Learning from Failure: Hybrid Fabrication of a Gridshell Canopy Structure Using Timber Battens and 3D Printers

Building construction is a crucial part of the architectural profession. As a process, it encompasses all aspects related to the planning, fabrication, production, management, assembly, and disassembly of buildings. Consequently, it is closely tied to the state of the construction industry, which has undergone significant changes in recent years. In response to the demands of 21st-century global society, construction professions have had to become more multidisciplinary, particularly in areas related to technology and management (Chan Edwin H. W. et al., 2002).

At universities, technical courses are often taught through lecture-based sessions. However, studies have shown that these types of sessions tend to result in low student engagement (Shareef & Farivarsadri, 2020). Therefore, to properly prepare students for professional practice, universities must move away from traditional curricula and instead adopt teaching and learning materials that embrace a more proactive and multidisciplinary approach to building construction courses. In this context, preparing students for multi-disciplinary practice means that, as professionals, they will not be confined by the traditional roles of architects, builders, engineers, or surveyors (Chan Edwin H. W. et al., 2002).

During their education, architecture students often lack practical assignments that involve building at 1:1 scale. As a result, they rarely experience the full design cycle, which begins with conceptual elaborations and ends up with a real scale physical object.

This gap is particularly problematic in building construction courses, as these courses should be designed to provide students with both the applied knowledge of building construction and the management skills necessary for their professional activities.

Working with real-scale prototypes promotes an active learning environment, where students learn by making. By physically engaging with their subject matter, students can fully develop their creative thinking and problem-solving abilities. The hands-on experience they gain enhances their design capabilities, increases their awareness of their own learning progress, and prepares them for professional practice (Morales-Beltran, 2023). The ability to understand the entire design process – from conception to delivery – is a particularly needed in a construction industry where professionals are expected to oversee and manage project development and the construction process.

The widespread integration of digital technologies in the construction industry is reshaping business models and driving firms to focus on resilience and adaptability. This “digital shift” necessitates a corresponding “cultural shift,” emphasizing the development of soft skills such as collaboration, flexibility, and adaptability. To succeed in a digitally-driven environment, students must learn to work across traditional silos, embrace risk-taking, and avoid an overreliance on commercial software (Papadonikolaki et al., 2020).

Moreover, the increasing role of digital tools in architectural practice highlights the importance of students acquiring both basic computational skills and an understanding of new technologies to accelerate design, fabrication, and assembly processes. The rapid development of digital tools in architecture has raised the need to integrate these technologies into architectural curricula more effectively. Graduating architects must be equipped with the ability to use digital modes of design and fabrication, which integrate traditional skills of drawing and making into computational contexts. This approach not only enhances design thinking but fosters new conceptualizations of space and geometry (Soliman et al., 2019; Tepavčević, 2017)

Hybrid materiality refers to the integration of both digital fabrication techniques and traditional craft practices, allowing for a blend of modern technology with age-old methods. Hybrid design expands on this idea, merging the forms and qualities of traditional craftsmanship with digital fabrication to explore new creative possibilities (Bernabei & Power, 2018). This fusion enables greater flexibility in architectural design, fostering innovation in both form and function. While digital fabrication, especially 3D printing, presents exciting opportunities for customization and efficiency, it is inherently limited by its reproducibility, unlike handmade artefacts which offer uniqueness and individuality. A key example of this is seen in the practice of combining digital fabrication with object restoration, where 3D printing is used to repair or recreate crafted items (Zoran & Buechley, 2013).

In architectural education, the hybrid use of both digital and analogue fabrication techniques plays a significant role in enriching design learning. By engaging students in both methods, they gain deeper insights into the advantages and limitations of each technique. These active learning experiences help students explore the relationships between space, geometry, and material performance, and encourage creative thinking. This process, enhanced by computational tools like parametric design software, not only accelerates the design process but also enables exploration of non-conventional geometries and structural forms, contributing to innovation in both pedagogy and architectural production methods (Morales-Beltran et al., 2021).

