home *** CD-ROM | disk | FTP | other *** search
- 10 being a reasonable starting point.
- Now, we return to the network editor and retrieve (or
- rewrite) the lowpass filter that we analyzed earlier. Then,
- we come back to the optimization section. When selecting the
- elements for inclusion in the optimization table, it is not
- possible to select certain elements such as OPN, SHT, ONE,
- TWO, IMP, and IMS, and none of the interconnections or trans-
- formations, as variables. Doing so will illicit an error mes-
- sage from the program. Also, the parameter number selected
- is checked against the number of parameters for that element.
- Selecting inappropriate parameter numbers will also result in
- an error message.
- In this example, we will select all three elements as
- variables for optimization. We do this by selecting the ADD
- option from the menu, entering the element number, and then
- the parameter number, which in all three cases here is one.
- Our listing should look like this:
-
- ELEMENT PARAMETER GRADIENT
- TYPE NUM NUM VALUE
- 1 CPP 1 1 1.967
- 2 INS 2 1 12.875
- 3 CPP 3 1 6.366
-
- For the sake of curiosity, let's do a sensitivity analy-
- sis of our network. Upon return, we observe the following:
-
- ELEMENT PARAMETER GRADIENT
- TYPE NUM NUM VALUE
- 1 CPP 1 1 1.967 -163.439
- 2 INS 2 1 12.875 9.968
- 3 CPP 3 1 6.366 72.048
-
- ITERATION #0 ERROR = 125.9
-
- We can now proceed right into the optimization procedure
- by selecting option 5. The program now asks if hard copies of
- the intermediate results are desired. This is optional. Next,
- we are asked for the number of random moves desired. Since
- this is a very simple problem, we will not require any random
- searches, so we enter 0. The final result of all this is:
-
- ELEMENT PARAMETER GRADIENT
- TYPE NUM NUM VALUE
- 1 CPP 1 1 3.341 -3.631
- 2 INS 2 1 13.646 0.259
- 3 CPP 3 1 6.007 0.374
-
- ITERATION #38 ERROR = 0.24
-
- This is quite an improvement. Notice that the first ele-
- ment made the most dramatic amount of change, whereas the sec-
-
-
-
- - 31 -