Design of textile structures: methodology and data

2 downloads 0 Views 2MB Size Report
the designer (i.e. more generally the engineering office) and the manufacturer are ... The third is centred on the structural analysis of the textile and metallic struc-.
PERGAMON

Computers and Structures 67 (1998) 309±317

Design of textile structures: methodology and data architecture P. VeÂron a,*, J.C. LeÂon a, P. Trompette b a

Laboratoire des Sols, Solides, Structures UMR-CNRS 5521, BP 53, 38041 Grenoble Cedex 9, France Laboratoire de BiomeÂcanique du Mouvement, Universite Claude Bernard, 69622 Villeurbanne Cedex, France

b

Abstract A new methodology dedicated to the design of textile structures is introduced. It is based on a collaborative environment to carry out the dialogues between the actors participating in the design process. A speci®c data architecture has been set up to allow ecient collaborative design loops, to manage the data and check their coherence during the overall design process. # 1998 Elsevier Science Ltd. All rights reserved.

1. Introduction Nowadays, textile structures occupy an increasingly important place in the civil engineering architecture. Their lightness and their aesthetic aspect justify their wide use to cover large areas such as platforms, halls of railway stations, entrances of buildings, areas used for exhibitions and meetings, etc. From the de®nition of the needs to the manufacture of the structure, the design of textile structures is a tightly coupled process. The architect draws a sketch of the three-dimensional shape of the textile structure in accordance with the customer's requirements and his own aesthetic criteria. From the stress state point of view, the 3D shape that has been de®ned must be under a biaxial tension everywhere. For manufacturing purposes, the 3D shape of the structure has to be divided into several 3D cutting patterns which must meet the architect's aesthetic criteria. Afterwards, a ¯attening process takes place to produce the corresponding 2D shapes which must satisfy the manufacturer's speci®c criteria. Such requirements explain why the current linear or concurrent design methodologies cannot be applied straightforwardly because the design process of textile structures is a set of tightly coupled steps. The method-

* Corresponding author.

ology proposed here divides the overall design process into three phases which make it possible to uncouple the de®nition of the 3D shape of the textile structure, its mechanical analysis and the de®nition of its manufacturing features. Di€erent actors such as the customer, the architect, the designer (i.e. more generally the engineering oce) and the manufacturer are involved in the design process. Several agreements between them are required and obtainable only through speci®c dialogue objects adapted to the objectives and to the knowledge of the participating actors. To this end, several views of the same structure are used at di€erent stages of the design process in accordance with the de®nition level of the structure and the subset of the actors involved. Moreover, because the location of these actors may be geographically di€erent, multimedia and collaborative tools are used to help them to work together. A speci®c data architecture has been developed to organize the data required at each step of the design process and to preserve their coherence during the numerous interactive modi®cations necessary to produce real collaborative design loops. The conversions between the models associated with the di€erent views of the structure are carried out automatically in order to allow ecient and robust treatments. Moreover, the speci®c knowledge associated to these models is not necessarily understood by the related users.

0045-7949/98/$19.00 # 1998 Elsevier Science Ltd. All rights reserved. PII: S 0 0 4 5 - 7 9 4 9 ( 9 7 ) 0 0 1 4 0 - 5

310

P. VeÂron et al. / Computers and Structures 67 (1998) 309±317

Fig. 1. The actors of the design process and the main data exchanged between them.

2. The actors of the design process and their objectives Several actors take part in the design of a textile structure. First of all, the customer expresses his needs. He often does not have any technical knowledge about the textile structure but his objectives are mainly economical, functional, and rely on the architect's proposals. The architect converts the customer's needs into a sketch of the textile structure, mainly, using his own aesthetic criteria. This sketch is the starting point for the work of the engineering oce which plays the central role in the current design methodology. The main actors of the engineering oce are the designer and the structural engineer. They have to ®nd a compromise between aesthetic, functional and mechanical criteria in order to produce a feasible textile structure. Finally, the manufacturer is in charge of the cutting and assembly of the textile strips. Fig. 1 illustrates the di€erent actors of the design process and the main data exchanged between them. 3. The design methodology: an integrated approach Owing to the multiple actors involved in the design process of a textile structure, an integrated approach has been set up to carry out ecient dialogues and to allow extended modi®cations of design parameters during the design process. Moreover, integration makes it possible to manage real collaborative design sessions. The integrated approach is centred on the tasks of the engineering oce and facilitates the con-

