Training Module #14: Quantify Producibility with HyperXpert

Fundamental to understanding HyperXpert is understanding Design Variation - i.e. the change in cross-sectional dimension across consecutive Zones in a selection. That selection of Zones, in the physical world, could be considered a part surface. The more variation that is seen between Zones, the harder that part will be to produce (as tooling may vary, time to build may increase, etc.).

Take the wing model shown below. For simplicity's sake, the upper skin is sized as a metal plate. A range of possible thicknesses is given in the assigned Design Property (i.e. the design space).

Running a preliminary Sizing on the entire upper skin, HyperX will find the thinnest plate that satisfies all strength and stability criteria and apply that to the entire skin. This design, "Design #1", is the most producible design possible since there is no thickness variation across the part surface. However, it is typical aerospace practice to allow particular design variables - in this case, thickness - of each panel vary across the part surface. In this instance, HyperX will find the thinnest plate that satisfy the loads and criteria for each panel, rather than assigning the controlling thickness to the entire part. This is the weight optimum design, but comes with increased variation in thickness across the part surface (and therefore a decreased level of producibility as compared to "Design #1"). This is "Design #2".

For each design, which are both considered viable options for this part, a variation score (based on the amount of change in thickness across adjacent panels on the part surface) and corresponding weight are calculated. These are the bookends of the problem. But, based on the number of cross-sectional candidates defined on the Design Property, there are potentially 1000s more viable part designs.

Running a full factorial design of experiments (DOE) of all possible thickness combinations across the part surface, HyperX systematically identifies all positive margin wing skin designs, calculates the corresponding weight and variation, and plots them accordingly. Notice that the high weight, low variation (i.e. high producibility) designs trend towards the upper lefthand side of the plot; while the low weight, high variation (i.e. low producibility) designs trend toward the bottom right.

HyperXpert offers two methods for quantifying Design Variation across a part:

The first algorithm quantifies variation by determining the total number of variables within a part.

A Constant dimension is a cross-sectional dimension that is same across all Zones in the part. The thickness in Design #1of the Simple Example given in the previous article is an example. Variables, then, are the opposite - dimensions that are not the same across Zones. The more Constant cross-sectional dimensions, the less Design Variation there is across a part, and the more producible it becomes.

To mirror previously explained trends, this data is plotted in terms of Design Variability (# of Variables). So, the more Design Variability along a part surface, the harder it is to manufacture - thus ensuring that the lightest weight design with the most variation is found in the bottom right of the plot. In the example Design Property below, there are 3 Variables (indicated by the black boxes). All other inputs are Constants (i.e. all Zones will have the same dimensions for these values).

HyperXpert runs full factorial design of experiments (DOE) of all possible ways design variables and their prescribed values can be forced constant across a part surface. For each solution to the DOE, the number of Constants vs. Variables are stored. The weight of each solution is then plotted with respect to the total number of Variables in the design. See the example plot below.

Using the tree on the left, each Data Point can be identified if it is a design that is held constant at the selected dimension. For example, all red points in the image above represent designs in which all Zones have a stiffener height of 0.7in. For more information on identifying points and operating the plot interface, see The HyperXpert Viewer.

Further explaining the example above:

Starting on the right where #variables = all* - This one datapoint has maximum variability. This represents the typical HyperX Sizing result that allow every dimension and every Material/Laminate to be different throughout the Structure, across all Zones. This one data point is the lightest design from the lightest Sizing result of every Zone.  

Moving to the left, the axis with #variables = all* - 1 - Represent designs that have one constant input across all Zones. Such as H=1.2” or Spacing = 7”. Only the lightest design of each unique variable is displayed.

Moving furthest left where #variables= 0 - These datapoints have all Constant input values. These are designs that have every dimension and every Material/Laminate the same across all Zones. As such, these designs have to carry the highest load of any Zone and are the heaviest.

Variability scoring counts the number of times a particular variable changes within a design datapoint. In the example below, the height of the stiffener changes 39 times within the design. This is an indication of how producible the corresponding design is.

Start by opening the HyperXpert Runner from the Run tab of the Ribbon.

Then, use the resulting form to define the Run Set: