May i know how i can assure that time step and solver used for a given nonlinear model such as Van der pol or nonlinear model having term such as x^4*xdot are accurate enough with respect to error flow in solution. Any class note and article with examples may be more useful.
Error is with respect to true solution. If the equations have a closed form solution derived from first principles, you may plot difference between simulation and true solution. If that is not possible, the only way is to compare simulation data with other solvers in Matlab or other algorithms/softwares or an implementation of state-of-the-art techniques.
Firstly you can try raising the tolerances and see if the simulation is significantly different. Or, try the comparison with a different (but not inappropriate) solver.