Advertisement

Nexus Network Journal

, Volume 16, Issue 1, pp 69–87 | Cite as

Design and Fabrication of Free-Form Reciprocal Structures

  • Dario ParigiEmail author
  • Poul Henning Kirkegaard
Research

Abstract

Due to their non-hierarchical nature, the geometry of reciprocal assemblies cannot be described conveniently with the available CAD modelling tools or by hierarchical, associative parametric modellers. The geometry of a network of reciprocally connected elements is a characteristic that emerges, bottom-up, from the complex interaction between all the elements’ shape, topology and position, and requires numerical solution of the elements’ geometric compatibility. A computational method, the “Reciprocalizer”, has been developed by the authors to predict and control the geometry of large networks of reciprocally connected elements, and it has been now perfected and included in a larger procedure that can be regarded as an extremely flexible and capable design tool for the generation of free-form reciprocal structures. The design tool has been applied for the design and realization of a free-form structure composed of 506 round, un-notched wooden elements with a diameter of 22 mm. This paper focuses on the geometry of reciprocal systems and the unique issues of fabrication posed by such assemblies.

Keywords

Reciprocal frames Free-form Geometry Fabrication 

Introduction

Reciprocal structures have been studied and used in the past for different needs and purposes, and their presence throughout history is scattered and discontinuous (Pugnale and Kirkegaard 2011). Recently they have been classified along other woven structures such as knitting, fabrics and basket works. This is due to the fact that the elements forming them are interwoven with one another, with the peculiarity that reciprocal structures use elements which are stiff and short compared to the size of the entire structure (Baverel and Popovic 2011). In the world of construction, the application of the principle of reciprocity requires:
  1. i.

    the presence of at least two elements allowing the generation of a certain forced interaction;

     
  2. ii.

    that each element of the composition must support and be supported by another one;

     
  3. iii.

    that every supported element must meet its support along the span and never in the vertices (Pugnale and Kirkegaard 2011).

     

When a superimposition joint is used (i.e., un-notched bars sit on the top and on the bottom of each other, as shown in Fig. 4a or 5a) reciprocal structures develop naturally out-of-plane because the elements’ axes are not aligned; for this reason they can be defined as intrinsically three-dimensional.

Due to their non-hierarchical nature, the geometry of a reciprocal assembly is extremely difficult to predict and control, and it cannot be described with available CAD software or by hierarchical, associative parametric modellers. The geometry of a network of reciprocally connected elements is a characteristic that emerges, bottom-up, from the complex interaction between all the elements shape, topology and position.

The three-dimensionality can be regarded as a design opportunity, for the possibility to create three-dimensional configurations with simple joints and standardized elements. A computational method called the “Reciprocalizer” has been developed by the authors to predict and control the geometry of large networks of reciprocally connected elements (Parigi and Sassone 2012; Parigi and Kirkegaard 2013, 2014).

In the first part of the present article, the peculiarities of reciprocal structures geometry are presented, a tentative classification of different configurations on the basis of the joint type is discussed, and an explanation is provided of the way the Reciprocalizer handles those and the morphological characteristics of the resulting assemblies. In the second part, we deal with the issues of fabrication and the solutions devised for an efficient fabrication process. In the third part, the entire process from design to fabrication of a free-form reciprocal structure composed of 506 round, un-notched wooden sticks of 22 mm of diameter (Fig. 1) with a particular focus on the fabrication and the passages from digital model to the prototype (Fig. 2), is described.
Fig. 1

a Starting free-form geometry, b reciprocal mesh generated to closely fit the starting geometry, c the geometry and the mesh overlapped

Fig. 2

a Digital model of the three-dimensional reciprocal geometry and b realized prototype of the three-dimensional reciprocal geometry

Geometry of Reciprocal Structures

As we said, reciprocal structures that use a superimposition joint are intrinsically three-dimensional. The extent and direction of the out-of-plane deviation depend on the values of three geometric parameters necessary to describe each connection between two elements b i and b j .
  1. i.
    the engagement length l ij , which measures the position where each element is supported along the supporting element (Fig. 3a)
    Fig. 3

    a Engagement length lij and b effect of engagement length in the assembly morphology

