Glued laminated Robinia hardwood timber for structural use

Dill-Langer, G., Nieberle, R., and Hänsel, A. (2025). "Glued laminated Robinia hardwood timber for structural use," BioResources 20(2), 3848-3865.

Abstract

Robinia wood has a high technical and economic potential due to its future availability, its mechanical properties, and its durability. This also applies to its use in structural timber construction. Such applications require the manufacture of bonded construction products from this type of wood in order to compensate for dimensional shortcomings. In an application-oriented research project, tests were carried out for manufacturing technologies and resulting properties of glued laminated timber made from the hardwood species Robinia for structural purposes. Based on adapted visual grading rules, the strength and stiffness profiles of Robinia laminations were evaluated. The influence of different adhesive types and of production parameters on the strength and durability properties of glued finger joints and glulam bond-lines were characterised. By means of a simulation model based on X-FEM methods in combination with Monte Carlo simulations, the property potential of glued Robinia laminated timber was calculated. The model was calibrated by means of input parameters from the empirical lamination and finger joint test data and verified by a small series of full-scale glulam tests. The investigations showed the great potential of Robinia glulam, especially in highly loaded and heavily weathered applications.

Download PDF

Full Article

**Glued Laminated Robinia Hardwood Timber for Structural Use**

Gerhard Dill-Langer,^a,* Rupert Nieberle,^b and Andreas Hänsel ,^b

DOI: 10.15376/biores.20.2.3848-3865

Keywords: Robinia pseudoacacia; Bonding; Finger joints; Glued laminated timber; Timber construction; Delamination

Contact information: a: Department of Timber Constructions, Material Testing Institute, University of Stuttgart, Pfaffenwaldring 32, D-70569 Stuttgart, Germany; b: Duale Hochschule Sachsen (Cooperative State University of Saxony), Campus Dresden, Hans-Grundig-Str. 25, D-01307 Dresden, Germany;

* Corresponding author: Gerhard.Dill-Langer@mpa.uni-stuttgart.de

INTRODUCTION AND STATE OF THE ART

Robinia pseudoacacia

The cultivation of Robinia pseudoacacia (common name: black locust) as an economically relevant wood species is increasing throughout Europe. Today, European Robinia wood mainly comes from the south-eastern European region. However, Robinia is increasingly being cultivated in other parts of Europe, too. In Germany, cultivation is occurring especially on degraded sites such as former open-cast mines or slag heaps because of the species’ low demands on soil quality and its positive properties as a soil conditioner. In addition, the high tolerance to drought makes this species interesting in view of low precipitation levels over the past years, which can be regarded as probably typical for the effects of the ongoing global climate change.

Robinia has a height of about 20 m when fully grown in Germany. However, the branch-free trunk length is usually only a maximum of 10 m. In addition, the stem diameters are limited, which considerably restricts the species’ potential use as a non-glued structural building products today. Moreover, Robinia is extremely prone to crookedness, out-of-roundness of the trunk cross-section, and already deep-set twig growth (Grosser 1998).

Robinia wood is characterised by a high density and – at least on the level of clear wood – the mechanical properties surpass all other species growing in Central Europe. The excellent strength, hardness, and stiffness values makes Robinia wood an interesting candidate for demanding structural application. However, the mechanical properties decrease in many cases due to the aforementioned growth characteristics. For example, the tensile strength is already reduced by 20 to 22% at a fiber deviation of 3° (Göhre 1952).

The unique advantage of Robinia heartwood and the main reason for increasing interest in the application of this species for structural purposes are its excellent biological resistance properties. According to EN 350, it is assigned to the resistance class 1 to 2 against fungal attack (“very durable” to “durable”). This means that it surpasses the properties of all native European wood species and is in direct competition even with very durable tropical hardwoods. According to EN 460, Robinia wood can thus usually be used in hazard class 3.2 according to EN 350, i. e., permanent weathering without additional protection or even in hazard class 4 (in contact with soil or fresh water) without additional – for example, chemical – protection. Thus, Robinia wood with soil contact has a service life of over 25 years, and even 60 to 80 years for direct weathering without soil contact (Müller 2006). However, according to Koch and Dünisch (2008), there are some observations that this excellent resistance is restricted to mature wood, and the durability of the juvenile wood may be considerably lower.

Due to its typical growth characteristics, large straight, defect-free lengths can only very rarely be sawn from the trunks. One way to compensate for this is to use Robinia wood in the form of glulam. This offers the advantage that components of almost any dimensions can be produced from timber of short length and small dimensions. At the same time, local wood defects, such as knots or strong fiber deviations, can be compensated. However, the process of glulam production is a technologically demanding process.

Bonding of Robinia

The bonding of Robinia wood has already been investigated in a number of studies and research projects. The focus of such projects considering the main topic of this paper, the structural bonding of Robinia wood for load bearing building components, has been on surface bonding. However, in a limited number of investigations, finger-jointing has also been considered.

Surface bonding

In the literature, data on the surface bonding of Robinia wood can be found in various contexts. In the publications considered, adhesives based on one-component polyurethanes (PUR), phenol-resorcinol-formaldehyde (PRF), melamine-urea-formaldehyde (MUF), and epoxy resins (epoxy) were primarily investigated (Schickhofer and Hasewend 1999; Schickhofer and Obermayr 1999; Pitzner et al. 2001; Frühwald et al. 2003; Borysiuk et al. 2011; Voulgaridis et al. 2012; Konnerth et al. 2016; Kamperidou and Barboutis 2017). The bonding of Robinia is also a topic of current research (Berthold 2022).

