 |
|
To enter the Run Project Dialog, click on the ,
identified by the icon 
or select from the main menu.
Note:
The is enabled if and only if the Work Flow is well-formed and without any error messages.
The Run Project Dialog is divided into two parts:
The Project File Chooser where the project file can be located or saved.
The Project Preferences Panels where all the runtime parameters can be defined.
The Project File Chooser is a simple file system browser located on the left hand side of the Run Project Dialog (see Figure 8.2, "Run Project Dialog"). The user can browse the file system and save the project file in the preferred location. At the run time modeFRONTIERTM will create the Project Running Directory starting from the chosen location.
The Project Preferences Panels are a set of tabbed panels where the run time parameters for the analysis loop, for the RSM based optimization and for the "Multi Objective Robust Design Optimization (MORDO)" can be set up.
This panel (see Figure 8.2, "Run Project Dialog") includes the main run time parameters for the analysis loop:
Num. of Concurrent Design Evaluation is the number of concurrent (parallel threads) design evaluations the user is willing to run. The maximum number of concurrent design evaluations allowed is limited by the license available.
Clear Design Dir on Exit combo box defines the modeFRONTIERTM directory cleaning policy:
Never - This option will never remove the working directories at the end of each design evaluation.
Always - This option will always remove the working directory .
Not On Failure - This option will only remove the working directories associated to a successful design evaluation. The option is recommended where the design's run logs are important for the debugging activities.
Evaluate Repeated Designs check box, if enabled, will evaluate again the designs that are already present in the design database. Otherwise the values already present will be reused.
Save Repeated Design in DB check box, if enabled, adds the designs even if they are already present in the design database. The designs are copied and not re-evaluated. This is useful to see how the optimisation algorithms evolve.
Save Error Design in DB check box, if enabled, saves the failed designs in the design database. Otherwise they will be discarded.
Evaluate Unfeasible Designs check box, if disabled, checks if the constraints on the input variables are feasible or not. If at least one constraint is broken the design is not evaluated and marked as an error design.
Process Priority parameter specifies the priority of the running processes. You can choose between five values: High, Above Normal, Normal, Below Normal and Low. The high priority is the default value.
This panel (see Figure 8.3, "RSM Preferences Panel") includes the run time parameters related to the RSM driven optimization:
RSM Functions tree includes all the Response Surfaces (RSM) generated by the user. The user has to define at least one RSM function to be the default one for every output variable. To mark a function as default just click with the Right Mouse Button on the selected RSM function.
RSM Evaluation Percentage parameter specifies the percentage of the design evaluations made using the RSM function associated to each output variable.
This panel (see Figure 8.4, "MORDO Preferences Panel") includes the run time parameters related to Multi Objective Robust Design Optimization (MORDO):
Sampling Mode parameter defines the sampling algorithm. The options are:
Montecarlo Sampling - This option generates the samples using the standard Statistical Distributions defined by the user.
Latin Hypercube Sampling - This option generates the samples using the Latin Hypercube distributions as described in Chapter 6, Design of Experiments (DOE).
Seed parameter is the seed of the random numbers generator used in the sampling cycle.
Num. of Design Samples parameter specifies the number of design evaluations needed by the statistics of the MORDO algorithm.
Note:
The number of samples plays an important role in robust design optimization. The higher the number of samples, the closer will be the mean to the real expectation. Unfortunately, this parameter can significantly increase the analysis time. This problem can be partially solved by using virtual samples instead of real samples.
Num. of Virtual Samples parameter specifies the number of virtual design evaluations that can be used to increase the overall number of samples needed by the MORDO algorithm and speed up the computation. These virtual samples are evaluated with a Kriging Response Surface automatically trained on the previously evaluated real samples. Both real and virtual samples contribute to the computation of means and standard deviations but only the real samples appear in the Design Table. A high use of virtual designs reduces the reliability of the results.
Reject Out of Bounds Samples check box will let the sampling algorithm to automatically keep or reject values that do not respect input variables bounds.
Error Samples Acceptance Level percentage parameter determines the handling error designs policy. Its value represents the largest acceptable fraction of error designs: if the percentage of error samples is above this threshold, the correspondent robust design is set to error. For example:
→ 0 % - Intransigent policy: at limit, a single error sample is sufficient to consider the correspondent robust design as an error design. This policy corresponds to the Quality Control point of view.
→ 100 % - Tolerant policy: at limit, one single good sample is sufficient to retain the correspondent robust design. This choice permits to utilize all the available information, without any resource wasting.
