inverse_laplace_transform produces incorrect result (compared to WolframAlpha) about sympy

[oscarbenjamin]

CC @hanspi42

[crgfn]

Another example:

from sympy import symbols, inverse_laplace_transform, expand
s, t = symbols('s t')
Fs = (20000.0*s**2 + 1600.0*s + 30.0)/(100000.0*s**4 + 48000.0*s**3 + 6930.0*s**2 + 281.0*s + 3.0)
Ft = inverse_laplace_transform(Fs, s, t)
print(expand(Ft))

Sympy output:
-7.52218450407113*exp(-0.223493475978327*t)*Heaviside(t) + 22.1739130434783*exp(-0.2*t)*Heaviside(t) + 0.0299800831746754*exp(-0.0395265172332464*t)*Heaviside(t) + 0.100900073070386*exp(-0.016980006788427*t)*Heaviside(t)

WolframAlpha (difference is in coefficient of 2nd exponential):

InverseLaplaceTransform[(20000.0 s^2 + 1600.0 s + 30.)/(100000.0 s^4 + 48000.0 s^3 + 6930.0 s^2 + 281. s + 3.), s, t]

-7.52218/E^(0.223493 t) + 7.3913/E^(0.2 t) + 0.0299801/E^(0.0395265 t) + 0.1009/E^(0.01698 t)

[oscarbenjamin]

WolframAlpha produces 10 (instead of ~30) for first term.

Yes, that seems correct:

In [8]: apart(nsimplify(Fs), s)
Out[8]: 
      ⎛       2              ⎞       
   20⋅⎝10000⋅s  + 1800⋅s + 53⎠     10
- ────────────────────────────── + ──
         3         2               s 
  20000⋅s  + 5600⋅s  + 266⋅s + 3  

[hanspi42]

Thanks for reporting this!

This is the same as #25352, it has already been fixed in PR #25642, so it will be correct in SymPy 1.12.1.