abstract surface modeling for concurrent form finding and

Abstract Surface Modeling for concurrent Form Finding andClass A Surfacing in Computer-Aided Design

Sebastian MisztalUniversity of AppliedSciences and Arts

Ricklinger Stadtweg 12030459, Hanover,

Germanysebastian.misztal@hs-

hannover.de

Ingo GinkelUniversity of AppliedSciences and Arts

Ricklinger Stadtweg 12030459, Hanover,

Germanyingo.ginkel@hs-

hannover.de

ABSTRACTAs the missing link between designers and engineers, we introduce a new abstract modeling approach forcomputer-aided design systems. In contrast to existing solutions, our strategy is less geometry driven and lessbased on low-level aspects like control points or mesh elements. Instead we operate on the idea of a hierar-chical modular concept with abstract components like categories, parts and the features relations. The wholesurface structure of the model is composed of abstract areas represented by meshes. This allows designers with-out engineering background to concentrate intuitively on the form finding process as they easily model abstractcomponents and automatically generate high quality CAD-freeform equivalents suitable for computer-aided man-ufacturing. During the phase of construction, we focus on the designers intent and guide him through this processto enrich the model with semantic information. The goal is to describe the models structure, such that the auto-matically generated freeform surfaces not only meet correct geometry but also mirror the internal configurationof the model. Compared to common practice where design changes and updates lead to the reconstruction of thewhole model, our system accomplishes these kinds of alterations automatically, based on the hierarchical modelconfiguration, derived from the designers intent. So our approach of an abstract modeler is much faster, closesthe gap between creative design changes and technical model construction and captures this in one contemporarysystem and workflow.

KeywordsComputer-Aided Design (CAD), Computer-Aided Manufacturing (CAM), Geometric Modeling, Interaction Tech-niques, 3D Modeling, Class A Surfacing, Shape Design, Object Representations

1 INTRODUCTIONDesigning and manufacturing a product incorporatesvarious tasks to be performed by professionals withpartly very divergent backgrounds. These can beroughly classified into designers with creative mindson the one hand and engineers with technical expertiseon the other hand.

Designers, who are responsible for form finding, oftenrefuse technologies like splines and NURBS. So theyusually rely on 2D-sketches, real clay models or virtualmesh models of the object they are creating.

Permission to make digital or hard copies of all or part ofthis work for personal or classroom use is granted withoutfee provided that copies are not made or distributed for profitor commercial advantage and that copies bear this notice andthe full citation on the first page. To copy otherwise, or re-publish, to post on servers or to redistribute to lists, requiresprior specific permission and/or a fee.

- CAD construction based on mesh model- CAD construction based on laser-scan- Direct CAD construction based on sketches

Construction (engineer)- Refine and adjust model- Enhance surface quality- Feed model into milling machine, deep drawing etc.

Explore model quality

- Export mesh-model (conversion)- Mill real clay model- Throw away model, start new sketch model

Styling (designer)- Clay model- 2D sketches- Virtual mesh model

Digitize shape: „strake“

- Curvature plots- Lightlines etc.

Alter design

Class A Surfacing (engineer)- Replica of model by engineer- Complete CAD construction- Incorporate construction issues

Figure 1: Common design / CAD workflow

ISSN 2464-4617 (print)ISSN 2464-4625 (CD-ROM)

WSCG 2016 - 24th Conference on Computer Graphics, Visualization and Computer Vision 2016

Short Papers Proceedings 41 ISBN 978-80-86943-58-9

brought to you by COREView metadata, citation and similar papers at core.ac.uk

provided by DSpace at University of West Bohemia

https://core.ac.uk/display/295592337?utm_source=pdf&utm_medium=banner&utm_campaign=pdf-decoration-v1

Engineers have to transfer the model into a manufac-turable model in a further step. Based on a real claymodel or sketches, the whole technical CAD-model hasto be created. In the case of a virtual mesh modelone has to state that meshes can not provide a suitablemathematical surface quality, thus a CAD-replica of theshape is also necessary.