     
  2. ii.
    the eccentricity e ij , which measures the distance between elements axes, directly dependent on the elements thickness and shape (Fig. 4a);
    Fig. 4

    a Eccentricity eij and b effect of different eccentricity values in the assembly morphology

     
  3. iii.
    the specification of whether element b i sits on the top or on the bottom of element b j with respect to a reference vector r j whose tip indicates the top position (Fig. 5a).
    Fig. 5

    a Top/bottom position and b effect of different combinations of top/bottom support positions in the assembly morphology

     

It is possible to observe the effect of changing each of the geometric parameters in the morphology of a three-element fan (Figs. 3b, 4b, 5b).

The out-of-plane deviation increases with the increasing of the eccentricity value e ij ; on the contrary, for large values of the engagement length l ij , the out-of-plane deviation is low, and it greatly increases if the engagement length value is close to zero.

The top/bottom position involves a topological change in the nature of the joint. In a three-element fan, four possible combinations of top/bottom positions determine four distinct characteristic geometries. When all elements are placed on the top or in the bottom of each other respectively a dome-like or reversed dome-like geometry is created. On the other side, when elements connect with mixed top/bottom positions, geometries that will prove to be fundamental for the creation of free-form shapes are created.

As a consequence of the non-hierarchical nature of reciprocal assemblies, those parameters are not independent, and any change in one parameter value should be followed by the simultaneous adjustment of all the other parameters values in the assembly in order to maintain the geometric compatibility. This is why the geometry of a reciprocal assembly cannot be conveniently described with available CAD modelling tools or by hierarchical, associative parametric modellers. The geometry of a network of reciprocally connected elements requires the use of iterative numerical methods to solve the physical interplay between all sticks in the assembly.

Physical models can be used at an exploratory level, to discover new configurations (Parigi and Pugnale 2014). However, the use and development of the typology requires a design tool that allows for predicting and controlling the geometry of large networks of reciprocally connected elements. The Reciprocalizer mentioned earlier is capable of handling different families of reciprocal systems based on the joint type.

Classification and Morphology of Families of Reciprocal Structures Based on the Joint Type

In the simplest possible reciprocal configurations, each locally supported element sits on the top of the supporting element with a superimposition joint. When structures are composed with this rule, the friction of the joint is sufficient to keep them stable, and thus they can be erected by stacking one element on top of another. The arrangements sketched by Leonardo da Vinci in the Codex Atlanticus belong to this family (see Houlsby 2014). However, the rule places severe constraints on which geometries can be obtained; generally speaking, the structures obtained have positive Gaussian curvature.

When considering practical applications, a friction-only joint should be avoided in favour of a bilateral joint to improve the robustness of the assembly. When a bilateral joint is used instead of a friction-only joint, the structures continue to exhibit all of the properties that characterize reciprocal system defined in “Introduction”. When a bilateral joint is used, the supporting element can either sit on the top or on the bottom of the supported element; the joint will exhibit prevalent compression when sitting on the top and prevalent tension when sitting on the bottom of the supporting element. The use of both top/bottom support position activates one of the most intriguing characteristics of reciprocal systems: the possibility to create free-form shapes with standardized elements.

Additionally, when a bilateral joint is used elements could be arranged without the superimposition joint, with their axis aligned. Historically, the proposals for a reciprocal roof by Sebastiano Serlio belong to this family (see Houlsby 2014).

We can thus describe three families of reciprocal systems based on their joint type:
  1. i.

    friction-only joint (top support position);

     
  2. ii.

    bilateral superimposition joint/(top or bottom support position);

     
  3. iii.

    bilateral joint with aligned elements axes.

     

Each of these typologies presents interesting and unique properties, in terms of both geometry and possible practical applications.

Reciprocal Structures Families Obtained with the Reciprocalizer

The architecture of the “Reciprocalizer” algorithm was set up for the highest generality, to provide the possibility to generate configurations in any of the three families of reciprocal structures just listed.

The Reciprocalizer is thus capable of generating configurations in all three conditions: where each supported element meets the supporting element in the top position, where each supporting element is met indifferently in the top or in the bottom position, and where elements join with their axes aligned.