Despite the potential for innovation, there are challenges associated with hybrid fabrication techniques. While digital tools such as 3D printing are increasingly accessible, they are still not fully suitable for large-scale construction due to high costs and limitations in material applications. However, ongoing research is working to make 3D printing and other digital fabrication methods more feasible for broader use in construction (Lee et al., 2024). Combining additive and subtractive manufacturing techniques can lead to efficient production, where large, simple components are made using traditional methods, and more complex, customized elements are produced through digital fabrication. This hybrid approach offers exciting possibilities for both pedagogy in architectural education and practical applications in construction, as seen in projects like 3D-printed nodes for gridshell structures (Seifi et al., 2018) or bio-composite connections for bamboo space structures (van Wassenhove et al., 2021).

While the above components are essential to a comprehensive architectural curriculum, several challenges hinder their inclusion. The growing demand to reduce the number of courses, shorten course contents, and the shortage of specialized instructors leaves little room for integrating construction exercises that address the full design cycle within a single-semester course. Additionally, while the use of digital technologies (e.g., drawing software or 3D-printed modeling) is highly desirable, a significant portion of any course incorporating such technologies must be dedicated to teaching students how to use 3D modeling software. This further limits the time available for the design, fabrication, and assembly of physical prototypes.

To add to this, while studio classes are often compressed to just two days per week, regular courses (i.e. not studios) are confined to a few hours per week. This leaves students with even less time to engage deeply with the full scope of the course content.

This paper presents the settings and outcomes of an undergraduate elective course on structures, in which students designed and fabricated a full-scale prototype over 14 weeks. The primary objectives of the project were to provide students with hands-on experience in designing and constructing a gridshell prototype using a hybrid approach, combining timber battens and 3D-printed nodes, and to enhance students’ abilities to create free-form structures using standard building materials and common 3D printers. However, a series of both expected and unexpected events not only extended the duration of the whole exercise but also led to failure-driven design and learning processes.

To achieve the goal of designing and building a full-scale prototype, the course was organized as an extended design-build workshop. Inspired by the ProtoLAB workshops (Latka, 2023), the course centred around a specific topic, with students receiving short, application-based, and result-oriented inputs. It was divided into two main phases: design and build. The requirement that “it has to be built” informed all decisions throughout the process, driving constant discussions on feasibility (e.g., Can we build it within n weeks? Do we have all the materials? Can 11 people complete this?), budget (e.g., Can we afford X material? Who will pay for it?), and associated risks (e.g., Is the location safe? What happens if it collapses?).

The course was planned to enable 11 students to design and build a full-scale structure using hybrid production methods within a 14-week semester. The key components and materials used in the course are described below.

The course was organized into four major stages, each lasting three weeks (see Table 1):

Understanding – including lectures & workshops focusing on form-finding methods,

After the Designing stage, which coincided with the official midterm exam week, students were required to submit designs for all the nodes and have them ready for 3D printing.

Since students do not have enough time to perform any sort of structural analyses, a compulsory step in the process of prototyping any student’s design is the introduction of a structural logic as early as possible in the design process. To this end, the course offers the students two key inputs:

Physical form-finding workshop – for the students to grasp the physical principles that rule funicular structures.

2.3.1.

Workshop with fabric and melted candle wax

This simple workshop introduces students to funicular systems, whose forms are linked to the manner loads are transferred to the supports through simple surface stresses. Students use fabric, dipped in melted candle wax, to quickly realize how the material's type and shape influence the resulting form. The process involves (Figure 1):

Dipping 30 × 30 cm (approx.) pieces of fabric into melted wax.

Hanging the waxed fabric upside down on a cardboard base using clothespins.

Allowing the fabric to dry using temporary timber supports.

Once dried, flipping the fabric to form a compression membrane structure.

Figure 1.

Physical form-finding workshop: funicular membranes using fabric and candle wax

The internal competition is designed to choose a single design to be developed as a team. This design must have already incorporated a structural logic, increasing its “buildability”. Students are required to prepare proposals using the provided parametric GH file, which models a generic gridshell structure based on catenary curves.

To ensure the feasibility of the designs, students are limited to 110 nodes and a maximum timber batten length of 100 cm. These limitations are based on insights gained from previous similar projects (Morales-Beltran et al., 2021).