So for all of the usual design approaches, a hand-craftedrecreation of a whole model into a freeform surfacerepresentation is everyday business and of course verytime-consuming. Even more serious is the problemwhen a designer wants to alter the shape of a modelafterwards and the engineer has to rebuild the wholefreeform model again. Reverse eingineering in differ-ent process steps became mandatory [Sok05]. The me-dia disruption between a model created by a designerand the one formed by an engineer and its impact to theoverall workflow is depicted in figure 1.

The task to overcome the media disruption is either toteach an engineer a creative mind or to provide design-ers a suitable modeling tool for designing and produc-ing a manufacturable model simultaneously. The goalof our approach is the latter one, providing the designeran intuitively usable tool which guides him through themodeling process and tries to capture the design in-tent. Since neither meshes nor spline representationsare suitable for a designer to simultaneously shape thedesired object and create a manufacturable model, acombined or hybrid approach is necessary.

The idea is to build a system which allows modeling ona less detailed but hierarchical abstraction level. Themodel will both cover a mesh model and a spline rep-resentation but neither of them will be modified di-rectly in the sense of low level polygonal transfor-mations, control point movement, degree selection orparametrization changes. This kind of construction istoo much a low level approach and distracts the de-signer from his main task, namely to find the overallshape of the object.

Instead we provide the user with an abstract surfacemodel consisting of abstract surface areas, hierarchicaldependencies and enrich this model with semantic in-formation which will be queried from the user or canimplicitly be derived from the context (i.e. the usedconstruction tool, moment and region where and whenit is applied). Topological operations will not be per-formed on meshes or spline surfaces, but on an abstractlevel containing adjacency between abstract areas freedfrom low level aspects like mesh consistency or curva-ture continuous transitions.

Based on this abstract model, a mesh surface for pre-view purposes will automatically be generated deliver-ing a fast feedback. In contrast to a usual mesh model,additional sematic information, design intents etc. arestill stored in the abstract model. After some iterations

Hybrid Model (designer)- Virtual abstract surface model containing geometric and semantic information- Generated by using given toolset- Use meshes for preview, generate CAD model on demand

Alter design

- Use given virtual toolset- Change given model

Construction (engineer)- Refine and adjust model- Enhance surface quality- Feed model into milling machine, deep drawing etc.

Explore model quality- Curvature plots- Lightlines, etc.- Verify structure of model

Figure 2: Intended design / CAD workflow

of design changes and form finding, this informationcan be used to automatically create a class A surfacemodel which can be used as input for the constructionphase.

Eliminating the media disruption between the design-and the manufacturing-model (i.e. designing and con-structing in a joint hybrid model) the overall workflowcan be simplified substantially, as shown in figure 2.

2 MODELING CONCEPTS ANDTECHNOLOGIES

Modeling objects using virtual construction tools ba-sically allows three conceptually different approaches.The first one are mesh-modelers, providing the gen-eration of simple geometric shapes. Another solutionare CAD-construction tools using freeform surfaces forclass A surfacing. The third concept are solid model-ers, focussing on solid primitives, building constructiveelements. In any of these configurations a designer andan engineer still produce the before mentioned mediadisruption.

Most CAD/CAM tools focussing on the media dis-ruption between design and construction phase tryto simplify and automate the conversion between thegenerally used representations. This also holds for3D sketching approaches like in [Zel06] or [Die04].These systems are trying to guide the designer througha model finding process, but still require a conversionof the design- (i.e. the sketches) into a construction-model. Using a converter is of course much fasterthan hand-crafting the model again for construction,but converters often demand adaptions after everyconversion. Additionally with every conversion,valuable information get lost. Converting from adesign-representation to a construction model, designintents and specific shapes can be sacrificed for tech-nical restrictions. Converting a construction modelinto a design model often causes a loss of structuralinformation.

To recover these informations various startegies are pur-sued, e.g. shown in [Han00]. Reverse engineering ap-




proaches traverse the history-tree or -graph and thus allrecorded modeling operations and interpret their con-stellation to extract valuable meta information from it.But this requires an expressive tree- or graph-structure,holding the desired intent. Hint-based solutions analysegeometrical characteristics which also only qualify forcapturing geometrical features. Knowing the pure fea-tures without any context is insufficient knowledge tocreate an overall view of the designers purposes. More-over, relations are often not cosidered.

