ARQ, n. 63 Mecánica electrónica / Mechanics & electronics, Santiago, August, 2006, p. 30 - 35 .



On Shells and Blobs
Structural Surfaces in the Digital Age(1)


Martin Bechthold*

* Professor Harvard University Graduate School of Design, Cambridge, U.S.A.


New representation techniques have expanded the possibilities of architectural form, such as it is understood as an expansion for creative freedom. How can we relate this new potential with the engagement that gravity force (as weight, thrust and resistance) demands on built work?.

Key words: Shells, structural design, structural surfaces, CAD/CAM, paraboloid.

We all know that computer-aided design and manufacturing (CAD/CAM) technology has triggered a proliferation of complexly shaped building designs, including the free forms we call blobs. But we have forgotten that many of these fluid shapes resemble concrete roof shells that appeared in the first half of the 20th century –shells devised to cover long spans with a minimum of material. There is, of course, a difference, and it hinges on structure.
It is often assumed that all curved surfaces are necessarily stiff and thus useful as primary structural elements. Surface curvature, however, forms a structurally effective shell only if it enables efficient membrane stresses to develop, which in turn allow thin surfaces to carry significant loads. This structural elegance contrasts with the relative clumsiness of the support systems for digitally generated irregular shapes.
It is true that shells cannot offer the degree of formal freedom for building skins and roofs that blobs offer, but why, given their compensating efficient use of material, are they practically forgotten amidst the frenzy about blobs? Architecture schools generally offer digital design classes, but hardly any schools address the design of structural surfaces(2). This disinterest in shells dates back to the early 1980s and reflects an avoidance of the immense constructional challenges that these systems pose. But can CAD/CAM technology make the construction of material-efficient structural surfaces easier? Can digital technology not only assist in expanding formal design, but also enable the use of material-efficient structural systems?
With current discourse focusing on the underlying shaping principles of digital architecture(3), architects often overlook the fact that the formal complexity of the blob is achieved through conventional constructional and structural means –skeletons assembled from linear and curvilinear members that support secondary members and non-structural building skins. These systems rely heavily on bending stresses –the least efficient of the basic load-carrying methods. The visible surface of the free-form shape is structurally functionless (see image 1). The curvature present in free-form shapes rarely allows membrane stresses to develop, since the underlying shaping algorithms are optimized for visualization purposes and cannot accurately simulate and model structural behavior. These shapes leave little choice other than those structural systems that rely on inefficient bending action. The in-plane membrane stresses of the shell are vastly more efficient: less material is needed to carry comparable loads.
Structurally efficient shells use construction technology that has progressed only marginally since the 1980s. This stagnation in technology contrasts starkly with the significant advances in digital design and manufacturing systems for skeletal structures. Steel Detailing Software and the related computer numerical control (CNC) manufacturing facilities partially automate the design and fabrication of structural members of skeletal frames, including the complex support systems of free-form shapes. Now new computer-driven manufacturing technology is needed for shell construction if this material-efficient system is to become a viable alternative.

