Matlab suspension modeling

[nora.rh]

Thank you for your reaction!

Almost all of the analysis methods you have suggested are inappropriate for a non linear model.

Does than mean at least 1 of them is appropriate?

The easy way to do it is to apply a sine sweep excitation to the contact patch.

Doesn't applying a chirp signal as ground position count for this? Or do you mean defining a Frestimate input in this form? At the moment for the Frestimate method, I use frest.Sinestream(...).

[zeroracing]

I'm looking at getting matlab for home use, I am not a professional consultant and hobby level racer. Any advice on which add-ons I should get, at $ 45 each they could add up quick just buying for no real reason. Looking at doing suspension modeling, on a mac computer.

[Tim.Wright]

For all the analysis work I do I have the following setup:

Matlab
Simulink
Signal processing toolbox (required for decent filters)

Tim

[Greg Locock]

The ones I use mostly are signal processing, statistics, image processing, optimisation. I'd like to do more simulink, and there are some toolboxes there that look fun.

[Greg Locock]

@nora - I think you'll find a conventional chirp has too high a sweep rate, i wouldn't use much more than 60 hz/minute.

The point is that by driving the system with a sinewave you are linearising it by forcing it to respond at one frequency only (you ignore the harmonics).

[Jersey Tom]

at $ 45 each they could add up quick just buying for no real reason

Given that in the real world they add up at $1000/each that's a steal of steals

Depends what you wanna do, man. Signal processing and optimization are definitely pretty hot though.

[nora.rh]

Greg, thanks again for your answer.

With a sinewave input with an amplitude of 30 Hz all the methods I use still don't generate a decent frequency response. Are you familiar with any of these methods, which do you recommend I use?

[Greg Locock]

That's not an amplitude that is a single frequency.

My experience with modal analysis is primarily based in real world testing, not models. I have used tf and bode, but they didn't seem very useful to me. When i wish to do a frequency response analysis i use fft etc, which is more time consuming to write, and probably less efficient and less accurate, but that is the way I have done it for 32 years.

[DaveW]

Greg: I am sure that budding engineers would be interested in Octave (which you have mentioned in previous posts).

I would add SkyLab, a well respected alternative to MatLab. There are plenty of "tooboxes" listed on SourceForge...

Both are free (I dislike the MathWorks business model...)

p.s. I agree with you about "chirp" inputs. Tip: generate a swept frequency sine wave, pass it though a single pole I Hz low pass filter, and scale to get a swept frequency input with (roughly) constant peak velocity....

[Greg Locock]

That'd be Scilab, not the space station. If anyone is wondering, as a standalone system Scilab is probably better than Octave, BUT its syntax is slightly different to matlab, and trying to work in them both is infuriating.

Octave is pretty much code compatible with Matlab.

There's also Freemat which I have just picked up at the portable apps site, which means it'll run just fine from a USB key, no install on a host required. Sadly it seems more of a work in progress than a finished system, it broke on every matlab/octave script I tried to run.

[DaveW]

That'd be Scilab, not the space station. If anyone is wondering,

Apologies, a slip of the brain (whats left of it, anyway)

SourceForge is wonderful. I recently had some algebraic manipulation to do & discovered that I no longer had a working copy of Mathematica. I visited Wolfram's web site to find, I thought, that I could now only hire it. So I consulted SourceForce & discovered wxMaxima. It formulated, and then solved a fairly complicated transfer function for me. At one point it had:

gfactor((Ki*(−Mc*w^2+%i*Cs*w+Kc))/(Mc*Ms*w^4−%i*Cs*Ms*w^3−Kc*Ms*w^2−Ks*Mc*w^2−Ki*Mc*w^2−Kc*Mc*w^2+%i*Cs*Ks*w+%i*Cs*Ki*w+%i*Cs*Kc*w+Kc*Ks+Kc*Ki)−(Ki*(−Mc*w^2+%i*Cs*w+Kc)*(Mc*Ms*w^4−%i*Cs*Ms*w^3−Kc*Ms*w^2−Ks*Mc*w^2−Kc*Mc*w^2+%i*Cs*Ks*w+%i*Cs*Kc*w+Kc*Ks))/(Mc*Ms*w^4−%i*Cs*Ms*w^3−Kc*Ms*w^2−Ks*Mc*w^2−Ki*Mc*w^2−Kc*Mc*w^2+%i*Cs*Ks*w+%i*Cs*Ki*w+%i*Cs*Kc*w+Kc*Ks+Kc*Ki)^2);

The good bit is that it was right, and the resulting routine actually works...

[dynatune]

Depends on what you want to simulate. A "ride" model or a suspension kinematics model. If it is the latter it can be done extremely well in MS EXCEL as there is this tool on the market called Dynatune Suspension Design Module on http://www.dynatune-xl.com

[djhache]

Sorry to re-resurrect an ancient post, I'm working on the same type of Matlab modelisation and I am having a ton of trouble exporting a linear model of my Simscape model. I've based my work on this http://fr.mathworks.com/matlabcentral/f ... h-simscape model I found where I've added a sensor for what I think is the CPL. But just getting a linear model of the body displacement based on road inputs is difficult. The model works fine but if I want to do frequancy analysis I'm SOL, does anyone have any ideas to help me (and others maybe?) along?