Adaptive Metamodel of Optimal Prognosis (AMOP) Node

The Adaptive Metamodel of Optimal Prognosis is an iterative meta-modelling approach based on the MOP using an adaptive refinement of the data points. It performs similar as an optimizer, running a defined number of solver runs in several iterations. As convergence criterion a minimum value of the Coefficient of Prognosis has to be obtained for all selected responses.

Relevant Parameters

Only parameters with deterministic properties (Optimization or Opt.+Stoch. parameter) are relevant for the analysis.

Initialization Options

To access the options shown in the following table, double-click the AMOP system on the Scenery pane and switch to the Adaption tab.

Option	Description
Adaption
Refinement type	Select one of the following refinement options: Global: Adds additional samples to the existing data set by using global sampling sets as Latin Hypercube Sampling until the target CoP is reached for all responses. Local: The first iteration generates a deterministic or random sampling. In the following iterations, new data points are placed in the regions of maximum approximation errors or where a maximum improvement of the criteria is expected. In both refinement approaches, the position of the start designs are considered.
Maximum number of samples	Sets the total number of samples and adds the recommended number of samples in the advanced settings.
Show advanced settings	If selected, the advanced settings options are displayed.
Advanced settings
Start iteration
Use start designs only	When selected, the first iteration uses only start designs as support points and initial sampling is deactivated.
Sampling type	Specifies the type of sampling or deterministic design of experiment scheme for the initial iteration. All DOE schemes are available.
Number of samples/levels	Specifies either the sample number or discretization level of the initial DOE method.
Refinement (Global refinement type)
Sampling type	Specifies the type of sampling for the refinement iterations. Only random sampling schemes are available.
Number of samples	Specifies the sample number for each iteration of the refinement sampling method.
Refinement (Local refinement type)
Importance sliders	Specify the percentage of the usage of the different local refinement strategies: Importance of sample density: Only the sample density is considered in the generation of refinement points. Importance of local CoP: The refinement points are created in regions with large local approximation errors. Importance of optimization criteria: The objective functions and constraints are considered for new refinement points. In case of single and multiple objectives, the region with expected maximum improvement of the objective functions are refined. If only constraints are defined, the refinement points are created in regions, where the constraints are expected to be fulfilled.
Number of samples per iteration	Number of designs in the local refinement iterations.
Pareto dominance refinement	Use this setting when performing criteria based refinement with multiple objectives. When selected, performs a Pareto optimization update, which results in a maximum spread of the Pareto frontier. When cleared, uses the best compromise between the objectives as local update criterion.
Consider failed designs	Positions of previous failed designs are considered in the global and local refinement procedures. If unchecked, their positions are ignored in the density estimate.
Convergence criteria
Target CoP	Sets the minimum Coefficient of Prognosis, required for all selected responses. If this value is reached, the AMOP algorithm has converged. In the global refinement, the global CoP values of all responses are considered. Tn the local approach the minimum local CoPs of all samples are used as error measure for the local CoP refinement.
Maximum iterations	The algorithm stops at the maximum number of iterations, if the convergence criterion is not reached. This number considers the initial and the refinement iterations.
Stagnation iterations	The stagnation iterations are considered in the local criteria-based refinement for single-objective optimization. The refinement stops if no improvement of the best design is obtained within the specified number of iterations.

Additional Options

To access the options shown in the following table, in any tab, click Show additional options.

Option	Description
Limit maximum in parallel	Controls the resource usage of nodes in the system. When the check box is cleared (default), a value is chosen to ensure the best possible utilization of the child nodes. When the check box is selected, set the value manually to specify how many designs are sent to child nodes, limiting the maximum degree of parallelism for all children. Ansys recommends keeping the check box clear.
Auto-save behavior	Select one of the following options: No auto-save Actor execution finished Every n-th finished design (then select the number of designs from the text field) Iteration finished (for optimization, reliability) The project, including the database, is auto-saved (depending on defined interval) after calculating this node/system (either when the calculation succeeds or fails). By default, all parametric and algorithm systems have Every nth finished design 1 design(s) selected, all other nodes have No auto-save selected.

Option

Description

Limit maximum in parallel

Controls the resource usage of nodes in the system.

When the check box is cleared (default), a value is chosen to ensure the best possible utilization of the child nodes. When the check box is selected, set the value manually to specify how many designs are sent to child nodes, limiting the maximum degree of parallelism for all children. Ansys recommends keeping the check box clear.

Auto-save behavior

Select one of the following options:

No auto-save
Actor execution finished
Every n-th finished design (then select the number of designs from the text field)
Iteration finished (for optimization, reliability)

The project, including the database, is auto-saved (depending on defined interval) after calculating this node/system (either when the calculation succeeds or fails).

By default, all parametric and algorithm systems have Every nth finished design 1 design(s) selected, all other nodes have No auto-save selected.