Michael J. Wozny

Predicting the drape of woven cloth using interacting particles

We demonstrate a physically-based technique for predicting the drape of a wide variety of woven fabrics. The approach exploits a theoretical model that explicitly represents the microstructure of woven cloth with interacting particles, rather than utilizing a continuum approximation. By testing a cloth sample in a Kawabata fabric testing device, we obtain data that is used to tune the models energy functions, so that it reproduces the draping behavior of the original material. Photographs, comparing the drape of actual cloth with visualizations of simulation results, show that we are able to reliably model the unique large-scale draping characteristics of distinctly different fabric types.

An overview of automatic feature recognition techniques for computer-aided process planning

Abstract In recent years, various researchers have come up with different ways and means to integrate CAD and CAM. Automatic feature recognition from a CAD solid model, for downstream applications like process planning, greatly impacts the level of integration. A brief discussion and review of methods used in automatic feature recognition like cell division, cavity volume, convex hull, laminae slicing and other miscellaneous techniques which include graph-based and hint-based feature recognition methods have been presented. Automatic feature recognition for machining is essentially viewed as an exercise in decomposition. It is further confirmed that even in a “design by features” environment, the need for feature recognition is inevitable.

Fitting a woven-cloth model to a curved surface: mapping algorithms

Abstract The problem of fitting a 2D broadcloth composite ply (modelled as a piece of deformable woven cloth) onto a 3D curved surface (modelled with a NURBS surface) is considered. ‘Fitting’ means that a ply is deformed and applied to a surface so that it is everywhere in contact with the surface. A fitting method is presented that incorporates a flexible method for specifying initial conditions, and allows for the simulation of a wide range of fitting configurations.

Modeling methods for the design of 3D broadcloth composite parts

With the advent of lighter and stronger composite materials over the last two decades, more and more metals have been replaced by composite materials in aircraft and other vehicles. It is not suprising that the airframes of certain aircraft (e.g. Lear Fan 2100) are totally fabricated from composite materials. Composites consist of a reinforcing material suspended in a “matrix” material (e.g. epoxy) that bonds it to adjacent reinforcing materials. The three major composite forms are chopped fiber, unidirectional tape, and (bidirectional) broadcloth. Today’s aircraft mostly employ broadcloth composites partly because they have outstanding strength-to-weight ratio, and partly because their structural properties may be tailored to the expected load in different directions [16]. Broadcloth composites have both vertical and horizontal threads (weft and warp) interwoven to form a sheet of cloth as seen in Fig 1. They strongly resist stretching along thread directions, but can be deformed flexibly along thread diagonals by changing the angle between vertical and horizontal threads. This allows a composite broadcloth to be “formed” into virtually any curved surface. In order to reinforce aircraft structures, multiple sheets (plies) of broadcloth composites are usually laid one on top of the other, a lamination process. With this trend of replacing metals with composite materials, new lamination technologies have been developed to manufacture composite materials in order to form an airframe. During the lamination process, plies of composite sheets must be cut to the required size, with the required fiber orientation, and laid up onto the surface of a die or molding tool prior to curing. It is this process of deforming a broadcloth composite sheet into a 3D shape by pressing it onto a mold that we will refer to as a “fitting.” Unfortunately, the cutting and orientation process has not yet been fully automated, and it has been mostly implemented with manual procedures. It is therefore a time-consuming, non-reproducable, and inaccurate process in need of design automation tools. In this paper we present a number of modeling methods that may be used to automate the design of 3D broadcloth composite parts. First, we describe a model (a Tchebychev net) which allows us to simulate the deformation of woven materials into a specific 3D shape. Two algorithms are described for performing the actual fitting. The first algorithm simulates the fit by solving the Tchebychev net formula using a finite difference technique. The second algorithm simulates the fit by reducing the problem to a surface-surface intersection problem. Once we establish the techniques for simulating a fit, we can discuss the quality and acceptibility of the fit. In general, a good fit is the one that consumes the smallest area of the material, that has the smallest deformation energy, and that is free of manufacturing anomalies such as wrinkles and breaks. We will present mathematical tools that allow us to measure the “goodness” of a fit with a Tchebychev net, and that allow us to visually identify where a possible anomalous event may occur. Specifically, we will introduce a path-dependent Gaussian curvature integral that is defined at an arbitrary point in a surface region. With a path-dependent Gaussian curvature integral, we will show that it is possible not only to predict anomalous events, or wrinkles in particular, but to provide a solution to preventing them. Finally, we will propose three methods for preventing anomalous events: (1) automatic generation of a good initial condition, (2) dart insertion, and (3) surface shape modification. Providing different methods for preventing

Generating Topological Informa*ion from a "Bucket of Facets"

The STL de facto data exchange standard for Solid Freeform Fabrication represents CAD models as a collection of unordered triangular planar facets. No topological connectivity information is provided; hence the term “bucket of facets.” Such topological information can, however, be quite useful for performing model validity checking and speeding subsequent processing operations such as model slicing. This paper discusses model topology and how to derive it given a collection of unordered triangular facets which represent a valid model.

Polarization and birefringency considerations in rendering

In this work we render non-opaque anisotropic media. A mathematical formalism is described in which polarization effects resulting from light/material interactions are represented as transformation matrices. When applying the matrices a skewing is performed to ensure that like reference coordinates are used. The intensity and direction of an extraordinary ray is computed.

Fitting a woven cloth model to a curved surface: dart insertion

If a woven cloth composite ply cannot fit a surface exactly, darts must be cut. Algorithms that define these darts for a CAD system resolve anomalies in the 3D ply.

Utilizing Topological Information to Increase Scan Vector Generation Efficiency

Demands for increased Solid Freeform Fabrication precision and speed suggest the need for advanced scanning techniques, such as boundary tracing, half-lap and multiple orientation scanning, or OintelligentO scanning. Since most SFF processes construct parts from parallel material layers, separating model slicing and scan conversion functions appears to be a powerful approach. Both can benefit from increased topological information. This paper addresses the issue of improving model slicing by utilizing topological data to increase performance, and consequently, improve the efficiency with which scan vectors can be generated.

A Flexible File Format for Solid Freeform Fabrication

A flexible file format for Solid Freeform Fabrication data is presented which significantly improves on the de-facto industry standard STL format. The new format removes the redundancy present in STL files and can contain topological information. Its specification flexibility allows users to balance storage and processing costs. Since facet boundary models currently provide the greatest common denominator for data exchange between many CAD systems, they are supported by this format. Additionally, representation of CSG primitives is provided, as are capabilities to represent multiple instances of both facet and CSG solids. Format extensibility, without obsoleting existing programs, is made possible by interleaving the format schema with the data. User data can be added to existing entities, or new entities can be created. This allows the addition of NURBS based geometries in the future.

A method for generating volumetric features from surface features

Form features may be represented both as surfaces and as volumes. Volumetric features are neeessary in automated process planning for relating a feature to the extent of material to be removed from a ~ and for capturing the global chamcteristics of a part, such as tool accessibility. In this paper, a method witl be presented for generating volumetric features from surface features, based on the technique of face extension. The implementation is currently limited to polyhedral parts. Examples will be given to illustrate the algorithm.