Prev: Computational Geometry functions
Next: Datestr does not give out milliseconds
From: anzp Ng on 13 Apr 2010 05:34
Hi,



I did PID tuning from experimental methods and managed to get my PID constants.

my proportional constant is a negative value, but my derivative and integral constants are all positive.



My professor tells me that the system is not stable theoretically.

however, i have done several testing and it shows that it works fine.



Can anyone verify my problem, is a negative proportional constant PID system really not stable?
From: Yi Cao on 13 Apr 2010 07:00
"anzp Ng" <a_n_z_p(a)hotmail.com> wrote in message <hq1dqd$75o$1(a)fred.mathworks.com>...
> Hi,
>
>
>
> I did PID tuning from experimental methods and managed to get my PID constants.
>
> my proportional constant is a negative value, but my derivative and integral constants are all positive.
>
>
>
> My professor tells me that the system is not stable theoretically.
>
> however, i have done several testing and it shows that it works fine.
>
>
>
> Can anyone verify my problem, is a negative proportional constant PID system really not stable?

To stabilize a system, the proportional gain can be either possitive or negative depending on the system. Consider a tank level control problem, if you use a outlet valve (u), then when level (y) is higher than the setpoint (r), the control error e = r - y is negative, but you need to increase the valve opening (u>0) to reduce the level, therefore, for proportional control, u = ke, where k has to be negative. In another case, if you use the inlet valve to control the level, then k have to be possitive.

HTH
Yi
 | 
Pages: 1
Prev: Computational Geometry functions
Next: Datestr does not give out milliseconds