Interface Design Issues in a Solid Model Derivation System

Difficulties with part oriented solid modeling tools have motivated the development of a derivationbased solid modeling system to aid in preliminary submarine design tasks. The projected system is a graphically editable geometric rule system based on Extended Pattern Grammars with Variables (EPGVs). This paper reviews relevant characteristics of Boundary Representation Modelers, Featurebased Modelers and Extended Pattern Grammars, then establishes a fundamental relationship begveen the Inference Dependency Graph that is constructed during an EPGV derivation and a parametric system Feature tree. The primary technique involved with editing an EPGV derived model is identified with Dependency Driven Backtracking, and a description of the process is presented. This work is funded under NAVSEA contracts N00024-96-C-4180, and N00024-97-C-4186, in association with the Small Business Innovative Research (SBIR) program. Introduction required systems, compartment rearrangement is not critical. Early stage submarine design entails a multistep process wherein: ̄ Ship geometry is established and compartments are arranged to determine the minimum size ship that satisfies the mission requirements; ̄ Displacements and weights are calculated; ̄ If the ship’s displacement exceeds its weight, ballast is added to achieve neutral buoyancy and no further compartment arrangement work is required. However, if the ship weight exceeds its displacement, ship size is increased (usually by adding length) until weight and displacement are equal. Since this larger ship provides more volume and deck area than was necessary to accommodate the Past attempts to accomplish the first step quickly using conventional Computer Aided Design (CAD) systems or Solid Modelers have presented frustrating complications due to time consuming model construction, mass property calculation inaccuracies, complicated editing operations and sometimes model-corrupting software upgrades. The problems encountered are similar to those that limit the application of Solid Modelers to fast generation of building Architecture designs. Potential solutions to the problem were examined through a Small Business Innovative Research (SBIR) Phase evaluation program. A promising direction was found which takes advantage of compilation possibilities offered by viewing geometric construction as a grammatical generation proc132 MAICS-98 From:MAICS-98 Proceedings. Copyright © 1998, AAAI (www.aaai.org). All rights reserved. ess, and which utilizes an inferencing system within which productions of the grammar are derived. The resulting system holds the promise of defining a derivation-based Feature Solid Modeling environment, supporting more open editing at higher levels of abstraction. The following sections present a brief review of the involved technology, identify some problematic user interface issues and present the solutions which are currently being explored in Phase II of the SBIR work. Boundary Representation Modelers Most commercial architecture and engineering oriented solid modeling systems use some form of Boundary Representation (Brep) Modeler kernel. The currently predominant commercial kernels are Spatial Technologies’ ACIS system and EDS’s Parasolid system, both of which are original descendents from work on boundarydefined solids initiated in Cambridge England. (ACIS 96a,b,c) A notable exception to the use of commercial kernels is that of the Pro / Engineer solid modeling system which uses a proprietary modeler. Detailed descriptions of the Brep approach to solid modeling can be found in a number of texts. (HOFFMAN 89) (FOLEY 90) (MORTENSON 85) There a a few points which are important in the context of defining grammars in such environments: * Modern Brep kernels can represent both manifold and non-manifold constructions. ̄ Topological set definition is characterized by a complex graph of inter-linked entities. ̄ The inter-linked entities represent a selection of geometric primitive types, such as points, curves, surfaces and bodies, each of which is used to define boundaries for entities of higher dimension. The modelers also provide both data representation editing procedures and evaluation procedures that calculate functions of the data representation such as mass properties or surface areas. Feature-based Modelers The provision of only Boolean editing operations in Brep kernels is generally insufficient to support an adequate editing interface for engineering and architectural design work. Limitation to only these operations results in many unintuitive construction operations. For example, the inside of a compartment or room must be created as a void within a larger solid. From a designer’s point of view, a more natural approach is to construct an interior volume by the sequential introduction of bounding surfaces. Feature-based modelers have evolved this constructive approach toward the editing of solid models. Pro / Engineer was one of the first of this class of modelers, however today most part-oriented design modelers for example SolidWorks offer feature-based construction and editing mechanisms. (MCLELLAN 96) (SOLIDWORKS 97) Generally, a feature-based modeler supports user definition of 2D sketches using predefined primitives, then generates solids by the control of Brep operations such as sweeps or extrusions performed using the sketches. Complex 3D solids are the results of composition of the first order solids under the control of subsequent Boolean or other Brep operations. The key points here are that: ̄ Each feature can be regarded as mapping to an expression that is evaluated to construct entities composing the solid model. In some systems these expressions can be regarded as constituting a single function. Most commercial modelers use some variant of a relational constraint manager; such as D-Cubed’s 2D dimension constraint manager (DCM) that used in the evaluation of user defined 2D sketches in SolidWorks. The D-Cubed product