Concepts like [Li10] divide the whole modeling ar-rangement into smaller sub-parts to find symmetrieswhich makes it easier to detect features and the designintent. Conceivable is also to split the history tree intolittle pieces too. The limitations of history-trees is thefact that they map a chronological sequence. In con-trast, our idea is independent from the order of the op-erations by just using the structural and hierarchical in-formation.

Establishing a modeling methodology like [Bod14]with advanced rules, predetermining the sequenceof types of operations during the modeling process,lead to a clearer history-tree and a better mapping ofrelations between elements and thus to better condi-tions for mentioned intent recognition approaches. Butthis methodology still requires a strong contributionof the designer and is less technology driven likethe strategy of [Ald12] and [Dor13]. These attemptsdemand the designer to make annotations, describingtheir design intent which is helpful in further modelingsteps especially for other designers, working with anunfamiliar model. Unfortunately these annotations areinterpreted by human beeings and not by a machine.

Machine-interpretable technologies to capture the de-sign intent are given in [Kim06] by reasoning which isontology-based. Spatial relationships can be stored ina relational model. But the focus here lies on assemblydesign which is difficult to adapt to classical productdesign issues.

During the evolution of CAD systems, different gen-erations and paradigms emerged, dealing with the cap-turing of design intents in the broadest sense, e.g. pre-sented in [Tor10]. Parametric technologies and feature-as well as history-based systems in solid modelers arethe fundamental idea, capable to describe the modelprecisely enough to map geometrical relations and de-pendencies. One of the major drawbacks is the steeplearning curve. Direct or also explicit modeling toolshave the advantage of beeing very simple, easy to learnand fast. But the absence of a construction history andmissing relations require reverse engineering to capturefeatures and design intents.

So today vendors tend to incorporate both approachesto hybrid solutions in its literal sense. That meansthat these concepts are not completely combined. In-

stead they often operate synchronously where the de-signer can switch between them like in Siemens NXand SolidEdge. PTC Creo Parametric also allows theadding of constraints to the model during direct model-ing. But our goal is to create high quality freeform sur-faces straight from the explicit modeling object. Otherprofessional software like Autodesk Fusion 360 pro-vide tools like snapping which allows the user to cre-ate the freeform surfaces directly on the mesh but stillbeing a time-consuming low-level attempt. Declaredconverters like Kontenpunkt PointMaster produce cor-rect freeform surfaces but neither capture the design-ers intent nor provide an internal model structure thatis suitable for production without further handwork.Solid modelers like SolidWorks provide very nice in-ternal structures of the model, containing defined rela-tions between parts. They also provide precise shapesof primitives based on implicit representations whichare previewed by meshes. Their weak spot is the re-striction to fairly simple surfaces based on the solids.So creating a real freeform surface with a solid model-ing tool is very challenging.

To construct our system, we borrow ideas from eachof these approaches. We use meshes for preview, para-metric surfaces for construction models and an internalstructure of abstract regions that are aligned and interactvery similar to a solid modeling concept. This tool willallow arbitrary surface shapes, explicitly modeled in-ternal dependencies and hierarchies built by a designerwith limited technical skill or interest. The goal of ourapproach is to guide and help the designer to intuitivelyfind the form or shape of an object. Simultaneously weneed to establish the preconditions that allow an auto-matic retrieval of construction data, namely a high qual-ity CAD model covering both the shape and internalstructure (part groups, hierarchy etc.).

3 ABSTRACT MODELEROur abstract modeler is an extending modelingparadigm which is theoretically transferable to vari-ous modeling systems. We integrated our technicalimplementation into the CAD-system Rhinoceros 5as a plug-in. We used the existing surface represen-tations and tools as low-level objects and operationsrespectively and created our own high-level objectsand operations upon them. In this section we introducethe basic concepts of our paradigm and describe thestructure of our solution by an exemplarily workflow.

3.1 Modeling ElementsMeshes are consistent structures where the features arecharacterized by an absence of mesh elements or bya specific geometry or constellation of vertices. Thatmeans that features in a common mesh are not just in-tegrated into the whole construct, they are melded into




the mesh. Freeform surfaces expose their features in asimilar manner. They are defined by the mathematicalrepresentation of the surfaces and finally a geometricaldeclaration. Features in both technologies basically donot differ from the rest of their object, as the border be-tween a feature and the rest of the model is fluent. Anautomatized way to identify a feature would be very ex-pensive as they have to be retrieved afterwards by usingcomplex algorithms. And there is no guarantee that allfeatures and the hard boundaries between a feature andthe rest of the model will be found.

Nevertheless our system is based on the idea of a mod-ular concept with separable elements with a clear dis-tinction between them. The following properties stateour requirements to these elements:

• Each feature or part of the model is an individual,distinct and nameable model-element.

• Each model-element can be addressed and selected.

• A model-element has a designated state, carryingmeta information.

• The meta information of a model-element preciselydescribes its own geometry and its relation and con-nectivity to other model-elements.

• A model-element is a module, which can be ex-changed and integrated into another model.

• The consistent surface of the model is formedthrough the connections of a set of model-elements.

• Operations can either be performed on an elementindividually or by a related set of elements usingtheir meta information and geometry.

As mentioned before, common surface elements likemeshes and freeform surfaces are not the fundamentalobjects in our attempt to describe a model. The ele-ments we use are an abstract generalization not onlystoring geometrical information. The visual shape of amodel is composed of the two model-elements additiveand area.

Definition 1 (Area) An area is a distinct, boundedarea, lying directly on the geometrical surface of amodel. Together connected, areas form the outer hullof a model.

Areas are the basis element comparable to faces ofmeshes or freeform surfaces in typical CAD-models butwith fundamental differences. They do not form thewhole model. Instead they shape the coarse form of itand function as a carrier for other elements. In practicalapplication, areas not only have the task to structure amodel but also to organize other elements of the model.To describe the whole model, there is a further elementcalled additive.

Definition 2 (Additive) An additive is a combinationof arbitrary geometrical elements. As a closed unit, ad-ditives are modules with the ability to be attached toareas. An additive itself can again be composed of amodeling structure with areas and additives.

Additives represent a self-contained feature. In practi-cal application, whenever a designer wants to create apart of the model which can be named and modeled onits own, detached from the rest, he would create a newadditive. As areas only describe the coarse parts of themodel, additives are used when fine sections have to becreated.

Definition 3 (Layer) Each area can be layered. Alayer is a specific area with its attached additives ina system with under- and overlaying other areas withtheir additives. A layer has a level, synchronous to itstime of creation.

Basically a layer is not more than a specific decoratedarea instance which can be exchanged by other layers.In practical application, several layers are used whenthe designer wants to experiment with different shapesand additive constellations. As each layer has a level,according its chronological creation, the designer cantravel the progress time by switching the level of thelayer. The unique characteristic here is the fact thatindividual parts of the model can be layered and tra-versed. Thus it is possible to assemble and modify amodel with parts of several progress steps.

Definition 4 (Base Object) The base object is a com-pound of not overlapping areas, on not necessary equalleveled layers, forming a gapless (and solid) surfacestructure. Each not overlapping (and solid) constella-tion of areas forms a valid base object, representing oneinstance of the models surface.

Definition 5 (Model) A model is the entirety (set) ofall model-elements (additives and areas), layers, rela-tions and operations, allocated to this model.

Figure 3: Engine hood of a vehicle with attachments.




Figure 3 shows a part of a vehicle with different model-elements. The surface of the model is constructed ofareas (colored in yellow), all together forming the baseobject. Different shades of yellow indicate differentlayers. The spare tire and the headlights are designedas additives (colored in red and complete models them-selves) attached to the engine-hood-area. The connec-tivity of the spare tire to the engine hood is highlightedin green. The position of the tire can be adjusted by us-ing control points (also in green). Moving these controlpoints alters the areas geometry as the hole for the tireis moved too. Control points at the border of the mid-dle area adjust the size of this area which in turn hasinfluence on the tire depending on the relation betweenthem. Varieties of this impact are illustrated in figure 4.

3.2 Meta InformationAs all model-elements are defined as separable objects,relations are the connecting element between them, cre-ating the consistency of a model.

Definition 6 (Relation) A relation is a meta informa-tion, describing the state between two model-elementswith the following parameters:

• The connectivity, including the type (loose, fixed,fluent, static etc.) and adjacency.

• The dimension, including the sizing information.

• Distances, including absolute and relative lengths.

• Repetition, including the mirroring of elements us-ing determined distances.

• The rank, including priorities of elements and re-lations.

In practical application, relations come into play whenthe model or single parts of it are modified. These rela-tions describe the behavior of linked elements and themodified object itself.

Definition 7 (Operation) An operation is an action di-rectly applied on one or more model-elements, areasand/or additives. Operations are all kinds of transfor-mations on an element, not changing its relations. Re-lations of the considered and related elements are trig-gered through an operation. Operations aggregate thetype of transformation, its location and the applied toolwith parameters.Unlike relations, an operation is not a determined andfixed state. Operations are sequential and retraceableactions where the order matters. In practical applica-tion, operations are used to shape an element.

Definition 8 (Mode) A mode is a state, chosen by theuser, defining the impact of an operation on the modelby redefining its relations.

The idea is to attain different effects with the same op-eration just by switching the mode. Theoretically oursystems allows an unlimited number of modes, somepre-defined by the system, further modes can be cre-ated by the user himself. We introduce two fundamentalmodes, also used in our example. The geometry-modeis a state where geometrical operations on a selected el-ement have pure and isolated geometrical impact onlyon this particular element. The designer is performingplain low-level geometrical modifications on separatedparts of the model without further impact on other parts.This is the kind of behavior which a designer would ex-pect as he knows it from familiar modeling tools. Herethe geometry-mode is mostly used for shaping the outerform of the model or isolated features. The semantic-mode is a state where geometrical operations on an el-ement not only modifies its geometry but also all partsof the model which are semantically connected to it. Asdiscussed before, adjacent elements of the model can bemutually related. These relations are used to determinethe effect of an alteration.

m n

x*m x*n

m n

m n

m n

Figure 4: Tire as an additive, attached to an area. Left:Initial state. Right: Various impacts on the tire afterresizing the area depending on their relations.

Figure 4 illustrates a tire on an area, similar to the ex-ample in figure 3. The tire is attached through relationsbetween the border of the area and the border of the tirewith defined distances m and n. When the area is en-larged (right column), the impact on the attached tire




depends on the predefined relations between them, par-ticularly the parameters of the connections. The resiz-ing of the area results in a resizing of the tire by satis-fying the absolute distances m and n and its shape (firstimage), in a deformation of the tire by satisfying theabsolute distances m and n (second image), in a trans-lation of the tire by satisfying the relative distances mand n and its shape (third image) or in a mirroring ofthe tire by satisfying the absolute distances m and n andits shape (fourth image).

3.3 Mesh DisplayOur surface representation is an abstract constructwhich needs to be displayed in some way. So to renderour abstract surfaces we use polygonal meshes createdby tessellation techniques from Delaunay [DeB08] andextended methods from Chew [Che87] and Shewchuk[She05] called Constrained Delaunay. This kind oftessellation creates triangles whose interior angles alltend to have the same size as they vary just a minimumuser defined value from π

3 . The result are homogenouslooking meshes which is not only an advantage duringrendering. The structure of the faces of these meshesalso leave a more valuable impression for the designer.

Besides the pure display and tessellation, our systemalso uses subdivision techniques from Catmull-Clark[Cat98] and Loop [Loo87] and furthermore extendedmethods from Ginkel [Gin06]. Curved surfaces are cre-ated by subdividing the mesh representation of the ar-eas to be shaped. Therefor the abstract area is deco-rated with an subdivision operation. These operationsare editable and exchangeable. Other subdivision meth-ods can be performed on the area by changing parame-ters.

Figure 5: Blended transition area.

Figure 5 exemplarily shows the creation of a transitionbetween areas using Catmull-Clark subdivision tech-niques. The designer can expand this blended areathrough four control points at the corners and alter itsroundness through another control point in the mid-del. The border is highlighted in green color. Ordi-nary meshes or freeform surfaces would provide a muchlarger number of vertices or control points respectively.We use just five. The techniqual details are hidden fromthe designer to reduce complexity.

3.4 Micro Modeling WorkflowTo create a model with elements satisfying the beforementioned characteristics, we imply this coarse model-ing process in three steps:

1. First of all the designer has to be aware of what ex-actly he wants to do in each modeling step. There-fore our system guides the designer by showing pos-sible workflow sequences and suitable previews ofall operations.

2. After the designer disclosed his decision, each ac-tion must then be declared by him. Not by perform-ing low-level operations but instead by triggeringabstract high-level operations which are capable toclassify the designers action and to record the neces-sary meta information. These high-level operationscan achieve the same geometrical result as familiarlow-level operations. But they have a different struc-ture and procedure where the designer instructs thesystem to do the operation and does not do it by hisown. These operations are meaningful and summa-rize a set of single actions.

3. Because we do not want the designer to state all nec-essary meta information by himself, which wouldend in a long querying sequence, the system re-sponses high-level actions with appropriate defaultsolutions. Afterwards the designer can adjust thesesolutions by altering the meta information interac-tively.

During the whole process a lot of meta information iscollected but not all of it is entirely provided by the user.A lot of it arises implicitly by choosing a tool, by apply-ing an operation or by appropriate default settings. An-other speed-up comes from the interchangeability andautomation of our workflow, when the designer deco-rates new or empty parts of the model with already ap-plied operations and previously stored categories.

In our interpretation of a modern CAD-system therole of the designer changes. Unlike common systemswhich are often geometry driven, our concept does notfollow the What You See Is What You Get approachentirely. By restricting low-level attempts on the




surfaces geometry and emphasizing the use of abstractoperations, the designer becomes more an instructor.We picture the working procedure of him more likedrawing a construction plan and less in forming eachsingle shape and feature manually. Besides the creativeform finding process, we mainly want the designerto enrich the model with semantic information whichdistinguishes us from other modeling systems. Inthe following we will show how this can work in anexemplary workflow.

3.5 Exemplary WorkflowIn this section we describe the concept behind our ideathrough a typical workflow by means of the example ofa washing machine. Going through all modeling steps,we introduce all kinds of elements, operations and ourfundamental layer system. The model was kept delib-erately simple and could also be created by a construc-tive solid geometry (CSG) system. But our methods areconceived with the aim to construct freeform surfacesequivalent to the model.

Figure 6: Top-left: Base object; Top-right: Bendedfront area; Bottom-left: Front-area divided into panel-area (overhead) and door-area (below); Bottom-right:Panel-area divided into panel- and detergent-area. De-formation on door-area for the detergent dispenser anddeformation on the top-area where the transition be-tween the top-area and the rest was blended.

Modeling Structure

The structure of our model is designed as a modularsystem with the main focus on interchangeability. Ontop of that system and at the very beginning of theworkflow, there is the base object, the initial instance towork with, representing the coarse shape of the desired

model. In our example the base object is a cube (figure6). Each of its sides represent a specific user definedpart on the surface, the areas. An area can be shapedlike shown in our example where the designer bendedthe front area of the washing machine. From then ona bending operation is assigned to the front-area. Butthe main aim of areas is to organize the elements uponit. Therefor the designer can divide an area – like thebended front-area – into two separated areas. Each ofthem can then be modified individually like the door-area which was dented for the detergent dispenser.

Attaching Additives

Figure 7: Attached additivescolored in red.

Furthermore an areacan host supplemen-tary features, the ad-ditives. Such an ad-ditive can be an en-tire model itself witha base object and ar-eas or a plain geomet-rical object, dependingon its complexity. Inour example (figure 7)the panel-area is dec-orated with additivesrepresenting a display,

some buttons, a rotary control and a coating for the de-tergent dispenser. Other additives are attached to thedoor-area for the door and the doorhandle.

