Design Studies Reference

Manual

Activates the Design Table node, which allows you to create the designs to run in tabular form. See Creating Manual Designs.

Sweep

Activates the Design Table node. Edit this node to display the list of designs that are created based on the Input Parameters settings. The number of designs to run is determined by the number of variables and how many values are used for each variable.

Smart Sweep

Activates the Smart Sweep Settings node. To set the properties, refer to 智能扫掠设置参考.

DOE

Activates the DOE Settings node that allows you to set the DOE Type to run:

2 Level Full Factorial

Carries out a two-level full factorial DOE study.

3 Level Full Factorial

Carries out a three-level full factorial DOE study.

Latin Hypercube Sampling

Carries out a DOE study that uses Latin Hypercube sampling to determine the input parameter values. Activates the Sampling Settings node that allows you to set the total number of Designs to Run in the design study. The specified value must be equal or smaller than the number of possible designs as defined by the input parameters.

Adaptive Sampling

Activates the Adaptive Sampling Settings node with the following properties:

Designs to Run

Specifies the total number of designs to run in the design study.

Adaptive Strategy

Specifies the strategy to use during sampling from the following options:

• Balance Local and Global Accuracy (default)—This strategy generates a surrogate with a good overall fit across the entire design space and also increases the sampling in local regions of interest. For this strategy, search cycles initially focus heavily on exploring broader regions. As the algorithm learns the design space, it gradually reduces its broad focus and shifts to reducing approximation error and focusing on local properties.
• Prefer Global Accuracy—Generates a surrogate with good global fit. All search cycles focus heavily on exploring broader regions. Only a small percentage of samples focus on local properties.
• Prefer Local Accuracy—Generates a surrogate with good fit in regions of interest. All search cycles focus heavily on reducing approximation error. Only a small percentage of samples focus on exploring broader regions.
• Detect Edge Cases—Generates sample points around edges or discontinuities. Every search cycle focuses somewhat on all three algorithms. Focusing on local properties has the highest weighting to start with and its weight gradually increases in importance. The other two algorithms gradually decrease in subsequent search cycles. To help in locating the edges, this method uses a gradient calculation in determining which local regions to sample from.

Responses to Consider

Specifies study responses for surrogate generation.

Surrogate Types

Determines the surrogate types used in surrogate generation. Multi-selection is allowed. By default, both Radial Basis Functions and Kriging are selected. The adaptive sampling algorithm specifies the final surrogate type within the selected types. When surrogate type Kriging is included, you must additionally set the Tuning property in the Kriging child node. Choose between Fast Kriging, Precise Kriging, or a Guassian Process. See also 替代模型类型.

Optimization

Activates the Optimization Settings and Design Seeds nodes. The Optimization Settings node controls the following properties:

Optimization Type

Specifies the type of optimization to run.

Weighted sum of all objectives

Specifies a design study where all objectives are combined into a single objective, which is optimized by the algorithm. The objectives are combined using the performance function given by Eqn. (547).

Multiple objective tradeoff study (Pareto front)

Specifies a design study where each objective is independently considered by the algorithm.

This option activates the Archive Size child node. The archive size is the number of non-dominated designs that you want the algorithm to find during the search in order to resolve the Pareto front.

The value of the archive size must be evenly divisible by four and greater than or equal to 12. This value should also be greater than or equal to the number of input parameters in the design study. As an example, a Pareto optimization study with 8 variables would have an archive size of 12, while one with 13 variables would have an archive size of 16.

Single-objective and multi-objective studies use the SHERPA algorithm.

Designs to Run specifies the total number of designs to run in the design study. For single-objective Optimization studies, run at least 30 to 35 designs. As a guideline, set this value to 10 times the number of input parameters in your design study. For multi-objective Optimization studies, run at least 40 designs, which is the default value. As a guideline, for multi-objective Optimization studies, set this value to 8 to 12 times the archive size. If the archive size is larger than 20, the number of designs to run is multiplied by 2 to give 40 designs. The SHERPA algorithm works best when it can learn as many times as possible.

Sequential Quadratic Programming (SQP)

Specifies a gradient-based optimization method based on geometric and surface sensitivities. See also: 基于梯度优化.

The child node SQP Method node controls the following properties:

Maximum Outer Iterations

Defines the number of the SQP outer iterations. At every outer iteration, Simcenter STAR-CCM+ recalculates the sensitivities of the parameters.

The default value of 10 indicates that after calculating the sensitivities 10 times, the optimization stops.

Maximum Line Search Iteration:

Specifies the maximum number of designs that Simcenter STAR-CCM+ runs using the same sensitivities.

The default value of 5 indicates that you run a maximum of 5 designs before recalculating the sensitivities.

Starting Design

Specifies the starting point for the optimization search.

Robustness and Reliability

Activates the Robustness and Reliability Settings node with the following properties:

Designs to Run

Specifies the number of study to run. To obtain a meaningful distribution, set this value to 2000 or higher. For a robustness and reliability study, applying surrogates provides an efficient method that reduces computational time. See also 生成和应用替代模型.

Sampling Method

When you have a large number of Designs to Run, say 10000, the random sampling method Monte-Carlo is more efficient than the structured method Latin Hypercube.

CAD Robustness

Activates the CAD Robustness Settings node that allows you to set the Sampling Type for the input parameters:

Manual

Activates the Design Table node, which allows you to set the input parameter values for the designs to run in tabular form. See Creating Manual Designs.

Sweep

Activates the Design Table node. Edit this node to display the list of designs that are created based on the Input Parameters settings. The number of designs to run is determined by the number of variables and how many values are used for each variable.

Latin Hypercube

Uses Latin Hypercube sampling to determine the input parameter values. Activates the Sampling Settings node that allows you to set the total number of Designs to Run in the design study. The specified value must be equal or smaller than the number of possible designs as defined by the input parameters.

• Pie Chart
• History
• Parameter vs Response
• Response vs Response
• Pareto
• Bubble
• Parallel Coordinates
• Probability Distribution
• Surrogate
  • Surrogate Fit
  • Actual vs Residual

For more information, see Creating XY Plots and Creating Parallel Coordinates Plots.

• [design study]0.res
• [design study]0.gph
• [design study]0.hst
• [design study]0.plot
• DOE_0_[response].out