2 of my inertial drive designs.

[pinobot]

Here are 2 of my designs. 
btw they don't work but i just love them.

[pinobot]

Ok, both "work" by changing the radius or speed to increase centripetal force.   
Check this calculator: http://pacer.calpoly.edu/tri/pacer/gforce.html

If you look at the first design then the thickness (and weight) of the top part of the rubber band is one quarter that of the bottom band but the speed is twice as high.
Calculation:
speed top 1000 rpm
radius 100 mm
G-force = 112 g

speed bottom 250 rpm
radius 100 mm
G-force 7 g

So the difference in centripetal force is 16x but the difference in weight (rubber band) is only 4x.

[pinobot]

A more worked out design.

[pinobot]

Ok, very quiet here, so i'll do all the talking. 
This 'works' because the weight is shifted from right to left.
To be honest it's the only design i haven't been able to falsify, i just can't get my head around it .

[broli]

In your last design, why not just cut to the chase and have a pump circulating water in a tube.

However there's a flaw in that specific case.

The other concept can also be reduced to a tube with water. The interesting part about this is that velocity is linearly proportional to area and so is the mass in flow rate of liquids. But velocity and mass aren't linear in the force equation for centripetal motion.