Fourier analysis of cam profile

[ivymike1031]

(transplanted debate)

Originally posted by ivymike1031
... whether it was more appropriate to look just at the lift event, or to look at a whole cam cycle. Both methods have shown significant components up to, and sometimes beyond, 10th order. My current preference is to look at the lift events alone, w/o the flat parts in between.

Originally posted by fritz_269

If you don't mind, I'd like you to revisit that debate. It seems to me that, a priori, you should consider the entire circular profile. The Fourier transform is based on the idea of an infintely periodic signal. If the signal isn't infinite, you have to use a circular approximation and usually windowing functions to make it appear infinitely periodic (or you could just use finite time-base wavelet transforms).

I was just thinking about how cool it was that you actually wouldn't have to make any of those approximations by forcing periodicity into your calculation, becase it was just natural since the cam is rotating.

Can you give me any intuition as to why you wouldn't want include the base circle in your transform? It's completely non-obvious to me.

[ivymike1031]

My reasoning was anything but rigorous - I had observed that the vibrations I was interested in were excited very rapidly (showing up during the opening flank), and decayed pretty quickly after valve closure. Because of this, I figured that I'd just focus on the lift events themselves, without the period of decay, as though the valve was getting "bopped" repeatedly. It may have been a wrong-headed approach; that wouldn't surprise me at all. I wasn't entirely sure what I wanted to look for when I did it, and I didn't learn anything useful by doing it.

If I emailed you a "typical" lift profile, velocity profile, and acceleration profile, would you be interested in manipulating that data to see what you can learn about it? I'd certainly appreciate some input, especially since it sounds like you've got more experience with frequency-domain analyses than I have.

[ivymike1031]

I guess I should add (although it was already implied) that my present set of tools for valvetrain analysis are time-domain tools. I'm waiting for that elusive "free time" so that I can try to write some code to help me perform frequency-domain valvetrain analyses.

[Someguy]

This may help save a little code writing time:

http://www.fftw.org/

Anyway, I hope this topic sticks around for a bit. Work just "blew up", so I'm going to be busy, busy, busy...

[higgimonster]

ivymike, excuse my ignorance but I was wondering how you determine the cam profile. Is it based off of a continuos function or data analysis.
Basically the reason for the question is that I am working on a very basic simulation of a cam and valve/spring mechanism and I don't really know how to model the cam. Any help would be apperciated.

[ivymike1031]

My valvetrain dynamic simulation code offers several options for describing the cam lobe geometry, but they all basically come down to giving the base circle radius & cam "lift" versus cam angle (above the base circle). It can account for the effects of roller/follower radius, etc., on the "follower lift."

How to represent the cam profile for your software is a whole other question... IM (or PM) me and perhaps I can help you sort it out.

[ivymike1031]

I found a book about cam design which describes several fourier-series cam profiles, but it's coming from a "design a profile" end, not the "analyze an existing profile" end.

I've had another project recently that's required me to look into the problem further.. I'm starting to think that you were right (fritz) about representing the whole cam cycle rather than just the lift period.

I've also found some software that will let me quickly calculate mode shapes for a valvetrain (as soon as I learn how to use it). It's called Easy5, and if I'm not mistaken, it was developed (initially) by Boeing. I haven't got a license for it yet, but that's coming. I'll let you know how it goes when I get that.


Back to your original question about the HLAs, fritz, I've received confirmation from a much more experienced designer that HLAs are almost always bad for high-speed performance. I'm currently working on a valvetrain where the HLA is a major PITA. It's by far the most flexible component in the system. :P

anyway... I'll quit talking to myself now.

[Someguy]

Originally posted by ivymike1031

anyway... I'll quit talking to myself now.

I'm listening. Just been way to busy to post much. I have some friends that have vested interest in HLA at high speeds. 1,300 rwhp @ 9500 RPMs vested. Still an interesting discussion even if it is a monologue.