Link to Index Page

Evolurion of the objective during an execution of the PANDA2 processor, SUPEROPT

Fig. 43 from the 1996 enhancements paper. This slide shows optimization during a single execution of SUPEROPT of a hydrostatically compressed isogrid and ring-stiffened perfect cylindrical shell.

These results are from the “two-materials” model.

Each “spike” in the plot corresponds to a new “starting” design obtained randomly by the PANDA2 processor called AUTOCHANGE. Five PANDAOPTs per AUTOCHANGE were specified by the PANDA2 user for the execution of SUPEROPT.

In 1996 a SUPEROPT execution generated about 270 design iterations. Since then the number of design iterations in a SUPEROPT execution has been increased from about 270 to about 470. Also, a new processor, SUPERDUPEROPT, has been written which automatically executes SUPEROPT in a loop.

The purpose of SUPEROPT (and SUPERDUPEROPT) is to find a "global" optimum design. Notice that during the execution of SUPEROPT the design converges to local optima with different values of the objective (weight).

Page 6 / 38