You are here

Constellation: A Multidisciplinary Design Optimization Graphical User Interface

In 2011, the Computational Science and Engineering Laboratory (CSE Lab) developed a scalable, multidisciplinary material attribution method under Riverside Research’s newly established Independent Research and Development (IR&D) program. This material attribution method was implemented within the CSE Lab’s design optimization graphical user interface (GUI) known as Constellation.

Background
Constellation is a Riverside Research-developed design optimization GUI that breaks new ground in terms of its usability and sophistication. Its innovative design enables the user to specify design optimization parameters, select the appropriate physics kernels, launch other GUIs as needed to configure code-specific parameters, and define computer-aided design and material characteristics. Constellation is built on a revolutionary data model developed by the CSE Lab that cleanly separates the GUI from its underlying data structures. This model empowers the user through spreadsheet-like capabilities and provides real-time validation feedback that can be employed to guide a user through the process of defining a problem’s scenario. 

Approach
The physics-based simulations targeted by Constellation operate on mesh files that describe the structures being modeled. The CSE Lab implemented the multidisciplinary material attribution capability by assigning material or optimization attributes to geometric meshes. Attributes were implemented with the following classes:

  • Materials – These are the constitutive parameters that define how domain phenomena, such as electromagnetic fields, behave within the context of a curve, surface, or volume (e.g., how an antenna element behaves when excited by a source). The definition must also support the material’s properties across multiple disciplines (electromagnetic, infrared, thermal, computational fluid dynamics, etc.) and also for different types of geometry parts, such as curves, surfaces, and volumes. Accordingly, each material definition is actually a composition of one or more parameter sets that define the properties of the material for a unique discipline/part-element pair.
  • Discretely Varied Material Optimization – This class defines a set of candidate material definitions from which an optimizer may select a single definition. For example, when designing an antenna, a designer may have n different antenna substrates available for manufacture, each of which has different properties. In this situation, the optimizer may select any one material from those n candidates.
  • Continuously Varied Material Optimization – This class defines a single parametric material definition where rather than having all of its definition fixed prior to the optimization process, the definition is modified by the optimizer. For each optimizable parameter within the definition, the optimizer chooses an independent value from the continuous range assigned to that parameter. This differs from the discretely varied material optimization in that the materials being chosen are not pre-defined, but rather a single base material with properties that can be changed according to a set of constraints.
  • Discretely Varied Material Optimization Group – This class allows the optimizer to control k parts as a group by selecting 1 ≤ jk materials from a list of material definitions and applying those materials to the k parts. This allows those parts to operate as a group.
  • Continuously Varied Material Optimization Group – Much like the discretely varied material optimization group, this defines a set of k material optimizations that are then applied to k different parts. As in continuously varied material optimizations, the optimizer will choose a value for each parameter within that parameter’s defined range. Unlike the continuously varied material optimization, these parameters are controlled together as a group by a spline function. The spline function allows for smooth variations across the generated material definitions and also reduces the complexity of the optimization by having the optimizer drive only the spline control points rather than directly controlling the values of these parameters.

Accomplishments
The following features were added to Constellation under this IR&D project:

  • Multi-discipline material definition    
  • Multi-discipline material attribution    
  • Material optimization (continuous ranges)    
  • Backend data management enhancements
  • Geometric part grouping    
  • Mesh coloring by part/material
  • Notification system
© 2017 Riverside Research