Surface splits in outdoor undercover unfinished glued laminated timber

1 Introduction

Glued laminated timber products may be used in applications, where product surfaces are exposed to outdoor environmental conditions. Deterioration of surfaces by environmental agents are generally referred to collectively as “weathering”. This term covers multiple deterioration agents, including thermal, biological, mechanical, physical and chemical mechanisms (Williams 2010). Weathering of timber surfaces may be significant enough to adversely impact structural, fire, durability and/or aesthetic performance. Longitudinal surface fissures are one type of deterioration characteristic seen in weathered timber. These fractures occur in-service and are usually orientated along the grain, with a relatively narrow width perpendicular to grain and some depth into the surface on which they occur. These fractures are known to arise because of changes in atmospheric conditions, causing moisture gradients in wood and inducing surface stresses (Franke and Franke 2014; Peck 1957). This paper refers to these in-service surface fractures as “surface splits”, (Figure 1), rather than fractures, cracks, checks, fissures or other terms. Despite pre-seasoning of timber products to ambient service conditions and the use of finish systems, this type of surface deterioration can still be observed in-service. While surface finishes, like paints, stain and water repellents, are often used to minimize split development, a base study of unfinished glued laminated timber has not been done. In this study, “unfinished” means a timber surface in a raw prefinished state with only machine dressing having been applied to the surface, and without additional surface products applied.

Figure 1:

Typical surface splits in glued-laminated timber exposed to outdoor undercover environmental conditions, including direct solar radiation, but excluding precipitation. Splits occur within laminates and at or near gluelines.

Research methods investigating surface splits in glued laminated timber involve various approaches, including.

1. Qualitative studies of existing glued laminated timber surfaces in structures (Al Sabouni-Zawadzka et al. 2022; Blass and Frese 2007; Dietsch and Winter 2009).

2. Laboratory testing in climate chambers to simulate conditions, which induce surface splits (Angst and Malo 2012).

3. The modelling of surface splits by using observed field conditions and simulating surface performance, based on mass transfer physics and the known responses of wood to temperature and moisture (Fragiacomo et al. 2011; Huč 2019). Modelling methods generally examine stress occurrence, and the atmospheric moisture and temperature conditions at which an environmentally induced stress may exceed the strength of the timber, resulting in a surface split.

Field exposure trials are a common approach used in durability research, where the changing performance of a material is measured over time to better understand the deterioration initiation and development process, and to provide evidence or otherwise demonstrating a products in-service performance. An example of this approach in unglued timber is a 61 month outdoor exposure trial that studied “cracks” in Pinus sylvestris and Picea abies (Sandberg 1999; Sandberg and Söderström 2006). This study made findings relevant to the “weathering” of solid timber surfaces and showed the benefit of field exposure trials.

Several topics potentially influencing surface splits in glued laminated timber in outdoor undercover climates have not been well studied; including precipitation influences, potential surface size influences, solar radiation influences and adhesive joint influences. The stabilization of surface splits over time in glued laminated timber in outdoor undercover climates has also not been well researched.

The direct wetting of glued laminated timber surfaces is generally incompatible with manufacturer’s recommendations for reliable outdoor use (Glued Laminated Timber Association of Australia 2020). Shielding and capping of glued laminated timber is recommended building practice to avoid or minimize direct wetting. The direct wetting of glued laminated timber surfaces also has the potential to reduce the visibility of surface splits, due to swelling, mould and other biological activity, making it potentially difficult to study.

The author has observed that surface splits in wood appear to stabilize in size over time as a surface becomes more fractured and climatic induced surface stresses are relieved. The deterioration of an as-new (split free) surface, towards a stabilized surface condition, suggests a staged phenomenon.

