Is a PID controller suitable or should I use another algorithm?

Question

Electronics

Good afternoon!
The background is as follows:
There is an engine on the old equipment, it is necessary to maintain its speed, but the load on the engine is not uniform - either thick or empty, and it can change 10 times per second.
The engine is controlled by a transistor assembly, there is a sensor that measures the revolutions and outputs a logical 1 - if there are many revolutions and 0 if there are few revolutions (that is, no intermediate values, and even more pulses per revolution, simply or a lot or a little).
It was controlled simply - there are a lot of revolutions, transistors close, there are few revolutions - transistors open to the full.
A sort of PWM with feedback. But due to the non-instantaneous reaction of the sensor and the spinning of the engine, the revolutions "floated" too much - you need to try to keep them somehow more accurately.
There would be no question if it were possible to put an inverter on the engine or replace the sensor - no way.
Those. we believe that the engine is controlled by either logical 1 or 0. We also believe that the engine speed cannot be counted otherwise. You have to start from what you have.
The idea of ​​a PID algorithm immediately came to mind, but is it applicable under these conditions (I repeat that the load on the engine can change up to 10 times per second, and maybe more often in the future)?
Maybe there is a better algorithm?

Answer 1

Your problem is not at the regulation level, but at the level of the initial engineering design of the mechanism.
You have nothing, from the word at all.
If you want to get some margin in response time to load changes and without changing the existing control mechanism, put the flywheel on a heavier weight.
For any other options, in principle, there is little data, and not from the category "it can change 10 times" but specific calculations for the load, timing diagrams, etc.

Answer 2

with relay control, the inertia of the "slave" plays a role and if it is insufficient, this cannot be solved in any way, overshoot (rpm oscillations) cannot be avoided.
You can try to apply the pid if you take the transition time 0--1--0 from the sensor as the input action. for idling, you can estimate the average time and use it as a zero point. then try to pick up a law with a real load for this. will work with approximately similar "time-distributed" loads.
engineering here is 0 and success will be determined only by which of the crutches is suitable.