When regarding the results of these publications, a high variety of the scientific statements are noticeable. This occurs both between different publications, and also within single publications. Thus, the bonding quality showed a strong dependency on the specific adhesive product or manufacturer, even within an adhesive family (Pitzner et al. 2001). The respective environmental conditions, and the wood moisture content, obviously also had a significant effect on the bonding quality, whereby the strength of the influence varied strongly in the different studies. Another parameter with reported effects is the applied clamping pressure during the bonding process (Kamperidou and Barboutis 2017). In addition, surface pre-treatments for optimising bonding were tested in a series of investigations (Varga and van der Zee 2008; Borysiuk et al. 2011; Lehringer 2014).

In general, however, a very heterogeneous picture is observed, so that no clear and consistent trends can be identified. In total, the PRF and MUF resins in particular showed positive results, primarily with regard to the shear strength achieved and also with regard to high wood fraction of the fracture surfaces. In addition, some PUR adhesives stood out as well.

However, the durability of the bond-line integrity – usually assessed by delamination tests – was studied only in very few investigations (Konnerth 2016). Being a generally accepted pre-requisite for the use in load-bearing timber structures, the lack of research on this topic is one of the current barriers for an introduction of bonded Robinia elements for structural use.

Glued finger joints

Due to the very uneven growth behaviour of Robinia and the resulting mostly very short sections of defect-free Robinia wood, a process for finger-jointing is urgently needed in order to be able to produce lamellas with a practical length. For non-loadbearing use, finger-jointed products are available on the market (e.g., Eurobinia). Examples of applications here are decking boards or cladding. The production of finger-jointed Robinia has proved to be problematic with regard to its suitability for use in load-bearing structures (Bernasconi 2004). However, some mechanical tests with structural finger joints are reported in some studies. A detailed paper on this subject is presented in Müller (2006), where different adhesives and finger joint geometries are investigated, and the results are compared. Frühwald et al. (2003) describe investigations on PUR-bonded finger joints and presents these as a significant strength weakness of Robinia laminated timber. However, it is problematic to derive clear lines or trends for certain processes or results in the available documentations. This again shows the difficulties in bonding Robinia wood and the very individual approach that is necessary when considering appropriate bonding methods.

EXPERIMENTAL

Material and Sampling

The investigations were carried out within an applied research project. The aim was to investigate the bonding properties of Robinia wood with regard to applications in the construction sector.

The Robinia timber used for the test program was of western Serbian origin, especially from the area of Bajina Basta. The trees used for the timber were about 30 to 50 years old. The saw mill, where the boards were sawn, usually produces timber for cladding, facades and other non- (or semi-) structural purposes. The timber was kiln-dried to a target moisture content of about 12%. Based on experience with glued non-structural products, only boards with small thicknesses of about 26 to 30 mm (after kiln-drying, but before planing) with target thicknesses for laminations in glued-laminated timber (after planing) of 20 mm were used. Most boards had a width of 120 mm, and some smaller boards with 50 mm width were produced. The boards were pre-graded and cut in two different ways for different purposes:

1. For the laboratory bonding tests of finger joints, short boards with 1 m length, free of knots and strong fiber deviations near the end-grain faces, i. e. graded according to the requirements of EN 14080, Annex B.3.1, were produced.

2. For the grading and tensile testing of boards without finger joints, boards with a length of 2 m and nearly saw falling quality (unsorted quality) were cut. Only a minor pre-grading was applied: Boards with extreme knot sizes, fiber deviations or mechanical damages were excluded by the staff of the saw mill, following the internal rules for “usual production quality” normally used also for non-structural products.

3. For production of glued-laminated timber and also for the industrial finger joint specimens, boards of the same origin, dimensions, and quality as for the lamination testing were produced; however, these had a smaller length of about 1 m, being the typical length of boards available for the “usual production quality”, i.e. with some minimum grading requirements.

In total, an amount of about 10 m³ of Robinia timber was produced and sent to the cooperating laboratories in Dresden and Stuttgart. The test program comprised the following steps:

– Strength grading of the laminations

– Tensile tests of boards without finger joints according to EN 408

– Tensile and bending tests of glued finger joints produced under laboratory and industrial conditions according to EN 14080, Annex E

– Adhesive tests according to EN 302-2

– Bending tests of glued laminated timber produced under industrial conditions according to EN 408

For the tensile tests a servo-hydraulic tensile/compression testing machine (MFL/Zwick 600kN, made in Germany) was used. The bending tests with laminations were performed by means of a electro-mechanical testing machine (Zwick 100 kN, Germany), whereas the bending tests with glulam beam were performed by a hydraulic testing machine with 8 pistons of a capacity of 200 kN each (Schenk, Germany). The delamination tests were conducted by means of the testing system “easyQ DLA-T” consisting of a pressure/vacuum vessel and a kiln dying unit (by KEMPF GmbH Solution+Project, Germany)

Grading and Strength Assessment of Laminations without Finger Joints

Visual grading was performed for the available 89 Robinia boards with a length of 2 m and cross-sectional dimensions of thickness t = 20 mm and width w = 120 mm. The grading was based on the definitions and criteria of the German standard for grading of hardwood DIN 4074-5. However, the grading and subsequent strength assignment aimed at a more tailored classification deviating from the threshold values of the standard grades LS7, LS10, and LS13.

