The Evolution of a Decision Support Architecture for Polymer Composites Design

• Authors: Timothy J. Lenz;James K. McDowell;Martin C. Hawley;Ahmed Kamel;Jon Sticklen
• Venue: IEEE Expert: Intelligent Systems and Their Applications
• Year: 1996

Polymer composites--lightweight, strong, and energy-efficientmaterials--offer significant advantages to durable-goodsmanufacturers and to performance-driven markets such as theaerospace industry. In describing a knowledge-based architecturethey've developed for automating the design of such materials, theauthors trace the evolution of their approach to polymer compositesdesign.Over the last several years, an effort has been underway in theIntelligent Systems Laboratory at Michigan State University todevelop a knowledge-based system for designing polymer compositematerials. The Composite Materials & Structures Center, also atMichigan State, has helped facilitate much of this work. TheAdvanced Computing Thrust (ACT) combines the generic task approachof the ISL with the expert composites knowledge of the CMSC and hasled to the development of several successful prototypes for thedesign of polymer composite materials. This article looks at thedevelopment and evolution of both the ideology and methodologybehind the work on this project.Polymer composites consist of a reinforcing structuralconstituent and a protective polymer matrix. The properties of thecombined material are significantly better than the sum of theproperties of each component, giving materials with highstrength-to-weight ratios. As a result, polymer composite parts aregenerally 20 to 30% lighter than the corresponding metal parts. Thestudy of polymer composites focuses on the properties andfabrication of materials and is based largely on polymer scienceand chemical engineering. This area is inherentlymultidisciplinary. A key to realizing the potential of polymercomposite materials is enabling a rapid transition from the settingof material specifications to the successful commercialmanufacturing of a material that meets those specifications.Presenting an intelligent decision-support system for compositesdesign to address this challenge becomes increasingly important aspolymer composite materials penetrate the durable goods markets andas performance-driven markets, such as aerospace, become morecost-conscious. Knowledge-based systems, in the context of polymercomposites design, facilitate the use and reuse of engineeringdesign knowledge, thus enabling the transfer of expertise andfreeing design engineers for more creative tasks.We approach the intelligent decision-support system for polymercomposites design from the perspective of generic tasks. Forexample, in designing materials, we use routine design (a generictask). This type of design is possible in a domain where effectiveproblem decomposition and compiled design plans are explicitlyknown. Such a situation exists for polymer composite materials.Material design for polymer composites involves mapping fromenvironmental and performance requirements (mechanical, thermal,optical, electrical, and chemical) to choices for fiber, matrix,and chemical agents. The issues of manufacturing technology choice,processing parameters, and specific part architecture are alsoimportant in the design of polymer composites, and we consideredthese issues in developing the decision-support systems.Others have applied a knowledge-based approach to the design ofmaterials. Ingemar Hulthage et al. used both qualitative andquantitative information in designing aluminum alloys. RangarajanPitchumani et al. and Andreas Nitsche et al. each focused on anintegrated design system for composites, except that they addressedceramic and metal matrix composites instead of polymer matrixcomposites. The first of these employed a routine-designmethodology of sorts, but considered the decomposition ofcomposites design in a different manner than what we present. Thelatter used a more traditional approach to composites design:coupling finite-element modeling (FEM) with a database and a modelof material behavior.This article summarizes the salient points of routine designemployed throughout the development of the composites designarchitecture, and looks at the four material designers that thearchitecture implements using ParcPlace-Digitalk's Smalltalk. Wecompare these material designers by examining the dynamic globalview of polymer composites design and of material design. We alsoelucidate notable features of each system through a benchmarkdesign example. Finally, by comparing the initial and currentapproaches, we illustrate the differences in both methodology andideology that have arisen in the evolution of this knowledge-basedsystem.