The materials were largely available through leftovers from a previous research project, so students only needed to purchase a few items (such as bolts and cables). The materials included:

Software: Rhinoceros 3D – henceforth referred to as Rhino – and GH for modelling (McNeel, 2022), Prusa Slicer and Ultimaker Cura for preparing G-codes files

The students self-organized and proposed seven designs (Figure 2). The selection committee was composed of the students themselves, who presented their proposals to one another for discussion. The designs were evaluated based on factors such as size, apparent complexity, and other considerations related to time constraints, available materials, and fabrication techniques. Three of the proposals were relatively large, which presented logistical and implementation challenges. From the four designs that appeared more feasible in terms of scale, students – in an open vote – chose the cantilevered canopy proposal. The fact that most of the other proposals strongly resembled variations of the given gridshell design, may have been perceived as too similar to attract much interest. Interestingly, the students behind the selected canopy proposal moved away from this obvious solution. Instead, they developed a tree-like form using a pattern of Voronoi cells.

Figure 2.

The seven proposed designs (depicted at similar scale): three large gridshell structures on the left, smaller proposals – including the selected one – on the right

The first modification to the design was driven by the need to ensure the stability of the structure. Given the temporary nature of the prototype, anchoring was not a viable option to provide the cantilever structure with the required self-balance. As part of a collective class effort, several key design strategies were proposed: enlarging the base (the area in contact with the ground), increasing the canopy’s angle of inclination, and adding tension cables. These changes helped improve the stability of the prototype by lowering its centre of gravity (Figure 3a).

Figure 3.

(a) Pre- and post-revised versions of the prototype, and (b) sketch used during the discussion on the design of the canopy for structural enhancement

Next, the arrangement of battens in the cantilever section, henceforth referred to as the ‘canopy,’ needed to be revised to ensure that the length of the timber battens did not exceed one meter. Reducing the length of the battens required splitting them and adding new nodes to the structure. As a result, the redesign of the canopy became a trade-off between the number of battens and the number of nodes (Figure 3b). An additional constraint in this process was that no more than three battens should meet at a single node, with the exception of the nodes where the canopy connects to the supporting structure. Fortunately, this condition was met by adhering to the Voronoi pattern.

In line with the goal of bringing the centre of gravity closer to the axis of the vertical part of the structure, henceforth referred to as the ‘column,’ the cross sections of the timber battens vary depending on their position within the structure (Figure 4). Along the longitudinal axis, battens of varying section sizes are arranged to ensure a balanced distribution of the canopy’s mass. The use of smaller sections in the canopy (Table 2) reduced its overall weight, while positioning the larger sections in the column helped lower the centre of gravity, thereby increasing the overall stability of the structure.

Figure 4.

Distribution of the 168 timber battens across the structure

The final design included 168 timber battens and 103 nodes.

After the canopy was redesigned, a new digital model was created in Rhino, incorporating all the battens at their correct sizes. Next, the 103 nodes were distributed among the 11 students, with each of them assigned 9 to 10 nodes to design for 3D printing. Although there were no fixed rules for the design, students were instructed to minimize the size of the nodes in order to reduce material usage and printing time. In addition, the nodes had to include several compulsory components (Figure 5): boxes to hold the timber battens and holes for fixing the battens into the boxes. Some nodes also required additional holes for inserting tension cables.

Figure 5.

Standard design components of a node

While all nodes were designed to optimize strength and facilitate assembly with the wooden pieces, their characteristics varied depending on whether they connected battens with different cross sections. The variation in cross sections led to nodes with boxes of different sizes and wall thicknesses (Table 3). Since the battens are inserted into the nodes, not all nodes could be assembled by simply sliding the battens into the boxes. To address this, some nodes were designed to be split into two parts, allowing them to be connected after the battens were attached.