Detailed inspections of Robinia boards showed that the most decisive properties were on one hand the knot area ratio (KAR), i. e. the knot or knot cluster size related to the board width, and on the other hand the global slope of grain. Thus, the individual values for these two grading parameters for all specimens were evaluated in this study. All other parameters defined in DIN 4074-5 conformed to the threshold values of the visual grade LS10. Although the presented study relates to the German visual grading standard DIN 4074-5, the most relevant parameters (KAR and global slope of grain) are part of most visual grading standards around the world.

Table 1. Tensile Tests of Robinia Boards Without Finger Joints

Fig. 1. Correlations between visual grading parameters and tensile strength. a) Tensile strength vs. knot area ratio (KAR). b) Tensile strength vs. global slope of grain (GSG)

For all specimens, the mass, dimensions, and the natural frequency of longitudinal vibration were recorded, and the raw density and the dynamic MOE were determined. The moisture content was determined by oven-drying method according to EN 13183-1. The tensile tests parallel to fiber direction were performed according to EN 408, section 13, in a servohydraulic test machine with special hydraulic clamps, which enabled a smooth load transfer in order to avoid unintentional fractures at the clamps. The free span of w = 1080 mm conformed to EN 408 and EN 384.

During the ramp load tests, the strain was measured by means of an extensometer and the tensile MOE was determined according to EN 408, section 12. The MOE and the raw density were adjusted to the standard moisture content of 12% according to the provisions of EN 384, section 5.4.2. The tensile strength was determined from the ultimate loads and the individual cross-sectional dimensions. Table 1 summarises the statistical parameters derived from the test results for tensile strength, adjusted MOE, adjusted raw density and moisture content.

Figure 1 shows the correlations between tensile strength and the visual grading parameters knot area ratio (Fig. 1a) and global slope of grain (Fig. 1b). The correlation coefficient 0.23 for KAR vs. tensile strength was larger compared to the value 0.10 for global slope of grain (GSG) vs. tensile strength. A statistically verified correlation could not be found in either case.

In order to investigate the relation between the accessible mechanical properties and the possible yield, three visual grade models with one grade (model 1) and two grades (models 2 and 3) were applied and compared. In Table 2, the threshold values for the visual parameters are given for the different definitions. The resulting characteristic values – determined from the test results according to EN 14358 – for tensile strength, MOE and raw density as well as the resulting yield ratios are also given.

Figure 2 shows exemplarily for the grade models 1 and 2 the effect of visual grading on the cumulative frequency of tensile strength.

Fig. 2. Effect of visual grading on the strength distributions: Comparison of model 1 (only one grade, no rejects) and model 2 (left) and model 3 (right) with two grades each. The individual strength values and the respective lognormal distributions are given.

Table 2. Definitions and Results for Three Different Visual Grading Models

If significantly higher strength values are desired, then model 2 would be the appropriate choice: In the higher grade A – with a yield of about one third of the total population – a characteristic tensile strength value of more than 50 N/mm² was achievable, whereby the MOE was only slightly increased. It is interesting to note, that the lower class B for this model 2 shows only marginal reduction of strength and stiffness values compared to model 1 with only one strength class.

Model 3 is an example of a moderate increase of tensile strength in the higher grade A, but accompanied by a much higher yield of 60%. The lower grade B shows a somehow more pronounced reduction of strength and stiffness compared to the ungraded material. All three models show a very high total yield of 100% for model 1 and 94% for models 2 and 3.

Glued Finger Joints

The finger joints were produced both on a laboratory scale and in an industrial plant. The milling in both cases was done by means of an industrial device, and the adhesives were always applied to the fingers manually. In the case of the laboratory production, the cramping pressure was applied by a testing machine in order to carefully vary pressure and cramping time. In the case of industrial production, an assembly press was used. The finger joint profile conformed to the most commonly used profile in central Europe with a length of 15 mm and a pitch of 3.8 mm. For characterisation of the performance of the finger joints, flatwise bending tests and tensile tests according to EN 14080 and delamination tests according to EN 301 Annex A were performed.

Laboratory production

During the preparation of the finger joint specimens, the importance of the quality of the milling equipment became apparent. Tests were carried out using milling tools with different service lives. For example, results when using a freshly sharpened tool could be compared with those for a tool that was already close to the end of its service life. The samples were then bonded and tested with the same adhesives and under the same conditions. One effect of the freshly sharpened tool could be observed with respect to the gap tip of the glued finger joint: With the worn cutter, the gap tip dimensions were in the range of approximately 2 to 4.5 mm, beyond the standardised range of max. 1.2 mm (for 15 mm finger joint length). On the other hand, when using the freshly sharpened cutter, care had to be taken that the tip gap did not become too small (standard value: <0.15 mm), which would result in reduced cramping pressure at the finger flanks and splitting of the wood at the base of the joint.

For the laboratory finger joint specimens, three adhesive systems (PU, MUF, and PRF systems) were used that had already been tested positively according to EN 301 or EN 15425 for load-bearing timber bonding. The specimens had cross sections between 52 mm and 60 mm and thicknesses between 20 mm and 24 mm. For comparison, solid Robinia wood specimens with the same dimensions and from the same batch were also tested.