The diagram in Fig. 6a represents a starting test configuration made of nine elements connected at their ends and converging in groups of three to four central nodes. Each element is defined with the specification of the nodal coordinates of its end points. An additional table must indicate the topology of connections between elements. The Reciprocalizer determines the geometry of the reciprocal joint by computing the three geometric parameters—eccentricity (e ij ), engagement length (l ij ), and top/bottom support position for each b i –b j connection—and adjusting the values according to the goal configuration, which may fall into one of the three families described in “Classification and morphology of families of reciprocal structures based on the joint type”.
Fig. 6

a Starting configuration, b solution 1: top support configuration, c solution 2: mixed top/bottom support configuration and d solution 3: aligned axes joint

In solution 1 (Fig. 6b), the support position is set for all connections b i –b j to be the top position. Therefore, each locally supported element meets the supporting element in the top position. The resulting configuration can be erected with friction-only joints, stacking one element on the top of each other.

In solution 2 (Fig. 6c), the algorithm ignores the specification of the top/bottom support position. Therefore, any top/bottom position is accepted, and the only condition set at each connection b i –b j is that the geometric compatibility is assured by the contact position. As a consequence, the algorithm computes the reciprocal configuration which is spatially the closest to the starting configuration geometry, and it will generate configurations with mixed top/bottom support positions. With those settings the algorithm exhibits the ability to adapt reciprocal assemblies to any free-form shape. A requirement for such configurations is that the joint must be bilateral if the supported elements connect the supporting element in the bottom position.

In solution 3 (Fig. 6d), elements join with their axes aligned. In this case, the value of eccentricity is set to 0, regardless the elements shape. In those configurations, the elements’ shape does not affect the eccentricity value, and the choice of the top/bottom position parameter does not apply. This solution has interesting practical applications for the possibility to create flat configurations, and for the independence of the elements’ shape from the overall geometry of the assembly.

Issues of Fabrication

Despite the apparent simplicity of the superimposition joint, when large, controlled networks of elements must be assembled, reciprocal structures require remarkable precision and uncommon expedients for their fabrication, because the overall geometry of the assemblage is the result of the complex and simultaneous interaction between all elements’ position and size. Therefore, the values of the geometric parameters at each connection are dependent on the values of all the others in the assembly. The practical consequence is that the elements must be assembled according to the exact parameters values, because if even a single connection sits outside a small tolerance, all the other elements’ positions are influenced and an overall modification of the geometry is determined. The extent of these modifications might require adjustment of all the other connections in the assembly in order to restore the geometric compatibility. Therefore, if one element is misplaced during assembly, it can cause a geometric incompatibility in the other elements of the assembly assembled according to the design parameters values.

This requirement for precision is challenged by the fact that each joint parameter is unique in free-form geometries, and in principle the elements’ spatial position requires six coordinates in order to be determined (either the position of the two end points with their coordinates x, y, z, or alternatively, the coordinates of one end point of the element and its rotation around the x-axis φ x , around the y-axis φ y , around the z-axis φ z ).

For the realization of the reciprocal mesh dome of the Rokko Observatory, whose architectural concept by Architect Hiroshi Sambuichi had been constructively and structurally developed by ARUP (Kidokoro and Goto 2011), the complexity of assembling a three-dimensional reciprocal structure was immediately evident to the engineers and the contractors. The complexity was tackled by creating the geometry of the dome from the repetition in a symmetrical circular array of an identical “master” slice. This strategy enabled the possibility to create a single “gig” or temporary scaffolding for the master slice that supports the steel elements in their final position before they are fixed through soldering.

The Scaffolding-Free Fabrication

We propose a method for fabrication of free-form reciprocal structures that does not require a temporary scaffolding to keep the elements in their final position before they can be fixed. The method is based on considerations regarding the kinematical determinacy of reciprocal systems, with a particular focus on networks composed of three-element fans.

