This node is located in the Systems > Algorithms > Optimization folder in the Modules pane.
The Global and Local Approximation of Derivatives (GLAD) optimizer is an optimizer, which uses derivative information from the solver to speed up the optimization process. Specifically, the GLAD optimizer can actually consider the derivatives of response signals from Ansys HFSS. Based on the derivatives of the real and imaginary part for each available signal response, the dB values of the signals are assembled. These signal responses must be specified in the settings. In the current state, only a single-objective optimization considering constraints is possible. In the objective function every scalar, vector or signal output can be considered, but the focus should be on the dB scale of the antenna spectrum to achieve the optimal convergence of the optimizer. In each iteration, a global and local search can be performed to ensure the local convergence and keep global exploration of the algorithm. If no start designs are available, the global search starts with 10 Latin Hypercube samples in the first iteration.
In Ansys HFSS an ACT wizard is available, which helps to set up the optiSLang project including the necessary signal responses and derivatives registered in the AEDT integration node. After setting up the project with the wizard, it is necessary to define the optimization criteria in the GLAD system.
Node Settings
To access the options shown in the following table, double-click the Global and Local Approximation of Derivatives (GLAD for HFSS) system on the Scenery pane and switch to the Settings tab.
Option | Default Value | Description |
---|---|---|
Maximum number of iterations | 20 | Defines the maximum number of iterations, where each iteration may contain a local and/or a global search resulting in maximum two designs per iteration. |
Minimum number of iterations | 5 | Minimum number of iteration with local and/or global search. The convergence criterion is checked the first time, when the minimum number of iterations is reached. |
Stop after stagnation iterations | 5 | Number of iterations before the algorithm will stop if no further improvement can be found. |
Start local search at iteration | 3 | Start iteration of local search approach. If no global search is selected, the local search starts in iteration one independent of this setting. |
Number of search samples | 1000 | Number of search samples for the internal local and global search algorithm. These samples are generated only on the approximation and will increase the effort of the algorithmic algebra but not cause additional solver runs. |
Initial scale factor | 0.5 | Initial scale factor of the local search approach, which locates the search region around the best design found so far. |
Shrinking factor | 0.7 | The shrinking factor decreases the search domain of the local search in each iteration, until the minimum search range is reached. |
Minimum search range | 0.05 | Minimum range for the local search approach. |
Use global approximation | true | Enables global approximation for exploration, if disabled, only a local search is performed. |
Signal identifier | S11,S12,S22 | Specifies signals to consider in the approximation of the real and
imaginary parts. The naming should contain only the core of the response
name without the suffixes _real, _imag. and _dB. For example if you have
S11_real, S11_imag, and
S11_dB, just enter S11 .
More than one signal names can be specified by using a comma as a
separator. |
Write internal omdb files | false | The internal OMDB files can be used for debugging the local and global search samples in each iteration, generated on the local and global approximation models. |
Additional Options
To access the options shown in the following table, in any tab, click
.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:
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 selected, all other nodes have selected. |