Question: How do I avoid a singular matrix error in dsolve/numeric/bvp?

I've been trying to numerically solve (and then plot) this set of equations, to reproduce results from a scientific paper (Berro et al., BPJ 2007). Maple tells me that some matrix is singular - any idea what I can do differently?

Any help much appreciated.

> eqn1 := 15*(diff(theta(s), `$`(s, 2))) = N3(s)*cos(theta(s))-N1(s)*sin(theta(s));                                  
> eqn2 := diff(N1(s), s) = 0;    
> eqn3 := diff(N3(s), s) = 0;
> eqn4 := diff(x(s), s) = cos(theta(s));
> eqn5 := diff(z(s), s) = sin(theta(s));
> eqns := {eqn1, eqn2, eqn3, eqn4, eqn5};
> bcns := {x(0) = 0, x(1.5) = 1, z(0) = 0, z(1.5) = 0, theta(0) = 0, theta(1.5) = 0};
> sol := dsolve(`union`(eqns, bcns), [theta(s), N1(s), N3(s), x(s), z(s)], numeric, range = 0 .. 1.5, method = bvp[middefer]);
%;
Error, (in dsolve/numeric/bvp) matrix is singular
>