Relations

The areas are connected through relations by default.All attached additives have at least one relation to itsunderlying area or other additives, describing its con-nectivity and behavior. In figure 8 the relation betweenthe rotary control and the panel-area is displayed. Thisadditive is connected to its area with a static connection,which means that in case of a deformation of the area,the distance between the additive and the areas borderstays constant. The buttons on the panel-area are con-nected to the areas border and with each other. Controlpoints adjust distances and the location of attachment.

Layer-System

Apart from their geometrical form, and their function asa host for additives, areas are predestinated to organizethese additives by grouping them. But areas can alsobe layered. Like shown in figure 6 the door- and thepanel-area were layered upon the front-area which isstill available and not deleted. Moreover these layeredareas are logically connected. The designer can flickthrough these layers and modify a certain one whichhas direct influence on the other layered areas automat-ically. This gives him the ability of saving a snapshot




Figure 8: Relations between additives and their areawith control points.

of a concrete part of the model for testing purposes orother design playthings. So different layered areas canbe shaped differently and decorated with different ad-ditives, fully modular. As these areas can easily be ex-changed, the designer can try out various constellationsof the model in little amount of time.

This is another crucial distinction to other modelingsystems. Going back in time in our solution meanschanging a layer and not changing the memory state.Thus a modification in the past has automatic influencein the future because of the relations between the lay-ers. Layers give the designer the opportunity to justtime travel selected parts (e.g. the front of the washingmachine) and leave the rest of the model untouched. Sovarious states of operations from various points in timecan easily be combined.

Modeling ModesThe impact of an operation also depends on the cho-sen modeling-mode. In figure 9 the designer decided tomodify the shape of the panel- and the door-area. Onthe left picture we see the old model. The right pictureshows that he wanted the border between these two ar-eas less curvy. So he modified the geometry of the bor-der curve which is the coupler between both areas. Be-cause this modification is conducted in semantic-mode,not only the geometry of this particular element waschanged, also all relations to this curve are involved, in-cluding the adjacent areas and their attached additives.This means that obviously both areas were altered. Butalso the door was relocated, still fitting the same dis-tance to the border and still retaining its size. Samegoes for the buttons. The display was resized as theusers intent defined it has to satisfy a fixed distance tothe borders of the panel-area.

Figure 9: Semantic-mode: Altering the shape betweentwo areas. Left: Old shape with a curvy coupler be-tween the areas. Right: New shape with a less curvycoupler.

Figure 10: Geometry-mode:Geometrical modification onthe doorhandle.

Figure 10 shows adoorhandle which wasmodified in geometry-mode (left: old state,right: new state). Thisoperation has no im-pact an other elementsand the alteration onthis additive was onlyapplied on its geom-etry. Here we cansee both modes in con-trast and how the sameoperation in differentmodes varies in theirimpact. In figure9 the same operationin another mode alsochanged the whole ar-rangement of the addi-tives. Whereas an operation in semantic-mode incorpo-rates all gathered meta information, the geometry-modeonly considers the pure geometry.

Figure 11: Semantic-mode: Tilting the panel area.




Another use case is shown in figure 11 where the panel-area was tilted and the attached additives on that areawere automatically tilted too. The rotary control wasautomatically extruded, fitting its basis plane from be-fore the modification.

4 SUMMARY AND FUTURE WORKWe have presented a modeling approach that enablesa designer to construct a 3D model by building shapesand internal structures on an abstract level. Previewsare accomplished by meshes and subdivision tech-niques while derivability of a CAD model based onthe hierarchical structure of the abstract model is stillguaranteed. The designer is equipped with constructiontools and guided through the process of creating anobject. For now we have focussed on the creation of themodel starting from a base model. The presented toolswere mainly meant to add features, both in geometricand in an abstract sense.

