Global Structure Definition#
This block stores all information defining the structural design. The basic logic behind the parametric definition of a structure is as follows.
Create base designs of cross-sections
Identify designable parameters
Assign specific values to those parameters to define a specific design of the structure
For iterative running of the structural analysis (such as optimizations), choose designed parameters as design variables and specify their design space (range or set)
The overall layout is shown in Listing 1.
structure:
name: "..."
parameter:
...
distribution:
...
design:
...
cs:
...
cs:
- name: ""
parameter:
...
design:
...
model:
...
- name: ""
...
The structure block specifies the top-level (global) structural design for the current analysis or design optimization, such as a beam structure or a plate structure.
The cs block is a library storing a list of base designs of cross-sections that will be used for the global structure.
Note that both structure and cs blocks share some common input entries.
These entries have mostly the same input syntax and meanings.
Two key entries are the design, which specifies the base design, and parameter, which specifies the designable parameters.
Parameters (parameters) specify the values defining a specific design of the structure.
They can be given either in a list or as distributions with respect to the structural dimensions.
Supported parameters depend on the structural class.
Note
For more information on the parameterization of beam structures and cross-sections, please check the Section Parameterization of Composite Slender Structures.
Parameter specifications (parameter)#
In this section, structural design specifications are given in a list of pairs of parameter name and value. The basic syntax is shown in Listing 2.
structure:
parameter:
param_1: value_1
param_2: value_2
...
cs:
- name: "cs1"
parameter:
param_3: value_3
param_4: value_4
...
Parameter list can be used to specify a constant value for the whole structure. These parameters can also be varied by an optimization method. This is done by directly using the parameter name as the design variable name.
Distribution of parameters (distribution)#
In this section, structural parameters are defined as distribution functions with respect to the structural dimensions. For slender structures that will be analyzed using beam models, this means that parameters are functions of a single coordinate along the longitudinal direction (e.g., blade span). The basic syntax is shown in Listing 3.
structure:
distribution:
- name: "..."
function: "..."
...
- name: "..."
...
...
Name (name) can be arbitrary for each distribution.
Function (function) is specified by choosing one of the built-in types such as interpolation functions.
Note
Currently only interpolation function is supported.
Other specifications depend on the type of function selected.
Interpolation function (function: "interpolation")#
This option creates one or multiple distribution functions by interpolating a table of data.
structure:
distribution:
- name: "..."
function: "interpolation"
kind: "linear"
xnames: "x"
ynames: ["a", "b", "c"]
ytypes: ["float", "int", "string"]
data: |
x1, a1, b1, c1
x2, a2, b2, c2
...
This function supports two different kinds, indicated by the keyword kind: linear (linear) and previous (previous).
Suppose we want to get the value y at x which locates between x1 and x2.
Given two data (x1, y1) and (x2, y2), linear kind function means that y is calculated by linearly interpolating these two data.
previous kind function uses the value at the location no greater than x (y=y1 in this case).
The set of data is specified by multiple keywords.
xnames and ynames are used to specify the list of names or labels of the independent and dependent variables, respectively.
Generally, labels should be placed in square brackets, delimited by commas.
If there is only one label, the square brackets can be omitted.
Currently iVABS only create scalar functions with a single output.
Hence, the labels in ynames indicates all the functions that will be created.
data is used to place the actual data that will be interpolated.
This is a literal block indicated by the vertical bar (see Section Scalars).
Values are arranged in a tabular form.
Each row is an entry of the data and values are separated by commas.
If kind is linear, there should be as least two entries.
If kind is previous, a single entry is acceptable.
Each column is either an independent or dependent variable.
In the most general case, the first \(n\) columns correspond to the \(n\) xnames.
Then the next \(m\) columns correspond to the \(m\) ynames.
Hence, the number of columns should be the same as the number of labels in both xnames and ynames.
Different types of values can be specified using the keyword ytypes.
By default, all values are treated as real numbers (float).
Other supported types are integer numbers (int) and strings (string).
Different parameters (ynames) can have the same or different types.
If ytypes is a string, then the type will be applied to all parameters listed in ynames.
Otherwise, ytypes must have the same size as ynames.
All quantities in the data block can be marked as design variables that will be changed later by Dakota. The design variable name is prepended to the value followed by a colon:
ynames: ["a", "b", "c"]
data: |
x1, dv2: a1, dv3: b1, c1
dv1: x2, a2, dv4: b2, c2
...
In this case, “dv1” to “dv4” are four design variables used by Dakota. During any iterative process (parametric study, optimization, etc.), values marked by the design variable names will be substituted by new values generated in the current iteration. It is acceptable that a quantity is marked by a label but not used in Dakota. In other words, values after the colon can be treated as the default ones.