Size effects are well established for stressed timber members (Leicester 1973; Madsen 1992). Atmospheric induced stresses in timber are however slightly different from an external force or deformation, induced stress in a section. Size effect studies in timber are based on applied external stress and are theorised via the distribution of “weakest links” within the volume of a stressed member (Madsen and Buchanan 1986). The stress in a section will increase where a load is constant and a series of weak point failures occurs within a volume of wood, thus reducing the capacity of a member to withstand load, as fracture number and size increases. In the case of atmospheric induced shrinkage strains in wood, surface split occurrence may result in stress relief, reducing the stress in a member, rather than increasing it. This matter has been modelled (Fragiacomo et al. 2011) and it was found that size influenced mean moisture content but had less effect on near surface gradient and stresses. Perhaps larger surfaces, with more inherent weak points may provide greater stress relief and result in smaller and fewer surface splits? A study of split occurrence in different width glued laminated timber sections may provide evidence for a surface split size effect, or otherwise.

Solar radiation exposure is known to have an influence on surface split occurrence and development with more severe exposure being linked to more noticeable surface splits. For instance, the cracking of window frame joints due to solar radiation (Castenmiller 2004). The influence of direct solar radiation on surface split occurrence and development in glued laminated timber however has not been quantified.

A unique aspect of glued laminated timber relative to solid timber, is the use of an adhesive material between adjacent laminates, causing an inherent material discontinuity. This introduction of an additional material (adhesive) and its bonding characteristics to the wood, may also create potential for surface splits, distinct from splits within the wood structure itself. Were a glued joint separates in-service, it may be termed as delamination, implying a failure of the laminating process. However, where a surface split occurs near a glued joint there may be other mechanisms causing splits. In such cases it would not be appropriate to apply terms like delamination where other mechanisms may be the cause. In this study, the locational phase “at or near the glueline” has been used to locate surface splits without implying causal failure mechanisms.

Questions remain about the nature and causes of surface splits in glued laminated timber, i.e. Do surface splits occur and develop in stages, Do surface dimensions influence surface splits, How much does solar radiation influence surface splits, Do gluelines influence splits? It is hypothesised that a longitudinal field exposure study of surface splits in glue laminated timber, may gather evidence to address these questions.

2 Materials and methods

A field trial was designed to evaluate the identified research questions. The influence of solar radiation exposure was considered a key question with three surface orientations included, i.e. horizontal, vertical north facing and vertical south facing. Horizontal surface orientation was included, as this plane is the standard measurement orientation for solar radiation reaching the earth’s surface and represents an orientation plane for glued laminated timber elements in structures e.g. beam top edges, deck surfaces. The two vertical planes selected also represent planes found in structures, including column faces or the vertical faces of beam members.

The selected field trial site in Hervey Bay, Queensland, Australia, with a latitude of −25.306° results in a solar radiation daily dose on a vertical north face, which is usually less than the maximum horizontal daily dose at this location, except near the winter solstice where it is slightly greater due to the suns declination at this point in the solar cycle. A south facing vertical surface was selected as it represents a negligible solar radiation dose.

Product width was also considered a key parameter and three surface widths where included, i.e. 5 laminates (160 mm), 10 laminates (320 mm) and 15 laminates (480 mm). In this study on surface deterioration, the term “width” refers to the magnitude of the surface perpendicular to the longitudinal orientation of laminates. Longitudinal magnitude is referred to as “length”. In a beam application, width and length, may be referred to as depth and span respectively.

The species of timber used in the trial was a hybrid Pinus species developed and grown in plantations in Queensland Australia. This is a locally developed hybrid of slash pine (Pinus elliottii var. elliottii) and Caribbean pine (Pinus caribaea var. hondurensis). This resource is used commercially in Queensland, Australia to manufacture timber products, including glued laminated timber. The timber used in the trial was sourced from an earlier characterization study of this resource (Bailleres et al. 2019). This timber had been stored in dry open undercover conditions for several years in Brisbane (250 km south of the trial site), with good acclimatization to ambient environmental conditions. Sawn sections, approximately 100 × 38 mm in size were selected to have a stratified distribution of typical surface characteristics, including some natural characteristics, like bluestain and knots. Surface split like characteristics, such as drying checks and resin shakes were excluded.