Fig. 2. User choices for the creation of the manufacturing features of the textile structure and the associated results.

vergence of the design process. It is proposed to divide the tasks of the engineering oce into three phases. The ®rst is related to the de®nition of the 3D shape of the textile structure and the second focuses on the de®nition of its manufacturing features. These two phases are part of the designer's task. The third is centred on the structural analysis of the textile and metallic structures and it is undertaken by the structural engineer. At ®rst, the 3D shape de®nition phase produces a topological model of the textile structure which is de®ned both from its ®xed parts (i.e. the anchorage points, hoops, bars and arches which maintain the membrane) and from the panels (i.e. topological meshes) created from the previous ®xed parts. Then, a polyhedral model which de®nes a 3D shape approximation of the textile structure, is computed using a method based on the calculation of the equilibrium position of a rigid bar network [1]. This method makes it possible to de®ne feasible shapes only. The initial bar network used to approximate the 3D shape of the textile structure is an assembly of all the mesh edges previously de®ned, therefore it possesses the same topology. The improvements of the shape thus obtained (i.e. to satisfy functional and aesthetic needs) can be worked out interactively and in real time on this polyhedral model. They allow the user to incorporate the e€ects of cables placed between two anchorage points, to change the form of the free boundaries, or globally tighten or loosen a part of the shape of the textile structure. Finally, the polyhedral model is automatically converted into a G1 continuous surface model approximation containing BeÂzier patches [2±5]. The surface

P. VeÂron et al. / Computers and Structures 67 (1998) 309±317

311

Fig. 3. The integrated approach of the design process with the three main loops and the di€erent actors.

decomposition into patches is related to the topological meshes initially created. Each patch is computed to provide the best approximation of its bar network sub-

set counterpart. The surface model thus de®ned is the basis of the subsequent phases. The software architecture makes it possible to generate a real design loop

312

P. VeÂron et al. / Computers and Structures 67 (1998) 309±317

Fig. 4. InPerson example.

where the user can intervene at each of the above stages to modify his anterior data or choices (see Section 5). The second phase focuses on the creation of the manufacturing features of the textile structure. The surface model of the textile structure previously de®ned is at the basis of this phase because it allows the results of the ¯attening technique to be independent from the parameters used during the ®rst phase (i.e. for example the discretisation used to build the polyhedral model). Thus, the 3D cutting pattern boundaries are directly de®ned over the surface model according to the requirements of the architect, the designer and the manufacturer choices (see Fig. 2). Then, an automatic ¯attening process produces the 2D contours that will be cut in the fabric roll. Here, the user can choose one among several methods ranging from fast to robust ones in accordance with the quality desired [6±8]. Improvements of the quality of the cutting patterns obtained may be required when their 2D shapes have global curvatures that are not

acceptable from a manufacturing point of view because they would generate large wastes of textile and/or reduce the load capacity of the structure. A direct and interactive correction process applied to the boundaries of 3D cutting patterns helps the user to modify the ¯attened 2D cutting patterns in the area of interest. Finally, the engineering drawings are made after a length adjustment of the di€erent pieces has taken place. Because the ¯attening process is carried out on an independent basis for each cutting pattern, the 2D boundaries obtained for two adjacent cutting patterns usually have di€erent lengths. This justi®es the length adjustment process previously mentioned. The two sets of the 3D and 2D cutting pattern boundaries constitute the manufacturing model of the textile structure. Like the ®rst phase, the second one makes it possible to go back to any of the previous states and to modify various parameters. The interactive and real time improvement of the 2D cutting patterns (see Section 5) is an important process to obtain the validation of the cutting patterns by the di€erent actors.

P. VeÂron et al. / Computers and Structures 67 (1998) 309±317

313

Fig. 5. Annotator example.

