A simple algebraic calculation does not work properly!

Hi friends!

I have been strugling to tell matab to do a simple calculation for me but I failed, unfortunately. I explain my problem using the following

simple problem (although my expressions are extremely large and more complicated). Consider the following function f:

syms dt y y0 muY(y)
f=-(4*muY(y)-7677*dt*diff(muY(y),y,y)+4434*y*muY(y)^2-4212*y0*muY(y)^2 +332*y^2*diff(muY(y),y,y))/(48*dt^(3/2));

Now, I would like to approximate f by rounding the coefficients with any decimal accuracy I wish. The command vpa(f,4) gives me

-(0.02083*(4.0*muY(y) - 7677.0*dt*diff(muY(y), y, y) + 4434.0*y*muY(y)^2 - 4212.0*y0*muY(y)^2 + 332.0*y^2*diff(muY(y), y, y)))/dt^(3/2)

which is not what I like. I wish to get a solution as bellow:

(-0.0833*muY(y)+159.9375*dt*diff(muY(y), y, y)-92.3750*y*muY(y)^2+87.7500*y0*muY(y)^2-6.9167*y^2*diff(muY(y),y,y))/dt^(3/2)

where each coefficient is calculated using the command round(-----,4). So, in this answer each coefficient is rounded by 4 decimal degit. I

do not want to use vpa since there is no way to tell vpa to consider 4 decimal degits after the decimal point. Unfortunately, vpa(----,n) only consideres n decimal degits in total. On the other hand, I cannot use the 'round' command here since I have a simbolic expression. If I use the command vpa(f,4) then I get the following undesirable answer, unfortunately:

0.0001*round(-(625*(4*muY(y) - 7677*dt*diff(muY(y), y, y) + 4434*y*muY(y)^2 - 4212*y0*muY(y)^2 + 332*y^2*diff(muY(y), y, y)))/(3*dt^(3/2)))

Any idea?

Thanks for your help in advance!

Babak

Answer 1

[Dyuman Joshi]

syms dt y y0 muY(y)
f=-(4*muY(y)-7677*dt*diff(muY(y),y,y)+4434*y*muY(y)^2-4212*y0*muY(y)^2 +332*y^2*diff(muY(y),y,y))/(48*dt^(3/2));
vpa(simplify(f,3),7) %7 because the most amount of significant digits (159.9375)

ans = 

Answer 2

[Walter Roberson]

syms dt y y0 muY(y)
f = -(4*muY(y)-7677*dt*diff(muY(y),y,y)+4434*y*muY(y)^2-4212*y0*muY(y)^2 +332*y^2*diff(muY(y),y,y))/(48*dt^(3/2))

f = 

newF = mapSymType(f, 'constant', @(v) round(v,4))

newF = 

If you need the powers to be preserved as rationals, then that is possible, but more of a nuisance.