HansH said:
as I understand An inertial frame of reference is one in which the motion of a particle not subject to forces is in a straight line at constant speed
This definition is correct, but note that its actual meaning is different in relativity than in Newtonian physics. In Newtonian physics gravity is a force, but in relativity it is not. So in Newtonian physics a frame in which you, standing on the surface of the Earth, are at rest can be treated as an inertial frame, since the net force on you is zero (downward force of gravity exactly balances upward force of the Earth's surface pushing on you). But in relativity this is not the case; this frame is
not inertial, because gravity is not a force so the only force on you is the Earth's surface pushing upward. In relativity, an inertial frame in your vicinity would be one in which a freely falling object, such as a rock dropping in vacuum, would be at rest.
HansH said:
one rocket accelerates for a certain amount of time with say 1g. so this gives a force of 1 Newton per kg mass at that mass during this amount of time. then after some time the same rocket accelerates for (I assume) a same amount of time with -1g
Time defined by what? By the rocket's clock? That's the simplest answer, but note that it is not the only possible answer, and other answers result in a
different scenario being specified. (That would not be the case in Newtonian physics, which has absolute time, but there is no absolute time in relativity.)
HansH said:
after that there is no speed difference between the 2 rockets but only a certain distance in between. now they come together again by slow acelleration en de-aceleration and after that compare their watches. This means there is now a time difference between their watches related to the amount of time and acceleration that one of them acellerated
This is true, but it does not mean that acceleration "causes a difference in the rate of time flow". The fact that in this particular scenario, it happens to be possible to use the acceleration to compute the difference between the watches does not mean it is
always possible to do that.
The only general rule that works is the geometric one: the difference between the watches is due to the difference in arc length along the two worldlines between the two meeting events. The particular scenario you describe just happens to be set up so that the difference in arc lengths, and hence the difference between the watches, has a definite relationship to the acceleration of one of the objects (and the time by that object's clock that the acceleration lasts). But that won't always be true.