eval() returns NaN when my matrix gets really big.

Hi

Im making a model of a mass-spring system where I can increase the numbers of masses. I use modal analysis where I transform my eq's into the Q domain and then back again.

My final result is a [n,1] vector where n is the numbers of masses.

When I use:

n = 4         % Number os masses %

T = 1         % Time which is set as "syms T" %

eval(X)       % X is my [n,1] vector %

my result is something like:

ans =

    0.0158
    0.0739
    0.1778
    0.8150

but when I pick n to be more then 4 i just get "NaN" and I think it is because my vector gets too big for eval()..

When I look at my X vector it looks like this:

(227072026347336273211456024361092029598452603947709420093308928000*sin(T))/3496036292367407553353910200384804229923697463840499539488117510161+..... and it continues likes this for 2 full pages.

What to do?

Thank you!

Answer 1

[Alexander]

Please try the following:

double(vpa(evalin(symengine, ['subs(' char(X) ', T=' num2str(T) ')'])))

This does the following steps:

1. Substitute the sym T inside the symbolic engine.
2. Evaluate the result using variable-precision arithmetic.
3. Convert the result to double.

[Walter Roberson]

I think you should probably be able to simplify that to

double(subs(X))

[Adilet Nuratay]

Sorry, I have similar problem.

But I cannot understand the given command. Can anyone clarify it ?

Thank you in advance!

Answer 2

[Mads]

Thanks you two!! Both suggestions worked but Alexanders was for some reason like 20times faster

[Mads]

but I cant plot it like this:

plot(T,double(vpa(evalin(symengine, ['subs(' char(X) ', T=' num2str(T) ')']))))

Plotting plot(T,double(subs(X))) works fine

[Alexander]

Strange, I would have expected Walters suggestion to be much faster but give the NaN answer. Good to know, that both works.

[Walter Roberson]

If T is a vector then you would need 'T=[' num2str(T) ']'

[Alexander]

Yeah, you are right. My suggestion does not work for vectors, I did not think about that.