In the future we are going to extend the functionalityby tools that modify the shape of a surface area. Firstto mention there is deformation. Neither control-pointmoving in a spline sense nor mesh deformation tech-niques in the sense of smoothing operators will be suit-able to be used by a designer. Again we need a more ab-stract and non-technical view to the task. A surface de-formation tool must be imaginable in a designer-contextand could for example be interpreted as adding or re-moving clay from a certain surface region or to applypressure to a rubber surface. Depending on which levelour hierarchical model is attached to the surface area,we need to impose characteristics and restrictions to theallowed operations and of course to transfer the abstractmodeling operator into a low level mesh and/or splineequivalent. Again geometric deformation and structuralaspects will have to be executed simultaneously on theabstract model, the mesh and the spline surface-model.

5 REFERENCES[Ald12] Alducin-Quintero, G., Rojo, A., Plata, F., Her-

nandez, A. and Contero, M. 3D Model Annotationas a Tool for Improving Design Intent Commu-nication: A Case Study on its Impact in the En-gineering Change Process. Proceedings ASME45011, Volume 2: 32nd Computers and Informa-tion in Engineering Conference, 2012.

[Bod14] Bodein, Y., Rose, B. and Caillaud, E. Explicitreference modeling methodology in parametricCAD system. Computers in Industry 65 (1), 2014.

[Cat98] Catmull, E. and Clark, J. Recursively Gener-ated B-spline Surfaces on Arbitrary TopologicalMeshes. Seminal Graphics, 1998.

[Che87] Chew, L.P. Constrained Delaunay Triangula-tions. Proceedings of the Third Annual Sympo-sium on Computational Geometry, 1987.

[DeB08] de Berg, M., Cheong, O., van Kreveld,M. and Overmars, M. Computational Geome-try - Algorithms and Applications, Third Edition.Springer, 2008.

[Die04] Diehl, H., Mueller, F. and Lindemann, U.From raw 3D-Sketches to exact CAD productmodels - Concept for an assistant-system. Pro-ceedings of the First Eurographics Conference onSketch-Based Interfaces and Modeling, 2004.

[Dor13] Dorribo-Camba, J., Alducin-Quintero, G.,Perona, P. and Contero, M. Enhancing ModelReuse Through 3D Annotations: A Theoreti-cal Proposal for an Annotation-Centered De-sign Intent and Design Rationale Communication.ASME 2013 International Mechanical Engineer-ing Congress and Exposition, 2013.

[Gin06] Ginkel, I. and Umlauf, G. Loop subdivisionwith curvature control. Eurographics Symposiumon Geometry Processing, 2006.

[Han00] Han, J., Pratt, M. and Regli, W.C. Manufac-turing Feature Recognition from Solid Models:A Status Report. IEEE Trans. Robotics and Au-tomation 16 (6), 2000.

[Kim06] Kim, K.-Y., Manley, D.G. and Yang, H.Ontology-based assembly design and informationsharing for collaborative product development.Computer Aided Design 38 (12), 2006.

[Li10] Li, M., Langbein, F.C. and Martin, R.R. Detect-ing Design Intent in Approximate CAD ModelsUsing Symmetry. Computer-Aided Design 42 (3),2010.

[Loo87] Loop, C.T. Smooth Subdivision SurfacesBased on Triangles. Master Thesis, Universityof Utah, 1987.

[Mun03] Mun, D., Han, S., Kim, J. and Oh, Y. A setof standard modeling commands for the history-based parametric approach. Computer-Aided De-sign 35 (13), 2003.

[She05] Shewchuk, J.R. General-Dimensional Con-strained Delaunay and Constrained Regular Tri-angulations, I: Combinatorial Properties. Discreteand Computational Geometry 39, 2005.

[Sok05] Sokovic, M. and Kopac, J. RE (reverse engi-neering) as necessary phase by rapid product de-velopment. Journal of Material Processing Tech-nology, 2005.

[Tor10] Tornincasa, S. and Di Monaco, F. The Fu-ture and the Evolution of CAD. 14th InternationalResearch/Expert Conference, 2010.

[Zel06] Zeleznik, R.C., Herndon, K.P. and Hughes,J.F. SKETCH: An Interface for Sketching 3DScenes. ACM SIGGRAPH Courses, 2006.




abstract surface modeling for concurrent form finding and

Documents