solve() does not give all roots and sometimes crashes the symbolic toolbox

Question

Jori on 7 Nov 2012

0
Link

Direct link to this question

https://au.mathworks.com/matlabcentral/answers/53044-solve-does-not-give-all-roots-and-sometimes-crashes-the-symbolic-toolbox

I asked this question earlier ( see http://www.mathworks.nl/matlabcentral/answers/52958-solve-gives-incorrect-answer) and it seemed to be solved, but it keeps on returning. In hope of new answers I opened a new thread.

----------------------------

COMPUTER SPECIFICATIONS

----------------------------

Windows 7 Enterprise SP 1
64-bit OS
Intel Core i7-2670QM CPU 2.2 GHz
8,00 GB RAM

----------------------------

QUESTION

----------------------------

I have a function

>> clear all
>> syms x
>> G = (x - (23*x^2)/2 + 1/2)/(10*(x^5)^(1/2))

for which I want to find intersections with certain y-values. Let us first try to find the intersection with y = -1.

>> m = 0; s = 4; double(solve(G == (2*m-s)/s,'MaxDegree',4))
ans =
      1.0000
      0.4388

Which is correct. If I now try to find the intersection at y = -0.5, it gives me

>> m = 1; s = 4; double(solve(G == (2*m-s)/s,'MaxDegree',4))
ans =
      5.0935

It leaves out one of the intersections, which should be around 0.3 after inspection of ezplot(G,[0 1]).

Now, if I try to find the intersection at y = 0.5, it gives me

>> m = 3; s = 4; solve(G == (2*m-s)/s,'MaxDegree',4)
Error using mupadmex
Internal error with symbolic engine. Please quit and
restart MATLAB.
Error in mupadengine/evalin (line 97)
                [res,status] = mupadmex(statement);
Error in mupadengine/feval (line 150)
            [S,err] = evalin(engine,stmt);
Error in solve (line 160)
  sol = eng.feval('symobj::solvefull',eqns,vars);

It gets from bad to worse. Not using the ...,'MaxDegree',4) does not change anything. The answer to my previous question on this topic showed me that using that option should work, which it did for one instance, but as you can see, the problem has returned.

Thanks in advance, Jori