Despite the design process initially appearing straightforward during the planning stage, it became an iterative process in practice. This was because the initial designs featured mistakes, such as unnecessary use of material (e.g., excessive volume or the inclusion of additional elements), inconsistencies with assembly requirements (e.g., cable holes interfering with boxes or wrong size of the box), and issues with the required strength (e.g. holes placed at points where the structural integrity of the node, or part of it, was compromised) (Figure 6). As a result, students were required to make multiple design revisions based on feedback from the course instructor. After these iterations, the node designs became quite varied, with some even contrasting significantly (Figure 7).

Figure 6.

Most common mistakes during the early design of the nodes: (a) excessive use of material, (b) open edges in the models, (c) cable holes interfering with timber battens, and (d) wrong positioning of the bolt holes

Figure 7.

Contrasting sizes and complexities of the nodes. Node 58 connects battens of the smallest cross-section (3.5 × 15 cm), while node 30 connects battens of the largest cross-section (4.4 × 2.2 cm), as well as the “column” and “canopy” sections. It also includes holes to anchor the tension cables. Since node 30 is split for assembly, the letters U and B are added after the node number to indicate the “upper” and “bottom” positions

The assembly process, involving all students and the instructor, was carried out in two stages: column assembly and full prototype assembly (Figure 9). The column was assembled first, approximately three weeks before the canopy was completed. This stage took about six hours, with most of the time spent ensuring the battens fit correctly into the nodes.

Figure 9.

Assembly process of the first prototype in an open area of the campus

Unlike the column, which was assembled bottom-up, the canopy was assembled upside down. This approach was chosen because, by design, all perimeter battens were placed on the same virtual plane using the ground as a reference. As long as the perimeter battens were correctly connected, this strategy minimized the risk of assembly errors. The entire canopy assembly process took about 18 hours, with most of the time again dedicated to ensuring the battens fit properly into the nodes.

Once the canopy was assembled, it was flipped and placed on temporary supports. The team used benches, tables, ladders, and long sticks to lift the canopy until it could be positioned beneath the column. The final step involved connecting the canopy to the column by securing the appropriate nodes and tensioning cables to adjust the canopy into the correct shape. This took an additional three hours.

During the examination of the column, it became clear that most, if not all, of the broken nodes were the white ones (Figure 11a, b). These nodes were printed using the defective 3D printer – before the extent of the problem was fully realized. Worse, they were relatively small and/or had smaller boxes, which reduced their contact surface area with the battens. As a result, when subjected to strong wind-induced vibrations, the column’s ultimate strength capacity relied on the bolts, which, unsurprisingly, surpassed the shear capacity of the PLA+ wall layers, leading to local failure. Given the distribution of the white nodes in the column, local failure of these nodes ultimately led to the collapse of the entire structure.

Figure 11.

View of the (a) original column; details of some nodes (b) with poor performance, and (c) revised version of the column in PETG nodes

In addition to the role of the white nodes in the collapse, two other issues increased the likelihood of structural failure: wet battens and split nodes. A few days before the full assembly, the column section was exposed to rain, causing some battens to become wet and swell in size. This increase in size led to early cracks in some of the node boxes. Furthermore, since all nodes, except those in contact with the ground, were split into two parts, the overall stiffness of the structure was reduced. This allowed for small movements between the parts, which, over time, led to loosening.

The redesign of the column section was driven by avoiding all those previous mistakes. The new design (Figure 11c) included the following changes:

Material Change: All nodes, except those connecting the canopy section, were 3D printed using PETG instead of PLA+. This decision was based on the superior strength capacity of PETG-based components when subjected to similar structural demands (Morales-Beltran et al., 2022).

In total, seven nodes were completely redesigned due to their previous low structural performance (i.e., overall weakness and/or lack of stiffness). However, modifications to the position and length of the boxes were avoided, as doing so would have required changes to the corresponding timber battens. Instead, the modifications focused on increasing the wall thickness of the node boxes and/or increasing the infill percentage to 20%, in order to improve their strength and stiffness.

Out of the initial 11 students, only four remained actively involved in the redesign and reprinting of the now PETG nodes. This was likely due to the students' busy schedules, as this work was not officially part of their coursework. As a result, their involvement in the later stages of the project was voluntary. For the same reason, the design and printing processes took several weeks. Since there was no time available during the semester, the assembly of the new prototype was scheduled for the summer break.

