Innovative design in timber
The first double-layer timber gridshell in the UK has recently been completed as part of a new building for the Weald & Downland Open Air Museum in Sussex. Michael Dickson and Richard Harris of Buro Happold explain a unique engineering accomplishment and the construction processes behind it.
The Weald and Downland Open Air Museum is a leading international centre for historic timber buildings. The new purpose-built conservation centre and store will allow the museum to make its research, conservation and restoration programme for vernacular buildings accessible to the general public. In their commission, the Museum aspired to a modern structure, extending the lineage of timber buildings in the collection into the twenty-first century. The brief called for an exemplary structure for modern rural buildings. Sustainability was one of the key issues of the brief given to the team of architects, engineers, surveyors and constructors.
Despite costing only £1.3m, this construction has featured in the national press and created international interest because the building is so unusual in its construction technique and in the resulting architecture (Figures 1 and 2). Conceiving, creating and constructing this building can also be seen as an example of how engineering methods can turn problems into potential solutions.
The workshop is housed beneath a double-layer gridshell of oak laths supporting an insulated cladding system largely of lapped western red cedar boards and polycarbonate cellular glazing. The double-layer timber gridshell technique, despite being able to achieve large spans with lightweight but stiff construction, has only rarely been used. So this building contains a number of features that are innovative and could be adopted for future similar use. Here the workshop roof is a doubly curved, four-layer, oak gridshell, 48 m long, 16 m across at its widest and 11 m at the waist. Internal height varies between 7 m and 10 m. It uses oak laths of only 50 × 35 mm at 1000 mm or 500 mm spacing, depending upon stress concentrations. This space sits over a floor of 50 mm boards for the craftsman to work on that in turn encloses a conservation store nestled into the north-sloping chalk terraces of the Sussex Downs.
The deciduous wood on the slope to the south provides shading in summer, which, coupled to the thermal mass of the enclosing masonry walls and reinforced concrete floor to the store, helps reduce air-conditioning requirements.
At an early stage of the design, the engineers made a scale model, using wooden laths (Figure 3). The model-making process provided valuable experience on the behaviour of the shell during forming. The model itself was valuable in helping to inform the rest of the team and for use in fund raising.
Why a double-layer gridshell?
There are limitations on the tightness of curvature to which laths, even when green, can be bent. The solution to this problem was to use a double-layer gridshell, illustrated in Figure 4. The lattice is composed of four layers, effectively two single-layer mats sitting one upon the other. The laths are of sufficiently small section (50 mm wide × 35 mm thick) of oak to permit bending of the lattice into the desired geometry. Upon completion of forming, timber shear blocks were positioned between the lath layers and fixed with screws. These transfer horizontal shear between parallel layers and endow the lattice with the properties of a deeper section to resist out-of-plane bending.
Selection of timber
A number of timber species were considered for use in the gridshell. These included larch, douglas fir, chestnut and oak. They were selected for the following reasons:
They are all naturally durable, making it possible to omit the timber treatment.
Oak is the most common structural material used in the museum’s collection of buildings.
They are all readily available from sustainable sources in the UK.
Tests carried out on laths at Bath University indicated that the performance of the oak exceeded that of the other species. While it was stiffer than the other timbers tested, needing a larger force to achieve a given curvature, it had a considerably higher bending strength, achieving a smaller bending radius prior to failure. In addition the failure mode was not sudden; there was a degree of plasticity at failure.
A significant problem noted was the variability of the bending strength of oak due to ‘short grain’. The straightness of the grain along the length of the log is a function of the way in which the tree grows. This varies from species to species and, while there are trees that grow with a more spiral pattern, oak does not grow with the straightness of grain of a timber such as Hemlock (chosen for the Mannheim gridshell for this reason). It is difficult to follow the grain of oak accurately when sawing timber out of the log. This means that the saw cuts across the grain, producing timber with a lack of continuity of longitudinal fibres, leading to considerable reduction in strength. It is for this reason that oak structural members have traditionally had large cross-sectional dimensions. In a sufficiently large timber the twisted grain can be contained within the section. In small sections, the problem is considerably increased. To overcome the problem, the latest jointing technology was utilised to cut out the defects and finger joints used to fasten the lengths together, forming laths of the required length and of a consistently high quality, produced from normal grade timber. The technical term for this process is ‘optimisation’.
The gridshell was analysed and designed in accordance with the Eurocode 5: Design of timber structures, using a timber grade of D30. Eurocode 5 uses load and material factors in the design; timber grade D30 has a characteristic bending strength of 30 N/mm2. To meet the design requirements, the 5-percentile characteristic bending strength had to be equal to, or in excess of, this value. Preliminary testing had proved that solid oak laths had the required strength and could easily achieve the 6 m radius of curvature required by the lattice. Thus the solid oaks were satisfactory and attention turned to the specification of the joints.
Preparation of oak laths
The oak was felled in Normandy in October 2000, sawn to 3 m lengths, 53 mm × 38 mm in section, and then delivered to the UK for further processing to create the ‘improved’ laths.
