This component can be viewed as another type of multiple-run device, similar to the Multiple Run component. The major difference is that the Optimum Run component actually searches for (and converges to) the optimum design parameters. The Optimum Run approach can result in a huge time savings by drastically reducing the amount of runs required, as well as improving accuracy by converging to the exact design point.
An option is provided for the selection of different optimization techniques:
Golden Section: Suitable for optimization of a single REAL variable. The golden section, also known as the divine proportion, golden mean, or golden ratio, is a number often encountered when taking the ratios of distances in simple geometric figures.
Simplex: Suitable for optimization of several REAL (up to 20) variables. This method runs along polytope edges of the visualization solid to find the best answer.
Hooke-Jeeves: Suitable for optimization of several REAL variables.
Genetic Algorithm: Suitable for optimization of several REAL/INTEGER/LOGICAL variables. An adaptive stochastic optimization algorithm involving search and optimization that was first used by John Holland. Holland created an electronic organism as a binary string ('chromosome'), and then used genetic and evolutionary principles of fitness-proportionate selection for reproduction (including random crossover and mutation) to search enormous solution spaces efficiently.
Regardless of the optimization technique chosen, a user defined Objective Function (OF) is required as an input signal. From the value of this function, the optimization algorithm will determine a new set of output parameters each run and compare the difference in OF value to the input Tolerance. If the change in OF is less than the specified Tolerance, the multiple-run is stopped.
The output signal is an array with dimension specified within the component.
More: |
Name for Identification |
Text |
Optional text parameter for identification of the component. |
||
Optimization Method |
|
Choice |
|
Select Golden Intersection, Simplex, Hooke-Jeeves or Genetic Algorithm. See Description above for more details. |
|
|
|
|
|
Number of REAL Variables to Control in this Optimization |
|
INTEGER |
Literal |
Enter the number of REAL variables to control (maximum 20).
This input is disabled if Optimization Method | Golden Intersection is selected. |
|
|
|
|
|
Number of INTEGER Variables to Control in This Optimization |
|
INTEGER |
Literal |
Enter the number of INTEGER variables to control (maximum 20).
This input is enabled only if Optimization Method | Genetic Algorithm is selected. |
|
|
|
|
|
Number of LOGICAL Variables to Control in this Optimization |
|
INTEGER |
Literal |
Enter the number of LOGICAL variables to control (maximum 20).
This input is enabled only if Optimization Method | Genetic Algorithm is selected. |
|
|
|
|
|
Maximum Number of Multiple Runs |
|
INTEGER |
Literal |
Enter the maximum number of runs allowed if the specified Tolerance is not reached (maximum 10000). |
|
|
|
|
|
Tolerance |
|
REAL |
Literal |
Enter the Objective Function tolerance which needs to be satisfied before the optimization is abandoned.
At the last time step of each run, the component stores a single value for the Objective Function. These values are compared between runs and the simulation is stopped if the difference is less than the tolerance. |
|
|
|
|
|
This Optimum Run Enabled or Disabled? |
|
Choice |
|
Select Enabled or Disabled. |
Search Method |
|
Choice |
|
Select Interval Search or Auto Search.
If Interval Search is selected, the search interval is pre-specified (i.e. Left and Right Hand Point). Whereas Auto Search seeks out a minimum search interval, starting from an initial point. Once this search interval is found, a regular interval search is performed. |
|
|
|
|
|
Left Hand Point |
|
REAL |
Literal |
Enter the start point of the search interval.
This input is enabled only if Search Method | Interval Search is selected. |
|
|
|
|
|
Right Hand Point |
|
REAL |
Literal |
Enter the end point of the search interval.
This input is enabled only if Search Method | Interval Search is selected. |
|
|
|
|
|
Starting Point |
|
REAL |
Literal |
Enter the initial starting point to seek out a minimum search interval.
This input is enabled only if Search Method | Auto Search is selected. |
|
|
|
|
|
Initial Step Length |
|
REAL |
Literal |
Enter the initial step length to be taken while performing an Auto Search.
This input is enabled only if Search Method | Auto Search is selected. |
|
|
|
|
|
Step Elongation Factor |
|
REAL |
Literal |
Enter a factor to elongate the step length. This helps the algorithm to converge to the correct point once the initial search direction is determined.
This input is enabled only if Search Method | Auto Search is selected. |
|
|
|
|
|
Search Interval Boundary |
|
REAL |
Literal |
Enter the value of the search interval boundary.
This input is enabled only if Search Method | Auto Search is selected. |
Simplex / Hookes-JeevesSimplex / Hookes-Jeeves
Initial Step Size |
|
REAL |
Literal |
This value is used to determine all other points of the simplex object. |
|
|
|
|
|
Step Reduction Factor |
|
REAL |
Literal |
Enter a factor to help speed up convergence to the optimum point.
This input is enabled only if Optimization Method | Hookes-Jeeves is selected. |
|
|
|
|
|
Initial Condition for Variables |
|
REAL |
Literal |
Specify the initial conditions of the simplex object (i.e. the initial component output values). |
Genetic AlgorithmGenetic Algorithm
Initial Population |
|
INTEGER |
Literal |
Enter the initial population of the search space. A larger initial population will increase the likelihood of the optimization space being searched evenly. |
|
|
|
|
|
Population of the Surviving Generation |
|
INTEGER |
Literal |
Enter the population of the surviving generations. This number of chromosomes will be chosen form the Initial Population. This value is kept constant throughout the process |
|
|
|
|
|
Population of the Mating Pool |
|
INTEGER |
Literal |
Enter the population of the mating pool |
|
|
|
|
|
Binary Mutation Rate |
|
REAL |
Literal |
This value determines the number of chromosomes taking part in mutation. |
|
|
|
|
|
Real part Mutation Rate |
|
REAL |
Literal |
This value determines the number of chromosomes taking part in mutation. |
|
|
|
|
|
Pairing Method |
|
Choice |
|
Select Top to Bottom, Random, Rank Weighting, Cost Weighting or Tournament. |
Genetic Algorithm - Real SpecificationsGenetic Algorithm - Real Specifications
Lower/Upper Limits of Variable |
|
REAL |
Literal |
Enter the lower and upper limits of each variable |
Genetic Algorithm - Integer SpecificationGenetic Algorithm - Integer Specification
Number of States for Integer Value |
|
INTEGER |
Literal |
Minimum value is 3. The output will change between 0 and N-1, N being the number of states specified. If the actual value increment between states is not uniform, then user should model a lookup table to translate values from 0:N-1 to the required values. If the number of states is less than 3 (i.e. 2) then it should be modeled as a binary variable. |
Output ConfigurationOutput Configuration
Write Output File? |
|
Choice |
|
Select Yes or No to write an output text file. |
|
|
|
|
|
Output File |
|
Text |
|
Enter a name for the output file if Write Output File? | Yes is selected. |