The characteristic bending strengths of the finger joints bonded with PU and PRF adhesives exceeded with f_j,m,k (w= 62, PRF) = 71.8 N/mm² and f_j,m,k (w = 62, PU) = 72.5 N/mm². For comparison, the characteristic strength values of the solid Robinia wood from reference tests f_m,k (solid wood) was 68.5 N/mm², as usually requested for finger joints in softwoods. For the MUF adhesives, however, the characteristic value f_j_,m,k (w= 62, MUF) = 50.0 N/mm² was smaller than the reference values (for strength distributions, see Fig. 3).

The delamination tests revealed even greater differences. The PRF samples delivered the best values that were clearly below the limits of the requirements of EN 301. PU and especially MUF adhesives showed higher delamination values beyond the limits of the EN 301, with MUF performing the worst.

In summary, PRF adhesives performed best, fulfilling the standard requirements. The PU adhesives exhibited good strength values, but they did not fulfil the delamination requirements. The tested MUF adhesives showed the least promising behaviour both with respect to strength and delamination resistance. In subsequent research, it should be clarified whether a significant improvement of the finger joint performance could be achieved by special measures such as surface pre-treatment or specially structured milling cutters.

Fig. 3. Cumulative distributions of finger joint bending strength: results of three test series with specimen produced at laboratory conditions bonded with different adhesive types (MUF, PRF and 1K PU) and one test series with industrial specimens (PRF adhesive)

Industrial production

Based on the results from the laboratory tests, industrial scale trials were carried out to produce finger joints. Finger joint specimens with two pronouncedly different widths (w=50 mm and w=120 mm) were produced. The jointed Robinia laminations were obtained from the same batch as the non-jointed tensile test specimens. The timber conformed to grade model 1, i.e., no timbers were rejected according to visual grading rules. However, the board end grain faces were selected or cut in such a way that the vicinity of the finger joint was free of knots. A PRF adhesive system was used, and the adhesive was applied manually on the finger flanks of both adherends. Compared to the laboratory tests, the clamping pressure was significantly lower than in the laboratory tests, i.e. 10 MPa instead of 18 MPa. The lower clamping pressure had been chosen after some pilot tests in order to avoid occasionally observed crushing of the finger grounds. However, it became apparent that the lower clamping pressure contributed to far too large tip gaps in many cases and subsequently in partly very poor finger joint strength values.

The distribution of the test results of the bending tests with the specimens with smaller width (w = 50 mm) are given in Fig. 3 together with the laboratory tests. It turned out that the strength values were significantly smaller than the respective laboratory specimens glued with the same PRF adhesive system, however, with a still satisfying characteristic bending strength value of 48 N/mm². The bending strength distribution of the specimens with the larger width of 120 mm showed pronouncedly smaller values; however, due to less variation, they showed roughly the same value on the characteristic level.

The tensile tests were performed according to EN 14080, Annex E. The resulting finger joint tensile strength revealed a significant width effect: Whereas the characteristic strength for the smaller width was f_j,t,0,k (w=50) = 26.4 N/mm² the respective value for the larger width was considerably lower f_j,t,0,k (w=120) = 16.5 N/mm². The latter was caused by a rather large COV=35%. For an illustration of the strength distribution see Fig. 6. The resulting strength value on the characteristic level was rather low, even compared to softwood finger joints of lower softwood strength classes. Although the high strength values of the finger joints produced under laboratory conditions demonstrated the potential opportunities of finger joints in Robinia laminations, the results of the tests with industrially produced finger joints shows the risks of unfavourably adapted production parameters such as inadequately low cramping pressure.

Adhesive Tests

In Europe, the use of face-bonded glued timber products for structural use (including glued laminated timber (GLT), glued structural timber (GST) or cross-laminated timber (CLT)) is bound to a successful classification of the applied adhesives. In a two-step system, the adhesive is first assigned to an adhesive type for the bonding of spruce and pine following an exhaustive test programme defined in the standards EN 301 (PRF and MUF/MF adhesives), EN 15425 (1K PU adhesives) or EN 16254 (for EPI adhesives). In a second step, the applicability of the adhesive for other wood species (either hard- or softwoods) is evaluated by delamination tests according to EN 302-2 and a classification according to the requirements of EN 301. EN 302-2 prescribes not only the test procedure with wetting and drying cycles, but also the specific dimensions and layup of the small, laboratory-scale specimens.

Three different adhesives – one PRF adhesive and two different 1K PU adhesives – were tested according to EN 302-2 with Robinia laminations. All used adhesives were assigned to “type 1” for spruce and pine. According to the provisions in the test standard, the maximum thickness of the Robinia laminations – here 20 mm – was used. The timber was sampled from the same material used for bending tests; however, samples were specifically selected that were as defect-free as possible according to the criteria of EN 302-2. For one PU adhesive (“PU1”), two test series – one with planed laminations and one with sanded laminations – were performed. For the second PU adhesive (“PU2”), the – always planed – surface was treated in different ways: In addition to a reference series without treatment, five series with application of water based sprayed-on primer solutions were employed. The concentration of the primer solution ranged from 0% (only water without primer) to 2.5%, 5% and 10% and the relative amount of sprayed-on solution varied between 10 and 20 g/m².

Figure 4 gives an overview of the delamination results of the different configurations. In addition to the delamination percentages represented by columns, the requirement of EN 301 for the classification “type 1” in case of special, high density hardwoods, i.e. a maximum value of 8% delamination, is given as a red dotted line.