A total of 30, 1.8 m long pieces were selected and sorted into two groups based on sawing orientation, referred to as either tangential faced or radial faced. This was done to reflect marketplace manufacturing practices where adjacent laminates often have similar sawing orientations. A commercially available polyurethane adhesive (PUR), Loctite HB S309 Purbond, was used to manufacture two beams. All laminates were machined to a 32 mm dressed thickness in preparation for beam manufacture.

Each of the two beams were manufactured from 1.8 m long laminates and comprised 15 pieces in each beam, with no glue used for half the length between laminate 10 and 11. Each of the two beams (radial and tangential) were then docked to provide 2 sets of 9 sub-beams, as follows; 6 sub-beams at 480 mm × 290 mm (15 laminates), 6 sub-beams at 320 mm × 290 mm (10 laminates) and 6 sub-beams at 160 mm × 290 mm (5 laminates). Refer to Figure 2 which shows a cross section of each beam. Each of the 9 sub-beams docked from the two manufactured beams contained matched wood samples to help ensure that any observed surface split differences would not be due to wood variability. A total of 18 sub-beams at 290 mm in length and either 480, 320 or 160 in width, were manufactured, i.e. 2 beam types, by 3 beam widths, by 3 surface orientations. These sub-beams became the test specimens, with a single surface selected for exposure.

Figure 2:

End grain images of the 30 individual laminates used to manufacture the two beams used in the field trial, i.e. a predominantly tangential faced beam (lower image) and a predominantly radial faced beam (upper image). The images show the 15 individual laminates per beam machined to 32 mm per laminate. A total of 480 mm in width. Also shown via the dashed line is the separation interface between the 10 laminate (320 mm wide) and 5 laminate (160 mm wide) beams. The top face is the exposed surface in both beams with all other faces sealed, with a 0.63 mm thick aluminium foil adhered to the surface.

Following manufacture, the average density and average moisture content of the tangential faced beam was measured as 545 kg/m³ and 9.9 %, while the radial faced beam was measured as 510 kg/m³ and 10.0 %, respectively. All 18 trial sub-beams were sealed to restrict moisture movement on all, but the surface intended for exposure. Sealing was undertaken to isolate the test surface from climatic effects acting on adjacent surfaces, primarily the effect of relative humidity. Changing moisture on unsealed adjacent surfaces could cause either shrinkage or swelling, which in turn could influence the surface splits on the single surface being studied. The impact of adjacent surfaces on surface splits, has been excluded from this study as it was considered that surface splits on one surface should be understood before considering the climatic effects on combinations of surfaces. The sealing of specimen ends, sides and backs was undertaken with a two part polyester resin applied directly to the wood surface, followed immediately by a 0.63 mm aluminium foil applied directly to the tacky resin prior to its hardening. This effectively stuck the aluminium foil to the unexposed surfaces, providing a barrier to moisture vapour transfer and solar radiation, but not necessarily to temperature.

The 18 test specimens where then exposed to a sub-tropical environment located in Hervey Bay, Queensland, Australia, at a latitude of −25.3061° and a longitude of 152.8710°. The exposure site is 20 m above sea level and is located 2 km from an Australian Government Bureau of Meteorology (BOM) Weather Station (ID 040405, Latitude −25.32°, Longitude 152.88°, height 13 m). The climate in this region is classified as humid sub-tropical which is characterized by hot and humid summers and cool to mild winters (Peel et al. 2007). About 20 % of the world’s population live within this type of temperate zone with no dry season (Mellinger et al. 1999). This represents major climatic regions on all continents, expect Antarctica.

All test specimens were supported at least 400 mm above the ground with open ventilation to all exposed surfaces. Protection from direct precipitation was through the use of a 0.8 mm clear polycarbonate profiled sheet, positioned at least 200 mm above the horizontal specimens and overhanging both the horizontal and vertical specimens by a distance which formed an angle of approximately 40° from the sheet edge to the extremity of the specimen surfaces. Relative humidity and air temperature were both measured at 10 min intervals at the trial site using a commercial measuring device. The exposure trial commenced on the 1st of January 2021 and ceased on the 1st of July 2022, after 18 months. The trial commenced 11 days after the 2020 summer solstice, and it finished 11 days after the 2022 winter solstice.

