Similarly to this thread NSolve gives additional solutions that don't satisfy the equations!
NSolve returns "spurious" solutions, even increasing the working precision
fsys={1 - Cos[(11 \[Pi])/45] - (3 Cos[psi1])/4 - 2/3 Sin[phi + \[Pi]/6] == 0, h + 2/3 Cos[phi + \[Pi]/6] + Sin[(11 \[Pi])/45] - (3 Sin[psi1])/4 == 0, -0.0436952 - (3 Cos[psi2])/4 - 2/3 Sin[phi - \[Pi]/6] == 0,2/5 + h + 2/3 Cos[phi - \[Pi]/6] + Sin[(17 \[Pi])/180] - (3 Sin[psi2])/4 == 0}
NSolve[N[fsys], {phi, h, psi1, psi2}, WorkingPrecision -> 100]
The second solution returned is
{h -> -1.99515, psi2 -> -1.22627, psi1 -> 1.68803, phi -> 0.0618559,
Sin[psi1] -> -0.993136, Cos[psi1] -> -0.116967,
Sin[psi2] -> -0.941237, Cos[psi2] -> 0.337748, Sin[phi] -> 0.0618165,
Cos[phi] -> 0.998088}, {h -> -1.99515, psi2 -> 1.22627,
psi1 -> -1.68803, phi -> 0.0618559, Sin[psi1] -> -0.993136,
Cos[psi1] -> -0.116967, Sin[psi2] -> -0.941237,
Cos[psi2] -> 0.337748, Sin[phi] -> 0.0618165, Cos[phi] -> 0.998088}
,
I do not understand why both psi1 and Cos[psi1] are returned, with Sin[1.68803] actually not equal to 0.993136 (the sign is different)
Sin[...], Cos[...]bits (on version 9). – b.gatessucks Dec 17 '12 at 6:29NSolveis not the right tool as stated in the documentation and the warnings. Why not usingFindRoot? – b.gatessucks Dec 17 '12 at 6:34