I solve an ODE system with the ode15s solver.
The system has very big positiv and negative gradients. If a gradient is negative and the timestep is too big, my solution gets negative.
However, this result is not physicaly correct and my next timestep breaks down because I like to calculate with this negative temperature other values. Normaly the solver "sees" that the timestep was to big and adjusts it self. But in this case (the gradient was to negative) a other function returns an error because the negative temperature and the hole calculatin stops.
How can I modify my code that I don't come to this case? I know that my model is OK, just the automatic time step from the ODE solver is to big.
Thanks for help.