[SOLVED] Control system with open loop

[CataM]

If we desire the output to be the same as the input, why we just simply implement a controller transfer function inverse to the plant's transfer function ? Why we implement it with feedback instead with a simple open loop ?



Why we do not use the 2nd option from the above picture ?

 

[andre_luis]

The second option is just a blind attempt of producing the output according to the known behaviour of the plant, but do not perform any correction of actual value of output, which in real world could differ from the value expected due to many reasons do not predicted on the ideal plant model, which, no matter how good the model, in certain way is nothing more than an approximation of the real plant. The second control method may be easier to implement, but its applicability depends on how much error you can accept in the output.

 

[CataM]

Totally agree. This questions arise because I am always thinking on the ideal model that everything is perfect and is not like that in real life.

My original question is answered.

Talking about the error, if I implement in real life the 2nd option (in open loop) and by experimentation I am able to find a transfer function for the controller that fits very well (say 10% error) on following a step input (output=0.9 or 1.1), if then I change the input to a ramp for example, how would respond the system then? Same error ? increase error ? Need to adjust the controller again for that ramp input especially ?

 

[andre_luis]

if then I change the input to a ramp for example, how would respond the system then? Same error ? increase error ? Need to adjust the controller again for that ramp input especially ?

It is somewhat difficult to determine in the course of time, to which waveform the entry signal was changed unless you're able to concurrently run an algorithm to foresee this shape; This sounds as quite complex for me, perhaps lying in the field of adaptive control, which is a vast subject, but to be honest, I'm even not sure if it is usual to change the whole plant model in runtime, because should be calculated and loaded inner states of variables of each plant for that transition. In short words, seems unfeasible.