For the cases of untreated lamination surfaces, the PRF adhesive showed the best performance; however, it clearly still did not fulfil the requirements of EN 301 for adhesive type 1 – a maximum delamination of 8% is required here for specific hardwood species. PU 1 showed generally quite high delamination values up to more than 40% with slightly smaller values for sanded surfaces (over 30%). The delamination results for the adhesive “PU 2” without treatment were extremely high (over 70%). However, surface treatment with a water-based primer solution improved the results for PU 2 significantly, i.e. the delamination percentages could be reduced by factors between ca. 2.5 and 6. Irrespective of application amount and concentration, all delamination values were well below 20%. The best results – in this case even fulfilling the “type 1”-requirements (delamination percentage smaller than 8%) – were observed for the configuration with a relative application amount of 10 g/m² and a concentration of 10%.

Total delamination (%)

Fig. 4. Results of delamination tests for different adhesives and different surface treatments

Glued Laminated Timber

As a basis for experimental verification of strength modelling results, bending tests of full-scale glued-laminated Robinia timber beams were performed, as shown in Figs. 5 and 6. The beams consisted of 20 laminations resulting in a cross-sectional height H of 360 mm, a width W of 110 mm and a length L of 6700 mm. Due to limited Robinia material, the number of test specimens was limited to 4 specimens in total. The laminations were sampled from the same batch as the boards and finger-joint specimens and were finger jointed in the same industrial production process as described above. The four point bending tests were carried out according to the provisions of EN 408 by means of a hydraulic testing machine, whereby the span was 6480 mm (= 18 H), a distance between the two pistons of 2160 mm (=6 H) and a distance between each piston to the nearest support, i.e. the lever arm a = 2160 mm (= 6 H). The global and local deflections were measured and recorded during the tests. From the ultimate load, the bending strength and from the load-deflection curves the local and global MOE were determined. Table 3 gives the individual results and the statistical parameters of all four tests. All beams broke at the bending-tensile edge within the zone of maximal bending moment. The initial fracture was always located at a finger joint of the outer lamination. The COV of the bending strength values COV_beam = 10% was much smaller compared to the variations of the tensile strength of the laminations COV_lam = 39% and the variation of the finger joint tensile strength COV_fj = 35%.

Fig. 5. Test set-up of the full-scale bending tests with Robinia glulam (Schenk 8piston/200kN, Darmstadt, Germany)

Fig. 6. Typical fracture pattern of a Robinia glulam beam after failure

Table 3. Results of the Bending Tests of Full-scale Robinia Glulam Beams

FEM-Model and Monte Carlo Simulations

The computational modelling of the glulam beams – especially with regard to the calculation of bending strength and bending E-modulus – was implemented using a finite element model with an “X-FEM” approach to represent the full non-linear and damage behaviour. This advanced FE method can on the one hand handle yielding phenomena on the compression side and on the other hand accurately represent the softening damage behaviour on the tension side accompanied with localization of damage and load redistribution to adjacent laminations. The FE calculations were carried out using the ABAQUS program package, which is particularly suitable for nonlinear calculations. The model was fully parameterized so that Monte Carlo simulations could be carried out by means of targeted stochastic variation of the input variables, from which the distribution of the glulam strength and stiffness distributions, and finally the characteristic 5% quantile values were to be derived. The computational model also had the potential to predict the expected size-scale effect (i.e., the dependence of bending strength on beam height) and the effects of a possible combination of different strength classes in the glulam cross-section. The model had originally been developed for the simulation of bending properties of glulam made from oak laminations (Tapia and Aicher 2018; Tapia 2022) and was adjusted to the properties of Robinia timber.

Input Data and Simulations

For the stochastic FE-modelling, the distributions of strength and stiffness properties of the laminations as well as of the finger joint strengths are needed as input data. Furthermore, a strength model based on the local correlation between strength and stiffness and a damage/failure model are needed. The different steps for the lamination strength model have been described in detail (Tapia and Aicher 2019; Tapia and Aicher 2021; Tapia and Aicher 2022) for the example of oak boards. Due to a relatively similar defect structure of the species oak and Robinia, some of the more elaborate investigations on the correlation functions, the typical length of weak sections, the fracture energy, and the ratio between tensile and compressive strength were transferred from data collected on oak boards. For calibration to the wood species Robinia, the following data sets of the matched laminations sets were used:

Empirical tensile strength distribution and distribution of tensile E-Modulus of the laminations without finger joints (strength grading model 1),
Empirical tensile strength distribution of industrially produced finger joints,
Empirical data of distribution of distances between two finger joints within the laminations.

In order to estimate the bending strength and bending MOE distributions, a set of 1000 numerically simulated beams was performed.

Comparison of Empirical and Modelling Results

In Figs. 7 and 8, the simulation results of the Monte Carlo simulations of the FE-model are compared to the empirical data of four bending tests with glulam beams. In order to visualize the pronounced homogenisation effects, selected data of the laminations are also additionally plotted in the figures.