The third phase concentrates on the analysis of the mechanical behaviour of the textile structure. Here again, the surface model of the textile structure is used as an input parameter in order to ensure the independence of the data structures generated during the ®rst phase with regard to the mesh used for the ®nite element computations. The latter is a subset of a mechanical model for the membrane part of the textile structure. The ®rst step consists of checking the static equilibrium position of the textile structure. For this, a ®nite element (F.E.) mesh is created from the surface model and the stress state computation helps to verify the textile structure which should be under an adequate level of tension everywhere. Then, the boundary forces obtained from the previous computations are used for the dimensioning process of the metallic structure. Finally, the mechanical behaviour of the structure submitted to various load cases like the wind, the snow, etc., is studied. This helps to verify that the fabric membrane does not exhibit compressive stress states under such load cases and that the maximum strains

and displacements of the fabric cannot create baggy areas where water could accumulate. All the previous computations are carried out with a speci®c ®nite element software which uses an anisotropic material law, large displacement theory and initial stresses. A visualization module has also been developed to display the results under an adequate form. Fig. 3 illustrates the three phases of the design process and the associated actors.

4. The collaborative engineering architecture and the dialogue objects Seeking compromises or agreements between the actors at various stages of the design process through meetings and discussions requires a speci®c environment which includes CSCW (Computer Supported Collaborative Work) tools [9, 10] capable of exploiting computer objects to achieve collaborative work between actors [11].

314

P. VeÂron et al. / Computers and Structures 67 (1998) 309±317

Fig. 6. The collaborative engineering architecture.

The collaborative environment of the current approach is gained through the ability of the software to be combined with a communication and an annotation software that links together di€erent actors of the design process who are located at distinct remote sites. The various sites are interconnected through a computer network. These collaborations are carried out with a multimedia software tool InPerson2 developed by Silicon Graphics that allows the users to simultaneously share a 3D graphic scene displayed on their respective workstations. The collaborations between actors that have di€erent working themes require the use of dialogue objects which must be as realistic as possible and incorporate the data placed at the centre of the mediation [12]. The collaborative engineering sessions take place through a network connection between the actors involved in a common dialogue. Then, within such a computer environment, each of them can visualise and manipulate the same graphic objects within a shared viewer (see Fig. 4). Whereas InPerson allows synchronous communications between remote actors, Annotator2, also developed by Silicon Graphics, copes with asynchronous communication between actors. It allows each actor to annotate a 3D model with multimedia information. Thus, changes requested can be transferred between actors via electronic mail facilities. This way, the architect can annotate the 3D shape submitted by the designer through an OpenInventor ®le that is developed by Silicon Graphics Inc. to exchange graphic objects between computer applications (see Fig. 5).

In accordance with the design approach previously described, a ®rst dialogue is necessary to validate, with the architect and the customer, the feasible 3D shape of the textile structure de®ned by the designer. For this, the surface model of the textile structure is used as a dialogue object to support this discussion because it provides a realistic image of the future textile structure. Thus, the architect can appreciate the aesthetics of the structure. Moreover, the combination of the surface model with a complementary model of the metallic parts of the structure, and eventually its environment, gives the customer an ecient and realistic point of view of the whole structure. From a practical point of view, this model can be simultaneously visualised by the actors and it allows them to express their opinions and submit improvements. Afterwards, the designer carries out the modi®cations and submits a new shape. This task may be carried out within a few minutes using the design software, and transferred back into the communication software through copy and paste facilities that can be applied between the graphic objects of both software packages. Such functions create a very ecient dialogue environment. Two other dialogues must take place during the determination of the manufacturing features of the structure, to ®nd a compromise between the aesthetics of the future sewings and the 2D cutting patterns used for the manufacture of the structure. Two dialogue objects are used by the designer during thesecommunications. The ®rst one is the set of the 3D cutting patterns boundaries added to the surface model and used

P. VeÂron et al. / Computers and Structures 67 (1998) 309±317

315

Fig. 9. Data structure describing the surface model of a textile structure.

