3.1. Frequency-shaped cost functionals
This method was originally presented by Gupta (1980)
then discussed by Hall and Wereley (1989) and later
revised by Sievers et al. (1991). The control goal in
this approach is the suppression of persistent sinusoidal
disturbances at the process output by applying optimal
state feedback. In the original work, the states of the
process were assumed to be measurable; this, however,
is seldom the case in practice. Hence, two different
approaches can be taken: either the process states are
estimated by an estimator or the feedback gains related
to the process states are neglected. In the sequel, these
methods will be referred to as FSCF and SFSCF,
respectively. Both of the methods have some drawbacks
and require somewhat questionable assumptions to be
made. Namely, the use of state observer in the presence
of deterministic and persistent disturbance, which is not
included in the process dynamics results in biased state
estimates. However, the inclusion of the disturbance
dynamics in the model would make the use of the filtering
approaches pointless in the first place. Again, the
neglecting of some feedback terms is a rather questionable
and risky approach, as the optimal control performance
and the closed-loop stability are guaranteed
only for the full state feedback. In essence, the exclusion
of the process states from the feedback concentrates
the control effort to the frequency range of the
disturbances, implicitly specified by the applied filter.
Hence, the process dynamics at the outlying frequencies
are neglected, resulting in similar stability problems as
perceived with the IHC approach discussed in Section
3.3. Regardless of these issues, controllers using both of
the approaches are derived.
3.1. Frequency-shaped cost functionalsThis method was originally presented by Gupta (1980)then discussed by Hall and Wereley (1989) and laterrevised by Sievers et al. (1991). The control goal inthis approach is the suppression of persistent sinusoidaldisturbances at the process output by applying optimalstate feedback. In the original work, the states of theprocess were assumed to be measurable; this, however,is seldom the case in practice. Hence, two differentapproaches can be taken: either the process states areestimated by an estimator or the feedback gains relatedto the process states are neglected. In the sequel, thesemethods will be referred to as FSCF and SFSCF,respectively. Both of the methods have some drawbacksand require somewhat questionable assumptions to bemade. Namely, the use of state observer in the presenceof deterministic and persistent disturbance, which is notincluded in the process dynamics results in biased stateestimates. However, the inclusion of the disturbancedynamics in the model would make the use of the filteringapproaches pointless in the first place. Again, theneglecting of some feedback terms is a rather questionableand risky approach, as the optimal control performanceand the closed-loop stability are guaranteedonly for the full state feedback. In essence, the exclusionof the process states from the feedback concentratesthe control effort to the frequency range of thedisturbances, implicitly specified by the applied filter.Hence, the process dynamics at the outlying frequenciesare neglected, resulting in similar stability problems asperceived with the IHC approach discussed in Section3.3. Regardless of these issues, controllers using both ofthe approaches are derived.
การแปล กรุณารอสักครู่..
