Hi Jan,
Quick look at hte model shows that your 'u' signal hovers around '0' at around time 2.79. When that happens, your F_lat subsystem divides by 'u' getting close to a divide by zero condition, causing the calculations to go haywire.
Do you have the equation of motions written out by hand? Perhaps you can paste them into here and someone can double check that there were no modeling mistakes.
~Zack