Rockwell Automation Arena OptQuest Users Guide - Manual (Page 38)

A RENA O PT Q UEST U SER ’ S G UIDE
34
• •
•
•
•
Replications per simulation
You may specify the number of replications per simulation. This is the number of times
that the model will be run in every simulation. There are two choices:
Solutions Log
During an optimization, information is automatically logged to a file. The information
includes each solution that is tried, the value of the objective function, and the value of
constraints. You may direct where the log file should be kept by selecting the Browse
button and identifying the file name and location. The default file is OptQuest.log in your
Temp directory.
The option settings are automatically saved when you navigate to a different node or start
an optimization.
Suggested Solutions
Suggested Solutions are solutions that you believe may be close to the optimal answer.
Suggested solutions can shorten the time it takes to find an optimal solution. These
suggested solutions will be the first solutions evaluated when the optimization is run.
After the suggested solutions have been evaluated, OptQuest will begin its search for the
best solution. You can input your own suggested solutions, or you can select solutions
from previous optimizations and save them as suggested solutions.
Use a fixed number of
replications When this option is chosen, OptQuest instructs Arena to run the
indicated number of replications per simulation.
Vary the number of
replications When this option is chosen, OptQuest uses the given numbers as bounds
on the number of replications per simulation. This option allows
OptQuest to test for the statistical significance between the mean of the
objective function in the current simulation (the “current mean value”)
and the best value found in previous simulations (the “best value”). The
purpose of this test is to weed out inferior solutions without wasting too
much time on them.
The “inferior solution” test constructs a 95% confidence interval around
the current mean value, then checks to see if the best value falls within
this interval. If it does not, then the solution must be inferior and the
simulation is ended with no further replications. If the best value does
fall within the interval, then Arena runs another replication and checks
again. As long as the interval continues to include the best value, Arena
will continue to run replications up to the maximum number specified.