If we consider a single three-element fan where elements are connected with internal hinges, it is a kinematically determinate assembly in both two and three-dimensional configurations (i.e., it exhibits only the free-body motions when unconstrained to the ground; Parigi et al. 2014). As a direct consequence, each fan, when complete, is rigid and retains its shape (Fig. 7a).
Fig. 7

a A three-element assembly is internally kinematically determinate, and exhibits only the three rigid body motions and b a four-element assembly is kinematically indeterminate, and it exhibits one inextensional mechanism beside the three rigid body motions

The geometry of a network of fans can be understood as being formed at the local level of the rigid fan units, which, when connected among themselves into a network, determine the global geometry.

Therefore, if the network is assembled with the subsequent addition of individual elements, starting from an arbitrary position, each fan retains, upon completion, the final shape and position with the adjacent fans. The geometry is assembled bottom-up, by enabling, at the local level of the fan, the complex interaction between the elements shape, topology and position that is characteristic of reciprocal systems.

Our ability to build the overall geometry relies uniquely on our ability to precisely assemble each element with the adjacent one. In case circular elements are used, each connection between elements b i b j , is uniquely determined by a single contact point for each of the two connecting elements, whose position should be identified along their surface, and that we call P ij on element b i and P ji on element b j .

Then we should be able to connect the elements and so that P ij and P ji are coincident. A statically determinate three-element fan is formed upon completion of the connections between elements b i b j , b j b k , b k b i , or, respectively, the coupling of points P ij P ji , P jk P kj , P ki P ik . When the three connections are completed, the fan is rigid, and each element retains its three-dimensional positioning with respect to the others in the assembly (Fig. 8).
Fig. 8

The three connections required to form a three-element fan

The opportunity to build the assembly bottom-up is provided by the fact that the three-bar fan is kinematically determinate. When we consider on the other side a single four-element fan where elements are connected with internal hinges, it is a kinematically indeterminate assembly in both two- and three-dimensional configurations (i.e., it exhibits, besides the free body motions, one additional inextensional mechanism) (Fig. 7b). The assembly is a mechanism. However the contact point between bars does not change during the mechanism, therefore a similar assembly procedure might be adopted here as well. Further studies might suggest that it is convenient to suppress, at least temporarily, the additional inextensional mechanisms for the purpose of fabrication. A network of four-bar assemblies is kinematically indeterminate, but only if the assembly is formed by a regular repetition of fans, while the network is kinematically determinate if the assembly is formed by a non-regular repetition of fans (Parigi et al. 2014).

The Prototype: From Digital Design to Manufacturing

The Reciprocalizer was included in a larger procedure that can be regarded as an extremely flexible and capable design tool for free-form reciprocal structures. This design tool has been applied for the design and realization of a free-form reciprocal structure composed of 506 round, un-notched wooden sticks of 22 mm of diameter. This paper describes the overall process from the design to the fabrication and focuses especially on the unique issues of fabrication of such assemblies. The structure was built by the students of the Master of Science programme in “Architectural Design” during a 1-week construction workshop that we organized at Aalborg University in the 2012 fall semester (Fig. 9).
Fig. 9

Night shot of the final prototype exhibited along the Limfjord, Aalborg waterfront

The design process includes:
  1. i.

    the definition of the starting free-form surface;

     
  2. ii.

    the definition of a network of elements connected at their ends with no axes eccentricity;

     
  3. iii.

    the definition of the corresponding reciprocal configuration with the Reciprocalizer algorithm;

     
  4. iv.

    the calculation and output of the fabrication data;

     
  5. v.

    the fabrication.

     

The Starting Free-Form Surface

The design begins with the definition of a starting geometry. In this case a doubly-curved, free-form surface with no axis of symmetry (Fig. 10) was chosen in order to test both the capability of the reciprocal system to fit to highly irregular surfaces with both positive and negative Gaussian curvature (Fig. 11), and the ability of the Reciprocalizer to transform each joint into a correspondent reciprocal joint.
Fig. 10

The starting free-form geometry. a Elevation aa, b elevation bb, c plan and d perspective

Fig. 11

Curvature analysis on the starting free-form geometry. a Elevation aa, b elevation bb, c plan and d perspective