Figure 7 shows the bending MOE of glued laminated timber for the four empirically tested beams (filled blue symbols) and for the numerically simulated glulam FE-models (open blue symbols). Additionally, the empirically determined tensile MOEs of the laminations are plotted in the figure as filled green symbols. It can be clearly seen that the simulations gave a very good estimation of the mean glulam MOE and also of the expected variation. It is striking to see the extreme reduction of spread when the tensile MOE of laminations is compared to the calculated as well as the empirically measured glulam MOE Whereas the mean values of lamination MOE (E_{mean, lam} = 14200 N/mm²), simulated MOE of glulam (E_{mean, GL, simu} = 14340 N/mm²), and empirical bending MOE of glulam (E_{mean, GL, emp} = 14500) were close together, the COVs of the lamination MOE (13.7%) was by a factor of ca. 4 to 5 larger than the COV of the simulated glulam MOE (2.7%) and the COV of the empirically derived COV (3.7%).

Fig. 7. Comparison of E-Modulus results: Empirical data of laminations and glulam and simulation results for glulam

Figure 8 shows the bending strength data of glued laminated timber of the four empirically tested beam (filled blue symbols) compared to the results of the numerically simulated glulam FE-model calculations (open blue symbols). Additionally, the empirically determined tensile strength distributions of the Robinia boards and of the industrially produced finger joints are plotted as filled green symbols (boards) and filled orange symbols (finger joints). Obviously, the glulam simulation results showed a good agreement with the test data in a qualitative sense: Compared to the board strength values the variation was reduced to an extreme extent and the strength was shifted to lower values due to the influence of the relatively weak finger joints. Compared to the finger joint strength distribution, the glulam bending strength – for the simulation as well as for the test results – showed pronouncedly higher values resulting from a strong homogenisation effect. However, looking at the quantitative comparison between simulations and empirical glulam results in more detail, it is noticeable that the numerical results exhibited lower strength prediction on the mean and the lower 5%-quantile level. Only the upper branch of the distributions fit well with the maximum value of the empirical data. The reason for the underestimation of the glulam strength could for instance be due to the partial model calibration with oak instead of (not available) Robinia data. But it must also be considered that the empirical data set of only 4 beams – even if these 4 beams consisted of in total more than 500 different pieces of Robinia boards – was very limited. Thus, for the conclusions, the conservative results of the simulations – which were even lower than the full-scale glulam tests – shall be used.

Fig. 8. Comparison of strength results: Empirical data of tensile strength of laminations and finger joints and of bending strength of glulam and simulation results for bending strength of glulam

Discussion of the Glue-Laminated Timber of Robinia

The surface bonding of Robinia posed a great challenge in the project. Although good shear strengths were achieved with several adhesives, achieving a durable bond-line integrity that meets the specifications of the relevant standards – being mandatory for structural use – was a major hurdle. Even with a PRF adhesive, which is successfully used for the production of non-load-bearing Robinia products, a sufficient quality of the bond-lines was not achieved. By far the best results were achieved with adhesive tests on the scale of laboratory specimens by application of a PU adhesive combined with prior water-solution-based primer treatment. Despite the lack of optimisation of the bonding process, delamination values were achieved matching the target values of the relevant standards for a specific parameter set. Due to the boundary conditions of the industrial partner, however, it was not possible to apply the PU/primer technology on the industrial scale, because the application technology for use of one-component PU adhesives differs considerably from the technology for two-component polycondensation adhesives such as MUF or PRF. Here it can be assumed that further investigations and an improvement of the gluing process can lead to safe bond lines also under industrial production conditions.

In the production of finger joints, good values were achieved right away in terms of both the strengths and the durability with PRF on a pilot plant scale. However, implementation on an industrial scale still proved to be difficult, as the results obtained were very strongly dependent on the special production conditions. Further optimisation, especially with regard to cramping pressure and times as well as optimal adhesive application, is necessary here. These developments promise a high potential to achieve such high bonding properties. However, it should be kept in mind that today the industrial production of PRF bonding is being implemented less and less frequently. For this reason, further research effort should go into investigating alternatives to PRF adhesives in the production of finger joints and glulam production from Robinia. More comparative studies on adhesives, particularly those suited for structural applications under harsh environmental conditions, are necessary here. First steps towards the judgement of the durability under extremely varying moisture and temperature changes were done by the use of the delamination test method usually seen as the state-of-the-art procedure to predict long term behaviour of bond-lines. Further research should also focus on use of the latest development of novel hardwood adhesives (e.g. Holeček et al. 2023; Kariž et al. 2024).

During the implementation of the finger joint production from the laboratory tests into industry-related tests, two main parameters to be optimised were identified: The clamping pressure has to be chosen optimally to avoid splitting and on the other hand ensure sufficiently small finger tip dimensions. Furthermore, the wear of cutters has to be monitored carefully, as reduced milling quality significantly influenced finger joint bond strength.

Due to its typical growth characteristics, Robinia wood shows an extreme variance in the mechanical properties, with variation coefficients up to about 40%. Although the properties on the mean value level showed quite high values, the characteristic values were limited. The results of the presented study confirm that an application of sawn Robinia wood as a building product would lead to very low design values without further bonding. The application of newly developed visual grading models based on the German grading standard DIN 4074-5, but with deviating threshold values, resulted in increased characteristic levels, however, with the reduced yield or reduced available lengths of the timber members.

The production of glulam from Robinia lamellae showed a very pronounced lamination and homogenisation effect being considerably stronger than for glulam made from usual softwood species such as spruce. In spite of non-optimised finger joints, limited bond-line quality and usage of saw falling timber without grading, the variations of strength and stiffness parameters of the produced glulam were extremely reduced and comparably high characteristic values were achieved. Thus, Robinia glulam has a promising potential as a high-quality product for the building industry. The lamination and homogenisation effects can be modelled well with the stochastic Finite Element simulation tools adapted to the properties of Robinia wood in the project. A reliable prediction of the glulam properties derived from the properties of the components laminations and finger joints can thus be implemented even with limited experimental testing. This also opens up the possibility of accurately predicting the properties of bonded Robinia laminated timber from other Robinia wood qualities than those used in the project. In addition, the model also offers the possibility of predicting the effects of further optimisation of the production process.