Fig. 7. Data structure describing the topological model of a textile structure.

to validate the shapes of the future sewing paths with the architect. The second object consists of all the ¯attened cutting patterns which are used to obtain the manufacturer's agreement in terms of maximum dimensions, contours and wastes. Here again, the transfer of such data from the design to the communication software is achieved through the copy and paste functions applied to graphic objects standing for the design objects listed above. Fig. 6 shows the two main collaboration loops previously described. They are associated with an illustration of each actor's point of view. These illustrations show the necessity of high level dialogue objects similar to those provided by the current architecture. 5. The data architecture A speci®c data architecture has been developed to manage the data involved at each step of the design process and to make possible extended modi®cations at every stage of this process and cope with the various

Fig. 8. Data structure describing the polyhedral model of a textile structure.

views required by the actors [13]. The identi®cation of the successive models for the same structure (see Section 3) are the ®rst elements of the architecture for the design software of a textile structure. Indeed, these models de®ne several views of the structure adapted to its de®nition level and to the actors involved in its design at a given step of its design process. The links between these models (or views) are carried out through automatic treatments which automatically convert the data from one model to another. These automatic treatments are the second aspect of the elements of the architecture setup. Interactive treatments are used to allow the user to de®ne the textile structure or improve its de®nition and thus, converge towards the goal of the design process which consists in ®nding a compromise between all the actors. Finally, several data interfaces are used to transfer data from the design software of the textile structures to the speci®c tool used for the mechanical analysis of the structure, and also to produce the drawings of the structure. The de®nition of the topological model, which is the starting point of the design process at the engineering oce, requires many interactive actions through the user interface. All the initial data and panels de®ned can be modi®ed at every step of the design process. This model is used by the designer to create several versions of a structure with di€erent geometric dimensions. Moreover, topological models of already built common textile structures can be used as standards in redesign processes. Fig. 7 describes the data structure associated with the topological model.

Fig. 10. Data structure describing the manufacturing model of a textile structure.

316

P. VeÂron et al. / Computers and Structures 67 (1998) 309±317

Fig. 11. Data structure links allowing extended modi®cations and data coherence checking of the complete textile structure model.

The conversion of the topological model into the polyhedral one is entirely automatic and consists of an assembly of all the topological meshes into a bar network, which owns mechanical properties. Afterwards, this model is also used by the designer to carry out interactive improvements of the 3D shape previously obtained. Fig. 8 describes the data structure associated with the polyhedral model. The conversion of the polyhedral model into the surface one is also completely automatic. Firstly, it involves the assessment of the decomposition of the structure into patches. Then, a parametrization of the nodes of the bar network is built prior to the construc-

tion of the BeÂzier patches which smooth the polyhedral model. Afterwards, several continuity conditions are generated to carry out the ®nal G1 continuous surface. Fig. 9 describes the data structure associated with the surface model. The mesh supporting the ®nite element computations of the membrane part of the structure is automatically built from the surface model and exported to the mechanical analysis software through a speci®c format. This mesh can also be adapted (i.e. in terms of quality for the ®nite element computations and with the use of a priori criteria) by a speci®c software [14, 15], before it is inputted into ®nite element computations. The ®rst part of the manufacturing model (i.e. the set of the 3D cutting patterns) is automatically built from the surface model in accordance with the designer's parameters (see Section 3). The boundaries of the cutting patterns are trimmed curves lying on the BeÂzier patches of the surface. Thus, the surface of one cutting pattern is a trimmed patch. The second part of this model (i.e. the set of the 2D cutting patterns) is completely linked to the ®rst one and is produced by an automatic ¯attening process using a method chosen by the user. The data structure describing the manufacturing model is shown on Fig. 10.

Fig. 12. Architecture of the design software dedicated to the textile structures.