The Network of Elements

The next step is the definition of a network of elements connected at their ends with no axis eccentricity, drawn upon the previously defined surface. Elements at this stage have no thickness, and only their axes are represented. In this case, a Voronoi diagram was chosen to cover the free-form surface (Fig. 12). The resulting mesh is constituted from 506 elements distributed over the surface.
Fig. 12

Voronoi diagram on the starting free-form geometry. a Elevation aa, b elevation bb, c plan and d perspective

The Reciprocalizer

Starting from the Voronoi diagram, and after assigning the elements’ thickness, the corresponding reciprocal configuration is obtained. Each node of the Voronoi diagram where three elements converge is turned into a corresponding three-element fan (Fig. 13). The Reciprocalizer is set to ignore the specification on the top/bottom support position, and to only compute the geometric compatibility of elements by physical contact. As a consequence, the algorithm generates a configuration with mixed top/bottom support position and variable engagement lengths so that the overall geometry closely fits the starting free-form geometry.
Fig. 13

The reciprocal configuration. a (Above left) elevation aa, b (above right) elevation bb, c (below left) plan and d (below right) perspective

The Fabrication Data

A table of values outputs the data needed for the fabrication (i.e., for the element b i , the position of the contact point P ij with each of the b j connecting elements). Point P ij is located along element b i surface and its position can be described with two values: its distance from one reference end D ij , and the angle α ij that it creates with a reference line arbitrarily set on the side element, measured from the element axes and in a perpendicular plane (Fig. 14). Except for the elements on the boundaries, in general each element connects with four adjacent elements, and therefore four contact points must be described (Fig. 15).
Fig. 14

The distance D ij and the angle α ij showed on a sample element

Fig. 15

A 4 × 4 sample of the fabrication table. The complete table has dimension 2,024 × 4

In this case the table presents 2,024 lines, each one describing the position of one contact point. In principle, this table is the only document needed for the entire process of fabrication. However, the digital model of the structure was also supplied as a tool to visually find and double-check the position of the elements in the space.

The Fabrication Process

The fabrication process starts with cutting the elements to their length, labelling them with their identification ID and marking the position of the contact points P ij on each element (Fig. 16).
Fig. 16

The marking process

He marking process is relatively fast and low tech (a pencil and a paper ruler). First, a reference line is drawn on the side of the element and along it a distance D ij is marked. Then, starting from this distance, the angle α ij identifying the final position of D ij , is marked. In order to mark both D ij and α ij with the same measuring tool, the value of the angle α ij is converted into a distance A ij along the element circumference (Fig. 17) with the formula:
Fig. 17

Conversion of α ij into linear distance A ij

$$ A_{ij} = \alpha_{ij} r_{ij} \frac{\pi }{180}. $$
A paper ruler was chosen for its adaptability to measure distances along both straight paths, the case when measuring the distance D ij , and curved paths, the case of measuring the distance A ij along the element circumference (Fig. 18).
Fig. 18

The labelled and marked elements

For the coupling of the elements, two solutions of the joint were developed and tested, each requiring a different drill hole on the markings of the points P ij .

In joint 1, the depth and width of the drill hole are equal to the radius of a shear lock (a sphere in this case), which is then placed between the two connecting elements (Fig. 19), preventing elements from sliding off their respective positions. After the sphere is placed, the elements need to be fixed in position, and the knot in Fig. 21 is used.
Fig. 19

Joint 1

In joint 2, the drill hole goes through the elements, and a rope is threaded through both elements (Fig. 20), and an initial knot is used to keep the elements in place, before the final fixing knot of Fig. 21 is used. Joint 2 was used for the final realization of the prototype.
Fig. 20

Joint 2

Fig. 21

The knot fixing the joint

The sequence of construction starts from the centre and gradually extends to the sides. This assembling sequence allows for immediate detection of, if any, errors in the construction, because the geometric compatibility is checked each time a ring of the dome is complete (Fig. 22). The joint adopted allowed a two-step fixing: initially elements are presented and loosely fixed with a knot passing through the elements, and then the final fixing knot is performed in a second stage. The assembling process was smooth and done with the simultaneous collaboration of groups of around thirty people. The final prototype closely matches the digital model shape (Figs. 2, 23).
Fig. 22

a Plan of the construction sequence in steps, b step i, e step n. be Construction sequence

