can be used to guarantee both the closed-loop and
internal stability of the system, under the assumption
of no pole-zero cancellations. Because this study
involves MIMO processes, an extension of the criterion,
the generalized Nyquist criterion (Maciejowski,
1989) is used. The generalized Nyquist diagram is
formed by drawing the characteristic loci of the openloop
system over the whole frequency range, namely
detðGð j!ÞKð j!Þ ð!ÞIÞ ¼ 0, 8! 2 ½1,1: ð64Þ
In order for the closed-loop system to be stable, it is
required that there are as many encirclements of the
point 1 as there are unstable (Smith-McMillan)-
poles (Maciejowski, 1989). The Nyquist diagrams of
the different controlled processes with different controllers
are given in Figures 16–19.
Several trends are clearly visible in the above results.
First of all, the SISO systems have larger stability margins
compared with those of the MIMO systems, with a
few exceptions. This is an expected result arising from
the properties of linear systems. Another clear trend is
the significant decrease in the stability margins for the
IHC, SFSCF and FSCF methods when multiple disturbance
tones are present. Overall, the IDC method
yields the best margins in the multiple input processes,