The development in the 1920s of structural roof shells stemmed from a fascination with a new material –reinforced concrete– and the need to cover medium to large spans economically. Félix Candela in Mexico, Eduardo Torroja in Spain, Eladio Dieste in Uruguay, Franz Dischinger and Ulrich Finsterwalde in Germany, Heinz Isler in Switzerland, and Anton Tedesco in the United States were among the pioneer shell builders. The labor-intensive construction of the complex shape was economically justified through significant savings in materials: Candela’s 1958 shell for a restaurant in Xochimilco, Mexico (see image 2), spans 106 feet with concrete just 1.7 inches thick.
Concrete shells include single-curved shapes such as cylinders and cones, as well as a variety of double-curved geometries. Double-curvature synclastic (with curves running in same direction) and anticlastic (with curves running in opposite directions) shapes are structurally particularly efficient, but the construction of their formwork is technically more demanding. Hyperbolic paraboloids (HP) and hyperboloids (see images 3, 4) form a particular group within anticlastic shells. They combine an efficient load-bearing mechanism with relative ease of construction: the formwork for these surfaces can be made mostly from straight wooden boards. The majority of Candela’s concrete shells –as well as some timber shells by other designers– are based on HP shapes.
Early shell builders employed simple geometries that could easily be geometrically described and built(4). Heinz Isler developed a new concept for shells in the 1950s by deriving shell geometry from experiments with accurate physical models such as inflatable rubber membranes or hanging fabric. These shells are equilibrium shapes –their shapes balance loads such as the weight of the shell through membrane stresses. Also in the ’50s, researchers at Frei Otto’s Institute of Lightweight Construction at Stuttgart University experimented with form-finding methods for tensile systems by studying minimal surfaces using soap bubbles and other methods. Their physical models were later complemented and partially replaced by computational form-finding methods, applicable to both tensile systems and shells. Designing an equilibrium shell means defining a structurally efficient shape through the specification of its support conditions and loads. Each prescribed combination of support and load will yield a unique geometry.
The advances in computational form-finding of the 1970s came at a time when the interest in shells was rapidly fading. Fabric structures, cable nets, and space frames (triangulated bar networks) presented equally efficient structural solutions for spanning larger distances, but their constructional problems were more readily solved with the established building technology for skeletal structures. The few shells built after the 1970s were mostly grid shells, with the continuous surface replaced by linear or curvilinear interconnected members.
What are the current technical impediments to shell construction? The making of a complexly shaped surface is necessarily challenging, even more so if this surface becomes the primary structural element. Common shaping or forming techniques may be applicable to thin materials such as sheet metal or wooden boards, but these elements by themselves are insufficient for roof shells. Shell construction techniques have traditionally relied heavily on labor, and consequently are hampered by today’s high labor costs. Labor costs (not adjusted for inflation) between 1958 and 2002 increased between factors of eight (unskilled labor) (Williamson, 2003) and eleven (manufacturing labor)(5), whereas the cost of construction materials (not adjusted for inflation) increased during the same time period only by factors between 3,8 (steel milled products) and 4,8 (ready-mixed concrete)(6). Less labor-intensive construction techniques need to be devised if shell construction is to become feasible today. Formwork accounts for a major part of the cost of concrete shells, and only by reusing these expensive, large molds can shells be economical. Heinz Isler, one of the few remaining active shell builders, reuses formwork in different projects, accepting the severe design restriction to regular plan geometries, the repetition of identical shapes (see image 5), and the inability to adjust to local conditions.
Research at the Harvard Design School has been suggesting new processes for shell construction using CAD-CAM technology. The accuracy of CNC fabrication now enables the subdivision of larger shells into panels that can be prefabricated, transported to sites, and assembled into larger roof shells. Researchers are developing and prototyping three types: a laminated timber-sandwich shell, a ferro-cement folded plate system (see image 6) with thin cement slabs reinforced by steel mesh, and a concrete shell system constructed with prefabricated lost formwork that becomes embedded in the structure.
The timber shell is a rigid sandwich with a high-density foam core and laminated wood facings. The foam core of individual panels can be fabricated on a CNC milling machine. Multiple layers of precut strips of sliced thick veneer or plywood are then laminated over the surface and cured under vacuum pressure (see image 7). Specially engineered lap or finger joints connect individual panels and form the shell. The sandwich can be designed to satisfy current thermal insulation requirements, fire-resistance, and structural rigidity. The depth of the overall system is sufficient to embed service elements, thus further enhancing multifunctionality in the geometrically complex roof. The sandwich, in combination with certain core and facing materials, can generate enough stiffness to accommodate moderate bending stresses. Not only shells, but also free-form shapes can be manufactured according to this method, as long as the bending stiffness of the sandwich suffices. At last The Shell Catches Up With the Blob: a technology developed for structural surfaces can be a viable alternative to the skeletal construction of free-form shapes (Bechthold, 2001 and Schodek et al., 2004).
The ferro-cement folded plate system revisits both a forgotten system and an exciting material no longer used in industrialized countries. Folded plates are closely related to shells, because their load-bearing mechanism is principally derived from the in-plane stresses of a thin folded surface(7). The constructional problems of folded plates are the relative complexity of formwork in concrete systems and the difficulty of creating efficient connections between flat panels in timber systems. An obvious solution is to use a material that combines stiffness and a capability for folding without relying on elaborate on-site formwork: thin sheets of a mesh-reinforced composite –ferro-cement(8). Ferro-cement panels can be manufactured efficiently on flat, reusable formwork. Predetermined fold lines are not covered with mortar during this process –these lines effectively act as hinges in the folding process. Here the steel reinforcing mesh yields on bending, and flat concrete plates can be folded like origami into a three-dimensional structure. The open joints are covered with mortar after the folding. Based on a 1980 Australian patent application, this concept has not yet been pursued beyond small-scale experiments carried out at the University of Sydney(9).
CAD/CAM technology streamlines the design-to-production process. Complex folded systems are generated digitally, and integrated structural analysis tools provide rapid feedback on the feasibility of schemes. The components of the design models are then digitally flattened into 2-D production patterns that are produced on CNC-lasers or routers. The reinforced panels are fabricated through either manual or partially automated spraying of the mortar over the reinforcing mesh. The folds have to be carefully engineered, and excessive bending of the plates needs to be avoided during folding(10). A wide range of custom shapes can be produced by this method: the number of individually shaped panels in a project has little impact on manufacturing time and costs. Parametric variations of a folded plate system –inherently simple to generate using design development software such as SolidWorks or Catia– can be manufactured and assembled more efficiently than they could be using traditional on-site construction techniques.
A third research investigation is developing a system of shaped ferro-cement panels to be used on site as lost formwork for a load-bearing layer of cast-in-place concrete and reinforcement. This approach –closely related to Pier Luigi Nervi’s use of lost formwork in the ’50s and ’60s– specifically addresses the need for reduced formwork cost by making the formwork a structural part of the finished shell. The ferro-cement elements are manufactured accurately off-site using CNC machines and spraying of mortar that may be partially automated. Overlapping reinforcement bars that become embedded in the cast-in-place concrete secure structural connections between adjacent panels in the finished shell(11).
These three processes for the design and production of individualized and material-efficient structural surfaces use CAD/CAM technology to allow architects greater variety of shapes while reducing the additional cost normally associated with customized construction. The problem of having to employ identical elements repetitively is, however, not limited to shells; it is a part of much building construction. The use of standard products and components is a generally accepted design imperative. Everyday construction products –from plywood and standard steel shapes to suspended ceiling systems and light fixtures– come in few variations. The building industry operates largely under economic principles that originate in the laws of early industrial mass production, and the customization of products usually implies significant cost increases. Will CAD/CAM eliminate the need for repetition altogether, thus drastically affecting the way we design?
The possibility of thoroughly customized buildings with individualized components is attractive to architects. Mass customization has enabled the production of individualized products in other industries at a price similar to that of equivalent mass-produced items. Mass customized products range from customized books to individualized machines for industrial production. The building industry has been adapting to this trend: individualized windows are now designed online, and the components are usually manufactured automatically. Walls are constructed using sets of prefabricated building blocks that are designed, CNC cut, and delivered to sites just-in time, increasing construction speed and reducing waste. This incremental implementation of mass-customization in the building industry has hardly been noticed. It is likely that we will continue to use standard products and materials in buildings, but interesting opportunities for designers, engineers, and contractors may arise where they are least expected. A recent example is the manufacture of custom steel reinforcement mats for concrete slabs –digitally designed and robotically welded– that can save a substantial percentage of steel because each mat is precisely tailored to the stresses and deflections present in the slab.
What objective can, or rather, should customization serve? The study of shells demonstrates that it may enable structurally efficient construction systems, provide a rich spatial experience, and use material resources responsibly. Customization through CAD/CAM could and should be directed towards a more efficient response to performance requirements as diverse as program, structure, energy efficiency, lighting, and maintenance. Digital technology is not an end in itself but should play a role in creating a more human, socially responsible, and sustainable environment. Before long, today’s separate discourses on sustainability and digital design will, we can assume, productively connect.


1. This article was first published in the Harvard Design Magazine No. 19.
2. In April 2003, I conducted a survey of courses offered at nineteen major American architectural schools. Occasionally shells were briefly mentioned in survey structures courses. My course at Harvard Design School was the only one to cover shells in any depth. The schools included in the survey are Yale, Princeton, Cornell, the University of California at Berkeley, MIT, UCLA, the University of Colorado at Bolder, the University of Florida, Georgia Institute of Technology, IIT, Tulane, the University of Michigan, Columbia, Rhode Island School of Design, the State University of New York at Buffalo, Rensselaer, and the University of Texas at Austin.
3. See for example Journal of Architectural Education, November 2002, and many recent books and publications.
4. German engineers pursued a complete quantitative understanding of simple cylindrical shells prior to building prototypes. Candela’s more daring HP shapes were often built without complete theoretical understanding of their structural behavior –Candela has referred to his workers as the ones that solved many technical problems directly on site (Herzog y Moro, 1992).
5. U.S. Department of Labor, Bureau of Labor Statistics.
6. Ibid.
7. A recent folded-plate structure, albeit a hybrid through a combination with trusses, is the roof of the Yokohama Ferry Terminal by Foreign Office Architects.
8. J. Lambot in France invented ferro-cement in 1848, a year before Monier’s use of reinforced concrete. Ferro-cement features excellent crack control due to the large bonding surface between the mesh layers and the matrix. It is extremely versatile and has been successfully used for an extremely wide range of applications ranging from seagoing vessels and water tanks to prefabricated housing panels and medium-span roof systems.
9. In 1980 two lecturers in Civil Engineering at the University of Sydney, Wheen and Jackson, applied for a patent on the bending of ferro-cement slabs. R.J. Wheen and G.N. Jackson, “Method of Bending Hardened and Stiff Slabs”, Australian Patent Application No. PE 3167, April 1980.
10. My team and I used a comparatively simple folded plate system to develop a hinge detailing strategy and the folding apparatus. The process was then tested on a six-foot long prototype.
11. These ongoing research projects are conducted at and funded by the Harvard Design School. Team members include Jerome Chang, Jason Halaby, Chung-Ping Lee, Mark Oldham, Tyrone Yang and me.

Bechthold, Martin. Complex shapes in wood: Computer aided design and manufacturing techniques. Doctoral thesis, Harvard University, Cambridge, 2001.
Herzog, Thomas and José Luis Moro. “Zum Werk von Félix Candela”. ARCUS 18, 1992, pp. 10-22.
Schodek, Daniel; Bechthold, Martin; Griggs, James Kimo; Kao, Kenneth and Marco Steinberg. Digital design and manufacturing: CAD / CAM Applications in architecture and design, John Wiley & Sons, Hoboken, 2004.
Williamson, Samuel H. “The relative cost of unskilled labor in the United States, 1774-present”. Economic History Services, March 2003.