Re: Please help me to scale a Turbo Generator.

Dear A. R. Shaghaghian

I generally scale the inputs with respect to their maximum allowed variations.

For example, for the distillation column the nominal values of the two
inputs are u1=L=2.71 and u2=V=3.71. I have then assumed that the maximum
allowed change in u1 is 1 and in u2 is 1. Thus, the scaling factor is 1 in this
case.

More generally, the scaling method must be selected according to what you want
to use the model for. The scaling I propose for the inputs is reasonable
for controllability analysis, and is often the same as the scaling ("range")
you will find on the instruments. For other purposes, for example, use
of the model in controller design, the scaling you propose may be more
reasonable.

Best regards,
Sigurd Skogestad

>Dear Proff. Skogestad
I am a MSc. student majoring in control engineering. I'm doing my thesis on Control S
tructure Design; But I have some problem with scaling. I have reviewed your distillat
ion column case study. As it seems you have scaled output variables, using maximum ex
pected change in error signal which can be obtain simply by simulation. But it is not
 obvious that how you scale the input variables.
I have some ideas about scaling. I decided to obtain a systematic method to find the
scaling factors. I have used Singular Value Decomposition as a meature to know how mu
ch is the maximum effect of the inputs on the outputs? Also I obtain a RSE (Rational
Scaling Error) factor to determine the truth of the scaling. I used "RGA norm sum" an
d "condition number" to obtain the RSE. I test your distillation column scaling with
 this RSE meature. It has a very good scaling. Let me know your idea about what I hav
e done. Would you please help me to know your ideas about my theories? And would you
please write me how you scaled the input variables of your distallation column? Let m
e to appreciate you because of your kindness.

Sincerely yours

A. R. Shaghaghian
Received on Tue Aug 8 09:10:56 2000