In this scenario the effect of shortening the experimental time between each step test on an SDOE design is studied.
Note, however, that the SDOE efficiency at 3[tau] is still much higher than the PRSD at the recommended [t.
In the first scenario, a Box-Behnken design (46 runs) was used to generate the input sequence of the SDOE.
Table 6 shows the percent efficiency results and, as in the two factors case, it shows that SDOE is more efficient than PRSD.
The next scenario investigates the effect of reducing the SDOE experimental time on the efficiency.
In this scenario, the times between SDOE step tests were 15 (3[tau]), 20 (4[tau]) and 25 (5[tau]) time intervals.
As in the previous examples, SDOE is compared to PRSD.
Table 9 shows the results for the percent efficiency and, as in the two and five factor cases, it shows that SDOE is highly more efficient than PRSD.
Since PRSD tends to require many input changes, it will be considerably less rich in ultimate response information as compare to SDOE.
For a given situation, as the ratio of dynamic parameters to steady state parameters increases, the efficiency of PRSD to SDOE will increase but we do not believe that it will go above 100% because step tests are adequate to provide the necessary dynamic information and because of the information limitations of PRSD for ultimate response behaviour as mentioned above.
inputs), a PRSD could provide the superior design over SDOE since the SDOE would be limited in capturing dynamic information and the PRSD would likely capture better dynamic information over the spectrum of dynamic behaviour.
i] model parameters [tau] dynamic parameter v (t) true output variable Abbreviations CSTR Continuous Stirred Tank Reactor H-BEST Hammerstein Block-Oriented Exact Solution Technique MIMO Multiple-Input, Multiple-Output PRS Pseudo Random Sequence PRBS Pseudo Random Binary Sequence PRSD Pseudo Random Sequence Design SDOE Statistical Design of Experiments SISO Single-Input, Single-Output