At high load most of the input crank force is applied to the inertia wheel through the differential lever (see saw). The inertia wheel starts spinning and gains inertia but once up to speed the inertial force adds to the input crank force and they push/pull together. This is the torque multiplication by adding the forces. We can also see all the input energy is conserved as inertia until the output shaft moves. So the total force (input + inertia) keeps increasing until the output shaft moves.
This is the bit that doesn't work for me. Firstly, the inertia wheel doesn't start spinning. It is forced to oscillate back and forth, so its stored energy reaches zero (it is stationary) twice per revolution of the input crank.
Secondly, the stored energy in the inertia wheel cannot "add" to the input crank force. Or rather,
for each revolution of the input crank, it adds half the time and subtracts half the time, so there is no nett addition.
I'll explain. Begin by assuming the input crank is turning, but the vehicle is stationary, so the drive rods and ratchet are stationary. Therefore
the input motion from the input crank moves only the inertia wheel (ie the vehicle is "idling", as described in the article).
Referring to the first diagram in the article, imagine the input crank (driven by the engine crankshaft) starts at the '3' position on a clock face, and turns clockwise:
Positions 3 to 6 of the input crank: the engine accelerates the inertia wheel clockwise. The torque required to do that acts to oppose the torque from the input crank, and thus is subtracted from the torque available to move the car. Energy is added to, and stored in, the inertia wheel.
6 to 9: the inertia wheel is decelerated to a standstill. The torque required to decelerate it acts in the same direction as the torque from the input crank, so is added to the torque available to move the car. Energy is subtracted from the inertia wheel until it reaches zero when the crank reaches '9' and the inertia wheel is stationary.
9 to 12: the input crank accelerates the inertia wheel in the opposite (anticlockwise) direction, storing energy in the inertia wheel. The torque required to accelerate the inertia wheel is subtracted from that available to drive the car.
12 to 3: the inertia wheel is decelerated to a standstill again. The torque required to do that adds to the torque from the input crank.
In summary, the inertia wheel is accelerated from standstill, and then decelerated back to a standstill, twice per revolution of the input crank. When being accelerated it stores energy from the engine; when being decelerated it releases the stored energy. As the input crank (driven by the engine) rotates, the inertia wheel impedes it for 50% of the time, and assists it for 50% of the time.
There is no mechanism by which the inertia wheel can store energy for any longer than half an input crank rotation.
There is no mechanism by which the inertia wheel can provide continuous torque addition to the torque from the input crank (and thus from the engine).