The assembly took three full days and followed the same order as the previous prototype: first, the column section was assembled, followed by the canopy section, and finally, the two were connected to complete the prototype (Figure 12). An additional three students joined the team specifically for the assembly, bringing the total team size to eight.

Figure 12.

Assembly process of the second prototype

The second version of the canopy weighed approximately 35 kg, distributed as follows: 17 kg for the 3D-printed nodes, 15 kg for the timber, and 3 kg for bolts, washers, and nuts. Despite its lightweight design, the structure performed well in terms of strength and stability. However, a few days after assembly, the team noticed that the tip of the canopy had lowered significantly. By design, the tip of the canopy – the highest point of the structure – was intended to be 2.89 meters above the ground. Immediately after assembly, this height was 2.50 meters, which was initially attributed to natural adjustments in the structure due to the changes made during assembly.

A week after the assembly (Day 17), the deflection became more noticeable, so the team began measuring it (Figure 13). This coincided with a summer rain and a slight increase in temperature (above 20°C). Over the following days, temperatures rose above 30°C, causing the nodes to literally melt. This led to significant deflections in the canopy, and eventually, the tip of the canopy touched the ground.

Figure 13.

Deflection of the canopy

The water absorbed by the battens was not a factor in the overall strength of the new prototype (Figure 14a). However, due to the extreme deformations experienced by the structure, almost all nodes suffered some degree of damage. Common types of damage included torsional, shear, and bending moment-induced failure (Figure 14b). The size of the nodes also influenced their behavior during the extended deformations caused by the hot weather. For example, smaller nodes tended to become more flexible and absorbed most of the rotations by deforming their boxes. In contrast, most of the larger nodes developed cracks due to their lack of ductility – except for the node depicted at the bottom of Figure 14c, which suffered no damage whatsoever.

Figure 14.

Performance of the nodes during the extreme deformation of the canopy

During the production and assembly phases, the team learned that the design-to-production process is inherently iterative. Problems identified during production often required a return to the design phase, where adjustments were made. While the first two stages of the course were completed more or less according to the original plan, the design and 3D printing of the nodes took up most of the remaining time (Table 4). The most obvious omission in the original schedule was the lack of adequate time allocated for the iterative process – particularly for revising the node designs. The original versions were full of mistakes, and the reprinting of the nodes was required not only due to design changes but also because of adjustments in printing settings, such as infill percentage, printing orientation, and filament type.

Three major factors related to technology influenced the delays and iterations throughout the entire process. First, a lack of awareness regarding incompatibilities between Rhino and other software created complications in the production process and led to significant time losses. These experiences have shown that software compatibility should be considered at the outset of the design process to avoid such issues in the future.

Second, variations in printer quality further complicated the process, producing nodes with inconsistent strength. The resulting design changes in the canopy caused connection problems with the battens and extended both the redesign and reprinting phases.

This study also examined the structural capacity and durability of nodes produced with 3D printers, highlighting the strengths and limitations of the technology. The issues encountered with the PLA+ nodes in the column section of the first prototype underscored the importance of selecting the appropriate material based on the expected environmental conditions. This aligns with findings of previous research (Morales-Beltran et al., 2021), which emphasized the need for careful material selection and early-stage planning to mitigate structural integrity issues in digitally fabricated designs.

The project emphasized teamwork, allowing students to collaborate on the design, fabrication, assembly, and disassembly of a gridshell prototype. This required coordination among team members and provided them with valuable opportunities to develop communication skills. As Carpenter (1997) noted, “students learn that architecture is a collaborative effort and not an exercise in isolation.” For the students, adapting to challenges and rapidly solving problems based on lessons learned was crucial in overcoming obstacles. Effective time management also played a key role in the successful completion of the project, from initial design to final assembly. Creating and adhering to a clear timeline was essential to avoid delays.

This aligns with the construction industry's need for graduate architects who not only possess modern technological and computational design skills but also the ability to oversee and manage project development and construction processes (Chan Edwin H. W. et al., 2002). The knowledge and experience gained from this course will significantly enhance students’ ability to improve and expedite the design, production, and assembly of building structures and components.