Due to its acidity, oak is a notoriously difficult material to joint with adhesives, all the more so when green. The green oak used for the Downland gridshell had moisture contents of up to 65%. The Swiss adhesive manufacturer Collano has developed a one-part polyurethane liquid adhesive under the trade name Purbond® HB 530, which is not adversely affected by the acidity of green oak. This adhesive has been designed to cure under the influence of material moisture and humidity; the ideal moisture content for curing is 18% but moisture contents greater than this will not inhibit curing. Curing is rapid without the need for expensive high-frequency heating. This adhesive cleared the path for the use of green oak. The timber was optimised in a highly automated process using the GreCon Dimter OPTICUT 101 mechanised saw. Each length of oak was visually inspected by a skilled carpenter to identify knots, unacceptable slope of grain and other defects. Visual grading entailed systematically marking the timber with a fluorescent crayon. The OPTICUT 101 has an optimisation computer that reads the fluorescent crayon marking, cuts out the defect and sorts, in accordance with the designated grade. Exact logging of the production data showed that the highest grade material ranged in length from 0.3 m to 1.4 m with the average segment being 0.6 m long.
Finger jointing was performed using the GreCon Dimter SUPRA finger-jointing machine. This is a continuous feed system. Fingers were cut simultaneously into the ends of the oak segments and the Collano Purbond HB 530 polyurethane adhesive was applied with the FLANK JET system that combs the adhesive onto the fingers. The segments were then aligned and pressed together at a pressure of 4 N/mm2, on the 50 mm × 35 mm section, to form 6 m lengths. The gridshell lattice would require 6000 linear metres of lath; considering that individual pieces of graded lath averaged 0.6 m in length, this represented 10 000 finger joints. Using the specialist machines this work was completed in three shifts. Although the timber had to be transported to the specialist machine the total weight was only 6 tonnes. Such a small quantity poses little difficulty to transport in one load.
The advantage of the above approach to produce the ‘improved’ oak laths was that the quality of the material was maximised very quickly and cheaply with minimum wastage.
Figure 7 shows a typical finger joint after the laths had been planed down to 50 mm × 35 mm section. The finger joint is almost indistinguishable within the lath; the low visual impact is one of the advantages of this jointing method.
The next stage in the process was to join the 6 m lengths of ‘improved’ timber to produce continuous laths up to 37 m long for the lattice laths and 50 m long for the longitudinal rib laths. This work was carried out on site under the protection of a polytunnel. The 6 m lengths were joined using scarf joints with a slope of 1 in 7. This slope gives the scarf joints a glue-line area the same as that for the finger joints. Figure 8 shows the construction of a typical scarf joint. There is an interesting contrast in the two jointing methods used: the finger joints are the latest timber joining technology whereas the scarf joint has been used for centuries.
Of the 10 000 joints in the structure there were only about 145 breakages during forming. Almost all of these were failures of the finger joints, pinching of the lattice on the scaffold support, tight curvature, tension buildup because the relative sliding between the two layers was being restricted and dry joints. The simple repair technique consisted of introducing solid blocking at the point of failure.
The double-layer gridshell is a lattice system with, after finishing, two 50 mm wide by 35 mm deep laths placed one above the other, with the space between them being formed by the lath system running in the opposite direction. Upon completion of the forming process, shear blocks were inserted to join the two layers; this formed a composite section that has significantly greater strength than the individual laths. To form the shape from a flat mat, the nodes must allow rotation. Also, with a double-layer system, because of the difference in their curvature and thus relative lengths, the upper and lower layers must be able to slide relative to one another.
The Downland Gridshell team developed a patented nodal connection. The central plate has a pin that inserts into the central layers. This fixes the central layers in position so that the nodes are a constant 1 m apart, and also enables rotation. The outermost layers are effectively passengers that are free to slide relative to the central layers but will rotate in tandem with its associated central layer as it is forced to do so due to the bolt arrangement of the nodal connection. Furthermore, two opposing bolts may be lengthened enabling the attachment of the rib laths that stiffen the gridshell.
This short article has described the selection, specification and testing of the timber for the gridshell. It has not been possible to describe the formfinding, modelling and analysis that were also vital to the success of the project. The construction itself has been fully described in a recent paper in The Structural Engineer2, which gives a complete description of the forming techniques. A gridshell is not suitable for all roofs but neither are other techniques universally applicable. Timber gridshells can provide an efficient and architecturally expressive way of covering space. There is no reason, now that the technique has been demonstrated to be a safe and reliable form of construction, that more should not follow.
This has been a team project and many individuals, from the various organisations involved in the design and construction, have made significant contributions in realising this unique project. They are as follows:
Client: Weald and Downland Open Air Museum
Funding: The Heritage Lottery Fund
Architect: Edward Cullinan Architects Engineers: Buro Happold
Project Manager, Quantity Surveyor and Planning Supervisor: Boxall Sayer
Carpenter Green Oak Carpentry Company Ltd.
