of the recorded wave form. You can do what you want with that.
P.P.S. Earthworks is the company I'm familiar with that produces good measurement microphones:
Measurement Microphones — Earthworks Audio