Surface split occurrence was determined by visual means where distinct separation between adjacent sides of a split, using human vision, was detected. Human stereoscopic vision can discriminate binocular disparities as small as 2 s of arc (Kane et al. 2014). At a distance of 200 mm, the inspection distance used in this study, this represents a split width of 0.005 mm. The presence of a split gap of at least this width constituted the occurrence of a split and governed its length limits. Splits less than this width are likely to be discerned as just a line on the surface and are not considered in this study. Surface split length, was measured to the nearest 5 mm, using a conventional steel tape with 1 mm increments. Visual assessment was the preferred methodology for both split occurrence and length measurement, as this corresponds to human appraisals made in real structures by occupants, owners, builders, designers and material suppliers. The author’s eyes with reading glasses were the visual assessment tool used in this study. All measurements were undertaken in ambient daylight between 2 pm and 5 pm.

Surface split width and depth was not assessed in this study due to the smaller dimensional scale of these characteristics, the lower accuracy associated with measuring such small values, the variation in both width and depth along the split and the difficulty in determining split depth. It is assumed that split length is a general indicator of split dimension. Surface splits were also evaluated based on their location within the glued laminated surface. Splits wholly contained within the wood laminate where aggregated and reported as laminate splits. Splits at or near (within 5 mm) a glueline were aggregated and reported as glueline splits. This locational grouping of surface splits was intended to provide information about the impact of gluelines on surface splits relative to laminate splits alone.

The aggregate split length and aggregate number of splits for both laminates and for gluelines in each of the 18 specimens were determined at 3 to 4 daily intervals for 12 months and after that at about 14 day intervals. A total of 125 evaluations of this type were undertaken on all 18 specimens. Individual laminates and gluelines, within the 18 specimens were also evaluated. The aggregate split length and aggregate number of splits for individual laminates and for individual gluelines was determined at the start of every month. A total of 19 monthly evaluations were undertaken on the 180 laminates, and on the 162 gluelines.

It is assumed that the nature of surface split deterioration in surfaces may allow the use of a reliability function to analyse surface split measurements. A three parameter Weibull function, widely used in reliability, lifetime analysis and material science is proposed with the parameters being scale, shape (slope) and location. The fitting of the observed deterioration results to this model, expressed as an inverse sigmoid curve, together with the additional parameters of initial condition and final condition will be undertaken. The general model form is shown graphically in Figure 3. The model selected includes several parameters which are determined through optimization methods. The parameters giving the best fit to the model are then determined by minimizing the sum of squared residuals (SSR) using. The form of the inverse sigmoid expression used is shown in Equation (1), with each parameter explained below.

Figure 3:

General form of inverse sigmoid deterioration model used to interpret surface split results.

SSP: the surface split performance at a date.

Max.: the surface split condition, before exposure (1).

Min.: the stabilized surface split condition, after exposure (<1).

Slope: the rate of surface split per time, at the mid-point.

Date: the time after initial exposure.

MP: the mid-point time between max. and min.

The determination of a reliability value for each measured aggregated surface split length can be expressed on a 0–1 scale by comparing the measured aggregate surface split length with a standard length, relating to the length of the specimens. In the case of laminate splits, a standard length of 3 times the laminate length was selected, and for gluelines 1 times the glueline length was selected. The 3 and 1 values were chosen so that the maximum aggregate surface split length was about 0.5 on the 0 to 1 scale.

3 Results

3.1 Split occurrence and development characteristics

Surface split occurrence and development over the trial period is shown in Figures 4 and 5 for the 15 laminate wide tangential surfaces, exposed horizontally. These figures show variations in surface split lengths between measures, with increases and decreases occurring due to climatic changes. These figures highlight the magnitude differences between laminate surface splits, and surface splits at or near gluelines. They also highlight the increased variation in laminate surface split length and number, compared to gluelines. The relative surface split performance shown in these figures for this one test specimen, is similar to all other 17 specimens, except for magnitude differences due to surface and sawing orientations, as well as due to exposure. The remainder of this paper will focus on the aggregate length of surface splits in laminates and gluelines and not the number of splits. Figures 4 and 5 indicate stages of split occurrence and development, i.e. an initial period of no splits (glueline splits only), a period of surface split initiation and extension, and a period of relative stabilization of surface splits.