Main Contractor: E. A. Chiverton Specialist Scaffolding Contractor: PERI
The monitoring and recording of the process were made possible through a research grant from the Department of Trade & Industry (DTI) through the Fast Track Research Fund.
- New Civil Engineer, 23 April 1998. Letter from C.L. Wallis.
- O.J. Kelly, R.J.L. Harris, M.G.T. Dickson, J.A. Rowe (2001) ‘Construction of the Downland Gridshell’, The Structural Engineer, 79 (17), 4 September.
Forming the gridshell
What is a timber gridshell? Timber gridshells enable relatively complex forms to be created from an essentially flat but predetermined grid of members which can be ‘pushed up’ (or lowered down) in a shape accurately predetermined and analysed by extensive use of computer software.
The final form is generated by deformation of the original square grid into parallelograms of different included angle depending on the local curvatures of the final surface. The development of this doubly curved gridshell from the flat rectangular grid is made possible by the low bending and torsional stiffness of the individual 50 × 35 mm oak laths. During forming, the timber lattice must allow rotation at the nodes and bending and twisting of each constituent lath. Once the grid is formed, ‘shell’ action is accomplished by an additional layer of timbers (rib laths) that triangulate the parallelograms, thereby providing inplane shear stiffness. This is in contrast to the first double-layer gridshell erected for the Bundesgartenschau in Mannheim, Germany in 1975.1 At Mannheim, crossed steel tension cables provided this triangulation and the structure was then clad in a translucent PVC polyester membrane.
For the Downland gridshell, the bracing was formed with timber rib laths acting as ties, which also support the cladding. These triangulating members running longitudinally in the lower half of the shell and transversely across the crown of the shell were called ‘rib laths’ and fixed directly to the patented steel nodes of the gridshell. The development of a doubly curved gridshell from a flat, square or rectangular grid is made possible by the low torsional and bending stiffness of timber.
Specification and validation
The specification stipulated:
a maximum slope of grain on either face of 1:10
no dead knots; no live knots; small clusters of pin knots were allowed provided that they did not form more than 20% of the width of any one face
no shakes or splits
no sapwood (heartwood of oak is naturally durable and resistant to infestation but its sapwood is not).
Using the finger jointing technique, this specification could be achieved with almost any source of oak. The problem is that the lower the quality, the greater the number of defects and the larger the amount cut out and discarded. This adds significantly to the cost of the final product. It is necessary to find the balance between low cost, poor quality timber, with a high rejection rate, and higher cost source material with a lower rejection rate. For this reason, suppliers suggested that the timber should be sourced in Normandy.
The laths and finger joints were tested in a four-point bending test in accordance with sample dimensions given in BS EN 408:1995: Timber Structures – Structural Timber and Glue Laminated Timber – Determination of Some Physical and Mechanical Properties. The samples were tested green and not conditioned to the requirements given in BS EN 408:1995. At the early stages of the project, a whole series of finger joint testing was undertaken examining a range of variables including capacity about both axes, performance of different adhesives and effect of different production pressure. For the production stage, a quality control system was implemented to ensure the effectiveness of the finger joints. Quality control testing was carried out in two stages: pre-production tests and batch sampling. The results of these tests enabled a statistical analysis in accordance with EC5 Annex A, Section A2: ‘Determination of 5-Percentile Characteristic Values from Test Results and Acceptance Criteria for a Sample’. The pre-production tests determined the 5-percentile characteristic bending strength of the finger joints, and coefficient of variation for the running production control. The batch testing ensured that the probability of accepting a sample with an ultimate bending strength less that the desired 30 N/mm2 was within acceptable statistical limits set out in EC 5 Annex A.
Michael Dickson FREng and Richard Harris
Richard Harris is a structural Engineer working in the Buro Happold Bath office. He specialises in the field of timber engineering, leading Buro Happold's work in this field, working with researchers at Bath University, TRADA Technology and BRE, lecturing widely and organising the annual ‘Time for Timber’ conferences. Timber engineering projects include: Weald and Downland Museum Gridshell; Globe Theatre, London; Cambridge Botanic Garden Education and Interpretation Centre; Cork Airport terminal building; Norwich Cathedral Visitor Centre; Caerphilly Castle Visitor Centre; Sheffield Winter Garden; Roundwood buildings at Hooke Park in Dorset.
Michael Dickson is Chairman of Buro Happold and a founding partner. He is a Visiting Professor of Engineering Design at Bath School of Architecture & Civil Engineering and was a Member of the Standing Committee on Structural Safety. He was Chairman of the Construction Industry Council from June 2000 to June 2002 and is currently a board member of TRADA and Chairman of nCRISP, the Construction Research and Innovation Strategy Panel. His projects have ranged from cable nets in Jeddah and a tensioned stainless steel mesh aviary at Munich, to many integrated building projects, including BA Waterside, Al Faisaliah Complex, and the HQ for Wessex Water. His recent projects in timber include work on green forest thinnings at Hooke Park, Weald and Dowland Museum and Japanese Pavilion Hannover.