CONCLUSIONS

In the project, it was possible to produce glulam from Robinia wood with comparatively high strengths and a high modulus of elasticity.
The adapted visual strength grading of the Robinia wood used enabled high mechanical properties to be achieved with a high material yield at the same time.
Finger joints in Robinia wood lamellas were successfully produced using phenol-resorcinol-formaldehyde (PRF) adhesives. The normative specifications were also met in terms of durability. The quality of the finger joints depended heavily on the manufacturing conditions, such as the degree of wear of the milling tools used and the pressing pressure. When using PU and MUF adhesives, it was not possible to fully meet the requirements for delamination resistance in particular.
For the surface bonding of Robinia wood lamellas, the sufficient durability requirements could only be met when using a 1C PU adhesive system in combination with a primer pre-treatment. As Robinia is generally considered to be one of the most difficult wood species to bond, the combination of primer pre-treatment and 1C PU adhesives could also be considered promising for use with other hardwood species that are problematic to bond.
Using a combined X-FEM and Monte Carlo simulation, the properties of glued laminated timber made of Robinia could be predicted from the property distributions of the used Robinia wood and the finger joints. The developed methods can generally be used for modelling of the mechanical properties of upcoming other novel load-bearing hardwood timber products.
Further research is needed especially with respect to finger-joint bonding: Either surface pre-treatment methods should be found that can be applied to the finger-jointed flanks or newly developed (preferably bio-based) hardwood-adhesives should be tested for this particular application.
Due to the unpredictable availability of individual softwood and hardwood species in the future, generalizable testing and assessment methods should be developed and validated in the future to enable varying and possibly even mixed use in load-bearing timber construction products.

ACKNOWLEDGEMENTS

The underlying project was carried out in cooperation between STRAB Ingenieurholzbau Hermsdorf GmbH, WMS Werkzeugmaschinen-Service GmbH, the Materials Testing Institute of the Technical University of Stuttgart, the Fraunhofer Institute for Material and Beam Technology IWS, EBF Innovation GmbH and Eurobinia. The authors would like to thank all partners and the funding organisations. This project was supported by the Federal Ministry for Economic Affairs and Climate Action (BMWK) on the basis of a decision by the German Bundestag.

REFERENCES CITED

Bernasconi, A. (2004). “Verleimung von Laubholz für den tragenden Einsatz,” Schweizer Zeitschrift für Forstwesen 12, 533-539. DOI: 10.3188/szf.2004.0533

Berthold, D. (2022). “Brettschichtholz aus Robinie-Hartholz für mehr Klimaschutz in der Bauindustrie und Windkraftindustrie sowie klimaresistente Zukunftswälder,” Fraunhofer-Institut für Holzforschung Wilhelm-Klauditz-Institut WKI. [Online]. Available: https://www.wki.fraunhofer.de/de/forschungsprojekte/2022/ ROBINIA_Brettschichtholz-aus-Robinie.html. [Accessed on 19 January 2024].

Borysiuk, P., Jablosnki, M., Policinska-Serwa, A., and Ruzinska, E. (2011). “Mechanical properties of glue bonds in black locust wood treated with ammonia,” Annals of Warsaw University of Life Sciences – SGGW; Forestry and Wood Technology 73, 162-166.

DIN 4074-5 (2008). “Sortierung von Holz nach der Tragfähigkeit – Teil 5: Laubschnittholz”

EN 301 (2023). “Adhesives, phenolic and aminoplastic, for load-bearing timber structures – Classification and performance requirements,” European Committee for Standardization, Brussels, Belgium.

EN 302-2 (2023). “Adhesives for load-bearing timber structures – Test methods – Part 2: Determination of resistance to delamination,” European Committee for Standardization, Brussels, Belgium.

EN 350 (2016). “Durability of wood and wood-based products – Testing and classification of the durability to biological agents of wood and wood-based materials,” European Committee for Standardization, Brussels, Belgium.

EN 384 (2016). “Structural timber – Determination of characteristic values of mechanical properties and density,” European Committee for Standardization, Brussels, Belgium.

EN 408 (2012). “Timber structures – Structural timber and glued laminated timber – Determination of some physical and mechanical properties,” European Committee for Standardization, Brussels, Belgium.

EN 460 (2023). “Durability of wood and wood-based products – Natural durability of solid wood – Guide to the durability requirements for wood to be used in hazard classes,” European Committee for Standardization, Brussels, Belgium.

EN 13183-1 (2002). “Moisture content of a piece of sawn timber – Part 1: Determination by oven dry method,” European Committee for Standardization, Brussels, Belgium.

EN 14080 (2013). “Timber structures – Glued laminated timber and glued solid timber – Requirements,” European Committee for Standardization, Brussels, Belgium.

EN 14358 (2016). “Timber structures – Calculation and verification of characteristic values,” European Committee for Standardization, Brussels, Belgium.

EN 15425 (2023). “Adhesives – One component polyurethane (PUR) for load-bearing timber structures – Classification and performance requirements,” European Committee for Standardization, Brussels, Belgium.