Figure 4:

Aggregate surface split lengths (mm) in the 15 laminate (480 mm) tangential surface, orientated horizontally.

Figure 5:

Aggregate surface split numbers in the 15 laminate (480 mm) tangential surface, orientated horizontally.

The split performance measures (aggregate laminate surface split length) were converted to a reliability value from 0 to 1, where 1 represents no splits and 0 represents an aggregate split length equal to 3 times the surface length. The selection of a 3 times length, to compare surface splits in laminates, was also based on getting a good relative separation of different surface performances. The fitted deterioration models for the three matched 15 laminate tangential surfaces in the three exposures (Figure 6) allow surface split start, finish and duration times to be estimated. Comparing the three exposure orientations shows that deterioration in a horizontal exposed surfaces started on the 1/1/2021 (Day 1) and ended after 109 days on the 19/4/2021 with about a 26 % deterioration from the initial condition. The vertical north facing surfaces deteriorated from the 2/3/21 for about 200 days until the 17/9/2021 with about a 20 % deterioration from the initial condition. The vertical south facing surface deteriorated from the 5/1/2021 for about 301 days until the 2/11/2021 with about a 9 % deterioration from the initial condition.

Figure 6:

Reliability models fitted to aggregate laminate surface splits results for the three 480 mm (15 laminate) tangential surfaces. Vertical scale: 1 = no splits, 0 = 3 times the laminate length of the surface.

The 18 month aggregate solar radiation dose on a standard 320 × 290 mm surface, was determined to be 267 kW, 121 kW and 5 kW, for horizontal, vertical north and vertical south orientations, respectively. Figure 6 also shows that the surface splits in the vertical southern facing surfaces commenced before the surface splits in the vertically north oriented surfaces. This is consistent with increased solar radiation on the south facing surfaces at this time of year, given the more vertical position of the sun in the sky and it’s rising and setting in the southern sky. Figure 7 highlights the surface split developmental differences in matched glued laminated specimens. The models suggest that splits in surfaces with horizontal exposure commence almost immediately and this progresses over about 6 months to a stabilized split surface which is more severe relative to the other two exposure orientations. The vertical facing exposure results show surface splits developing at a slightly slower rate and not reaching the same final magnitude as the horizontally exposed surfaces. Surface splits at gluelines showed a similar staged evolution with a reduced overall split length for decreasing solar exposure.

Figure 7:

Reliability models fitted to aggregate laminate surface splits results for 9 tangential surfaces. Vertical scale: 1 = no splits, 0 = 3 times the laminate length of the surface.

3.3 Solar radiation

The three combinations of surface orientations and directional exposures adopted in this trial, represent solar radiation exposure which may be encountered in timber structures, i.e. from the most severe (horizontal) with full sun exposure, to partial sun exposure (north facing and vertical) and to negligible sun exposure (southern facing and vertical). Solar radiation varies throughout the day due to the suns position in the sky and the magnitude of any radiation obstruction, such as clouds, vegetation or other structures. Vegetation obscured some radiation in this trial, particularly in the early morning and late afternoon when the sun was low in the sky. At the latitude of this field trial, the relative solar radiation throughout the day and the year is shown in Figure 8 for the solar extremes of the solstice events. Figure 9 shows the daily radiation dose received on the surfaces, based on the horizontal solar radiation measurements at the nearby BOM weather station.

Figure 8:

Relative solar radiation in the three exposure orientations.

Figure 9:

Daily solar radiation doses on a horizontal surface from Bureau of Meteorology data. Vertical surface daily doses are calculated based on the angle to sun throughout the year.

