|
|
||
|
|
|
|
This is Fig. 4 from the 1990 GENOPT paper. This slide shows the decision variables LENGTH and WIDTH versus design iterations during a single execution of the GENOPT processor, OPTIMIZE.
Ordinarily one would execute OPTIMIZE several times in succession in order to obtain a local optimum design. For this simple case a local optimum design was obtained with only a single execution of OPTIMIZE.
In order to find a "global" optimum design one would execute the GENOPT processor called "SUPEROPT", which automatically executes "OPTIMIZE"/AUTOCHANGE many, many times. AUTOCHANGE is a GENOPT processor that automatically finds a new point in design space from which to restart optimization.. SUPEROPT did not exist when this 1990 GENOPT paper was written.
The word, "global", in the previous paragraph is in quotes because GENOPT does not rigorously find the global optimum design, but attempts to find a design the objective of which (e.g. objective = weight) is likely to be close to that of the global optimum design.
There now exists a GENOPT processor called "SUPERDUPEROPT", which automatically executes SUPEROPT many times in succession.
Page 28 / 190