EN 16254 (2023). “Adhesives – Emulsion polymerized isocyanate (EPI) for load-bearing timber structures – 5Classification and performance requirements,” European Committee for Standardization, Brussels, Belgium.

Eurobinia, Keilgezinkte Robinie? Das geht. Wir haben’s erfunden., [Online]. Available: http://www.eurobinia.de/index.php?p=about_keilzinken. [Accessed on 12 Septembre 2024].

Frühwald, A., Ressel, J. B., Becker, P., Pohlmann C. M., and Wonnemann, R. (2003). “Verwendung von Laubhölzern zur Herstellung von Leimholzelementen – Abschlussbericht,” Hamburg: Universität Hamburg – Zentrum Holzwirtschaft.

Göhre, K. (1952). Die Robinie und ihr Holz, Deutscher Bauernverlag, Berlin.

Grosser, D. (1998). Einheimische Nutzhölzer (Lose Blattsammlung) – Vorkommen, Baum- und Stammformen, Holzbeschreibung, Eigenschaften, Verwendung, Holzabsatzfonds, Bonn.

Holeček, T., Sikora, A., Šedivka, P., Cvejn, D., Sládek, D., Bárta, J., and Lagaňa, R. (2023). Novel hybrid polymer adhesives for laminated materials based on hardwood,” Composite Structures 308, article 116684. DOI: 10.1016/j.compstruct.2023.116684

Kamperidou, V., and Barboutis, I. (2017). “Bondability of black locust (Robinia pseudoacacia) and beech wood (Fagus sylvatica) with polyvinylacetate and polyurethane adhesives,” Maderas, Ciencia y tecnología 19(1), 87-94.

Kariž, M., Šega, B., Šernek, M., Žigon, J., and Merela, M. (2024). “Bonding properties of selected alien invasive wood species,” BioResources 19(2), 3078-3094. DOI: 10.15376/biores.19.2.3078-3094

Koch, G., and Dünisch, O. (2008). Juvenile wood in Robinie – Qualität von Robinienholz (Robinia pseudoacacia L.) und Folgerungen für die Holzbearbeitung, Fraunhofer IRB Verlag, Stuttgart, Germany.

Konnerth, J., Kluge, M., Schweizer, G., Miljovic, M., and Gindl-Altmutter, W. (2016). “Survey of selected adhesive bonding properties of nine European softwood and hardwood species,” European Journal of Wood and Wood Products 74, 809-819. DOI: 10.1007/s00107-016-1087-1

Lehringer, C. (2014). “Alles außer Fichte; Neuste Entwicklungen bei der Verklebung von alternativen Holzarten mit 1K-Purbond-Klebstoffen,” in: 2. Kooperationsforum Kleben von Holz und Holzwerkstoffen, Würzburg.

Müller, G. (2006). Untersuchungen zur Optimierung der Keilzinkenverklebung von Robinie, Diplomarbeit, Eberswalde, Germany.

Pitzner, B., Bernasconi, A., and Frühwald, A. (2001). Verklebung einheimischer dauerhafter Holzarten zur Sicherung von Marktbereichen im Außenbau, Hamburg.

Schickhofer, G., and Hasewend, B. (1999). “FFF Projekt: Entwicklung widerstandsfähiger Holzbauprodukte mit dem Hartholz Robinie Graz”.

Schickhofer, G., and Obermayr, B. (1999). “Development of high-resistance glued robinia products and an attempt assign such products to the European system of strength classes,” in: International Council for Research and Innovation in Building and Construction; Working commission W18 – Timber Structures, Graz, Austria.

Tapia Camú, C. (2022). Variation of Mechanical Properties in Oak Boards and its Effect on Glued Laminated Timber, Ph.D. Dissertation, University Stuttgart, Germany.

Tapia, C., and Aicher, S. (2018). “A stochastic finite element model for glulam beams of hardwoods,” in: World Conf. Timber Eng., Seoul, Republic of Korea.

Tapia, C., and Aicher, S. (2020). “Variation and serial correlation of modulus of elasticity between and within European oak boards (Quercus robur and Quercus petraea),” Holzforschung 74(1), 33-46. DOI: 10.1515/hf-2019-0038

Tapia, C., and Aicher, S. (2021). “Simulation of the localized modulus of elasticity of hardwood boards by means of an autoregressive model,” ASCE J. Mater. Civ. Eng. 33(6), article 04021132. DOI: 10.1061/(ASCE)MT.1943-5533.0003696

Tapia, C., and Aicher, S. (2022). “Survival analysis of tensile strength variation and simulated length-size effect along oak boards,” ASCE J. Eng. Mech. 148(1), article 04021130. DOI: 10.1061/(ASCE)EM.1943-7889.0002006

Varga, D., and van der Zee, M. (2008). “Influence of steaming on selected wood properties on four hardwood species,” Holz als Roh- und Werkstoff 66, 11-18. DOI: 10.1007/s00107-007-0205-5

Voulgaridis, E., Passialis, C., Negri, M., and Adamopoulos, S. (2012). “Shear bond strength of black locust wood glued with adhesive systems,” Wood Research 57(3), 489-496.

Article submitted: October 2, 2024; Peer review completed: November 23, 2024; Revised version received: January 15, 2025; Published: April 4, 2025.

DOI: 10.15376/biores.20.2.3848-3865