These figures highlight that the south facing surfaces received more solar radiation than the north facing surfaces at the summer solstice when the sun rises and sets in the south and is at its highest position in the sky. Also, the north facing surfaces received slightly more solar radiation than the horizontal surfaces at the winter solstice due to the angle of incidence resulting from the trial latitude. Figures 6 and 7 shows surface split differences between each of the 3 surface orientations and exposures. It is assumed that the relative humidity and temperature influence on splits is consistent for each surface orientation, as open air flow to all exposed surfaces results in similar atmospheric temperature and moisture conditions at each surface. This assumption allows the evaluation of split differences based on solar radiation alone.

An analysis of the stabilized split lengths for laminates and gluelines was undertaken and compared with the average daily solar radiation dose on the exposed surfaces. Figure 10 shows the relative laminate and glueline split deterioration at stabilization in each of the 3 surface widths by the two surface types (tangential and radial) comprising a total of 30 laminates and 28 gluelines. Each data point in Figure 10 represents an average of the stabilization levels in the 6 surfaces at each solar radiation exposure level. Relative deterioration is plotted against the aggregate solar radiation dose received on that surface over the 18 month exposure. It should be noted that the relative deterioration scale (y-axis) is different for laminate surface splits and for surface splits at or near gluelines. A value of one in both cases represents the initial exposure condition, i.e. no splits in laminates or at or near gluelines. A value of zero in the case of laminates, means that the aggregate length of laminate surface splits is 3 times the length of the surface in which they occur. A value of 0.66 for instance, represents an aggregate surface split length equal to the total laminate length of the surface. While for gluelines a zero represents a total aggregate length of surface splits at or near gluelines equal to the total length of gluelines. Horizontal surfaces, with higher solar radiation exposure, showed splits immediately while split initiation was delayed in the two vertical exposure orientations, with less solar radiation exposure. Increasing solar radiation exposure also increased the split rate, reduced the split initiation time, reduced the split duration time and reduced the time to split stabilization.

Figure 10:

Influence of 18 month average daily solar radiation (W/m²) surface dose on the relative laminate and glueline split deterioration at split stabilization. Surface splits for three different surface widths are combined with two surface types (tangential and radial) at the same solar radiation exposure.

4 Discussion

The occurrence and development of surface splits in unfinished glued laminated timber shows considerable variation and appears to be influenced by multiple parameters. This study has helped verify some of the anecdotal reports surrounding the occurrence and development of splits in glued laminated timber. In particular, the often noted observation that surface splits, stabilize over time; usually within a few months of exposure. Stabilization of splits was confirmed in all the surfaces trialled. The time to reach stabilization appears to be dependent on a range of climatic and wood factors.

Possible stages of surface split deterioration can be discerned from the results and interpreted via the proposed inverse sigmoid model shown in Figure 3. Table 2 summarizes four proposed stages in the progression of surface splits in glued laminated timber exposed to an outdoor undercover environment. A possible reason why surface splits may stabilize over time, is the stress relief provided by the split itself i.e. the ability of a split to open and close with changing conditions. Stabilization may be the point at which climatic conditions are no longer sufficient to cause new surface split formation, because existing splits open to accommodate all climate induced surface stresses.

The rate of surface split occurrence and development was found to increase with increasing solar radiation exposure (Refer Figures 6 and 7). The time to reach a stabilized deterioration level appears to be inversely proportional to the level of solar radiation exposure, i.e. the higher the solar radiation exposure, the shorter the time to split initiation and the higher the rate of surface split occurrence and development. A higher deterioration rate results in a shorter time to a stabilization and a more severely split surface.

Solar radiation has an obvious adverse effect on the occurrence, development and stabilization of surface splits in glued laminated timber. One explanation for this may lie in the impact of solar radiation on surface temperature, which is known to influence vapour diffusion in wood (Siau 1984).