P. VeÂron et al. / Computers and Structures 67 (1998) 309±317

Finally, the 2D contours of the cutting patterns are exported to the manufacturer as speci®c data ®les (DXF ®le format for example) directly used on numerically controlled machine tools. The links between the di€erent models are supported by their associated data structures. They allow the user to make extended modi®cations on the textile structure models. All the modi®cations generated through the interactive tools are propagated through every model in order to maintain and to check the data coherence of the complete textile structure odel. Therefore, the modi®cations become more ecient and can be carried out faster. For example, when the user modi®es the geometric parameters of a ®xed part, the associated panels are updated and in turn generate an update of the polyhedral model. The changes on the polyhedral model involve a surface model re-computation. Finally, the manufacturing model is also updated because of the surface model changes. Fig. 11 illustrates the links between the di€erent models through their data structures. The architecture of the software including the user interactions, the data exchange interfaces and the automatic treatments incorporated are summarised in Fig. 12. 6. Conclusion and future work The design methodology and the data architecture elaborated ®t into the design process of textile structures and incorporate the actors involved in this process as well as their know-how. The coupling between the geometry of the structure and the mechanical stress state of its fabric have been treated through a decomposition of the overall process into di€erent phases according to the knowledge of each actor. This decomposition has been worked out with the use of appropriate approximation methods and automatic conversions of the data between successive models. The identi®cation and adaption of the speci®c models (or views) of the structure to each of its de®nition level and to each of the actors using it, is a key point of the methodology introduced. The collaborative environment developed participates in the convergence of the design process and improves the eciency of the dialogues between the actors by the use of shared graphic objects and annotation tools. However, further progress can be made in this area through the use of true design objects to ease the modi®cations speci®ed during a dialogue. Future work will aim at developing an e€ective mechanical analysis view of the overall structure and at integrating this view within the whole design environment.

317

References [1] Scheck HJ. The force density method for form ®nding and computation of general networks. Computational Methods in Applied Mechanics and Engineering 1974;3:115±34. [2] BeÂzier P. Mathematical and practical possibilities of UNISURF. In Computer Aided Geometric Design, ed. R. E. Barnhill and R. F. Riesenfeld. Academic Press, London, 1974, pp. 127±152. [3] Coons SA. Surfaces for Computer Aided Design of Space Forms. In MIT Project MAC-TR-41. Massachussets Institute of Technology, Cambridge, Massachussets, 1967. [4] Farin G. Curves and Surfaces for Computer Aided Geometric Design. Academic Press, 1988. [5] LeÂon JC. ModeÂlisation et Construction de Surfaces pour la C.F.A.O. ed. HermeÁs. Paris, France, 1991. [6] Allera R. Mise en forme des structures textiles tendues, Ph.D. thesis, Institut National Polytechnique de Grenoble, Grenoble, France, 1992. [7] Galasko G, Trompette P. Calculation of cutting patterns using triangles. In Proceedings of the Euromechanics 334 International Conference on Textile Composites and Textile Structures. Lyon, France. 1995, pp. 327±337. [8] Shimada T, Tada Y. Development of a Curved Surface Using a Finite Element Method. In International Conference on Computer Aided Optimum Design of Structures: Recent Advances. Southampton, UK, 1989. [9] Saad M, Maher ML. Shared understanding in computersupported collaborative design. Computer-Aided Design 1996;28(3):183±92. [10] Gomez-Molinero V, Vilanova J, Grave M, Lang U. Use of Computer Supported Collaborative Work (CSCW) in Structural Analysis and Mechanical Testing of Spacecrafts. In Spacecraft Structures, Materials and Mechanical. Testing Conference, ESA/ESTEC Noordwijk, Amsterdam, 1996. [11] Favela J, Wong A, Chakravarthy A. Supporting collaborative engineering design. Engineering with Computers 1993;9:125±32. [12] Vinck D, Jeantet A. Mediating and Commissioning Objects in the sociotechnical process of product design: a conceptual approach. In the Third COST A3 Workshop on Management and New Technologies. Social Sciences Series, CCE, 1995. [13] Rosenman MA, Gero JS. Modelling multiple views of design objects in a collaborative CAD environment. Computer-Aided Design 1996;28(3):193±205. [14] Noel F, LeÂon JC, Trompette P. A new approach to freeform surface mesh control in a CAD environment. International Journal for Numerical Methods in Engineering 1995;38:3121±42. [15] Noel F. Mailleur auto-adaptatif pour des surfaces gauches en vue de la conception inteÂgreÂe, Ph.D. thesis, Institut National Polytechnique de Grenoble, Grenoble, France, 1994.