Fig. 23

Picture of the final prototype and side-by-side comparison with the digital design model. Photo: Mathias Sønderskov

Conclusions

Our experience has proved that the structural typology of reciprocally connected elements with superimposition joints, if supported by a custom developed design and fabrication process, is well-suited to build low-cost, free-form shapes with standardized elements. The design and fabrication of the prototype had two main objectives: (1) to test the results of the Reciprocalizer algorithm, and (2) to test a fast, efficient construction method, scaffolding-free, suited for any irregular free-form geometries. The experiment was successful in confirming and validating the output obtained from the Reciprocalizer algorithm and the ability to match the shape of the digital model into the prototype by connecting the elements according to the precise position indicated in the fabrication table. It was also successful in demonstrating the possibility to build the structure without scaffolding, by starting from a central point and gradually extending to the sides, taking advantage of the kinematical determinacy of the three-element reciprocal unit.

The construction of the dome also proved to be an engaging and almost magical experience when the expected shape emerged from the complex interaction of hundreds of precisely assembled elements with a low-tech fabrication technique.

Notes

Acknowledgments

The authors would like to thank the Department of Architecture, Design and Media Technology and the Department of Civil Engineering, Aalborg University, for providing working spaces and funds for the construction workshop, the students of the first semester of the Master of Science program in “Architectural Design” at Aalborg University for their intense participation and Eng. Ryota Kidokoro, invited guest of the workshop. All photos are by Dario Parigi except Fig. 23 by Mathias Sønderskov.

References

  1. Baverel, O., and O. Popovic Larsen. 2011. A review of woven structures with focus on reciprocal systems—nexorades. International Journal of Space Structures 26: 4.Google Scholar
  2. Houlsby, G.T. 2014. John Wallis and the numerical analysis of structures. Nexus Network Journal 16: 1. (in this same issue).CrossRefGoogle Scholar
  3. Kidokoro, R., and K. Goto. 2011. Rokko observatory—application of geometric engineering. In Proceedings of the international symposium on algorithmic design for architecture and urban design, 14–16 March 2011. Tokyo: ALGODE.Google Scholar
  4. Parigi, D., and P.H. Kirkegaard. 2013. The reciprocalizer: a design tool for reciprocal structures. In Proceedings of CC2013, 14th international conference on civil, structural and environmental engineering computing, 3–6 September 2013. Italy: Cagliari.Google Scholar
  5. Parigi, D., and P.H. Kirkegaard. 2014. The reciprocalizer: an agile design tool for reciprocal structures. Nexus Network Journal 16: 1. (in this same issue).CrossRefGoogle Scholar
  6. Parigi, D., Kirkegaard, P.H., and M. Sassone. 2012. Hybrid optimization in the design of reciprocal structures. In Proceedings of the IASS symposium 2012: from spatial structures to space structures, 21–24 May 2012, Seoul.Google Scholar
  7. Parigi, D., M. Sassone, P.H. Kirkegaard, and P. Napoli. 2014. General kinematic formulation of planar reciprocal systems. Nexus Network Journal 16: 1. (in this same issue).CrossRefGoogle Scholar
  8. Parigi, D., and A. Pugnale. 2014. Three-dimensionality in structural reciprocity: concepts and generative rules. Nexus Network Journal 16: 1. (in this same issue).CrossRefGoogle Scholar
  9. Pugnale, A., Parigi, D., Sassone, M., and P.H. Kirkegaard. 2011. The principle of structural reciprocity: history, properties and design issues. In Proceedings of the IABSE–IASS, symposium 2011: taller, longer, lighter, 20–23 September 2011, London.Google Scholar

Copyright information

© Kim Williams Books, Turin 2014

Authors and Affiliations

  1. 1.Department of Civil EngineeringAalborg UniversityAalborgDenmark
  2. 2.Department of Civil EngineeringAalborg UniversityAalborgDenmark
  3. 3.Department of Civil EngineeringAarhus UniversityAarhusDenmark

Personalised recommendations