A size effect relating to the occurrence and development of splits was not substantiated by this study. The evidence suggests that splits are no more or less likely to occur in wider or narrower surfaces. In hindsight, the trial design to evaluate a size effect was not ideal as surface width was limited to three widths only (480 mm, 320 mm, 160 mm) which only allowed a matched comparison of the 160 mm and 320 mm widths, to their matched surfaces within the 480 mm wide surfaces. The 160 mm wide surface comparison was also only based on 8 individual gluelines, which is statistically quite low. Glued laminated timber more than 320 mm in width is common, particularly in commercial and industrial applications, meaning that the section depth adopted in this trial is not as relevant to wider commercial beams. In principle, the resistance and consequence of stress/strain in a glued laminated timber beam surface, induced internally by climatic cycles is quite different from externally loaded timber members where a size effect is known. The cyclical nature of climatically induced surface stress, the surface only nature of the stress and the stress relieving nature of a surface split when it occurs, are not well aligned with the parallel establishment of a size effect in externally stressed timber members.

It appears that differences between latewood and earlywood play a part in surface split formation. Sandberg and Söderström (2006), commented that cracks on tangential sections first occur in the latewood. This study also found that latewood is an initiation point for splits on the tangential face. Lube (2015), videoed wood cracking during drying and this shows that denser latewood regions split before the lower density earlywood.

Wood gluing is based primarily on cohesive bonding of the adhesive to the wood. Adhesive penetration into the wood structure is essential to get mechanical interlocking so that the strength and stiffness of the wood itself controls bond performance rather than the adhesive. The achievement of a reliable mechanical interlock between the adhesive and the wood is fundamental to good gluing practices, with many manufacturing practices aimed at achieving reliable interlocking, i.e. adhesive application rate/methods, open and closed times, adhesive viscosity, wood temperature, clamping pressures and press times. Adhesive manufacturers provide generalized procedures for different wood species. In the case of some Pinus species, with a significant disparity between the earlywood and latewood wood structure, optimization of procedures may centre on the average density of the wood. Good adhesive interlocking in high density latewood may be difficult where procedures are optimized for average wood density, causing the type of delamination on the glueline seen in this trial.

5 Conclusions

An 18 month field exposure trial in sub-tropical Australia of 30 laminates in two unfinished glued laminated timber beams, each of which was docked into 9 sub-beams, with three different widths (480, 320 and 160 mm) and exposed to three different outdoor undercover environments (horizontal, vertical south facing, vertical north facing), resulted in splits in all 18 surfaces, leading to the following conclusions.

1. The nature of surface split deterioration in glued laminated timber can be characterised using a three parameter Weibull function expressed as an inverse sigmoid curve. Presenting deterioration data in this way allows key characteristics of surface split occurrence, development and stabilization to be estimated, i.e. the time from initial exposure to the start of surface splits, the rate of surface split occurrence, the time when stabilization of the surface splits has occurred and the relative deterioration of surface splits at the stabilization level. This characterisation approach provides a means of understanding and explaining surface split occurrence, development and stabilization.

2. The width of a glued laminated timber surface does not have an influence on the magnitude of surface splits resulting from undercover outdoor environmental exposure. The traditional size effect phenomena established for externally stressed timber does not appear to apply to cyclically stressed timber surfaces where surface splits provide stress relief rather than compounding the stress in the residual section.

3. Solar radiation exposure has a pronounced influence on surface splits in outdoor undercover glued laminated timber. Surface splits increase proportionally as the magnitude of solar radiation on a surface increases. A base level of surface splits occurs without any direct solar radiation. This is likely to be due to environmental cycling of air temperature and relative humidity, rather than any diffuse solar radiation impact.

4. Radially orientated splits on tangential surfaces are more likely to occur sooner and extend to a greater length than tangentially orientated splits on radial surfaces.

5. Surface splits in unfinished glued laminated timber exposed to outdoor undercover environments reach a stabilization level, where splits no longer initiate or extend, after about 6–9 months. This conclusion is based on an exposure commencement date of the 1st of January in the southern hemisphere. Other stabilization times may result where the exposure commences at a different time of year, or different conditions apply.

6. Latewood, in the Pinus species used in this trial, is more difficult to reliably glue relative to earlywood. Latewood on the glueline of surfaces was associated with more than 90 % of splits on the glueline.

7. The nature and location of latewood, both on exposed surfaces in laminates and at the surface in gluelines, will influence the occurrence and development of surface splits in glued laminated timber.