1 Introduction

Timber and timber-based composites are gaining popularity as environmentally friendly and cost-effective sustainable construction materials. Modern timber buildings often use hybrid and composite structural systems made from mass engineered timber products such as glue-laminated timber (Glulam), cross-laminated timber (CLT) and laminated veneer lumber (LVL) for their superior mechanical properties. Sawn timber produced from plantation forests have size restrictions, and hence, engineered timber products (ETP) are becoming popular for large scale building construction (Abed et al. 2022). Advancements in adhesive technologies enabled innovation in structural systems made from engineered timber whose performance could be significantly impacted by the fracture characteristics of both sawn timber and glued timber elements.

Three modes of fracture typically refer to the relationship between the crack tip stresses and the direction of crack propagation. Like other materials, timber fracture characteristics can be broadly classified as Mode I (tension is applied orthogonal to the crack surface), Mode-II (stress is applied parallel to the crack surface but orthogonal to the crack front), and Mode III (stress is applied parallel to the crack surface as well as to the crack front). It is worth noting that Mode-I and Mode-II represent failures typically observed in ETPs (Anderson 2017). Strain energy release rate in Mode-II fracture, which is also known as forward shear or sliding mode, is caused by a load applied perpendicular to the fibre direction of timber (Mohamadzadeh et al. 2018). Mode-II fracture condition in timber and ETPs are very common and are frequently observed in beams, longitudinal lap joints, holes, connections, and notches where load is applied perpendicular to the member. Sawn timber and timber composites are highly vulnerable to microcracks caused by moisture cycle, drying shrinkage, and natural defects where crack may propagate before reaching the design stress (Mohamadzadeh et al. 2018; Veigel et al. 2012). Despite the obvious significance of fracture characteristics, absence of any specific testing standard for timber has resulted in inconsistent fracture parameters obtained using ad-hoc testing techniques (Mohamadzadeh et al. 2018). The energy required for crack propagation is determined by the orientation of the fracture plane as well as the direction of load application (Kanninen et al. 1986). Timber is an orthotropic material consisting of three orthogonal axes such as longitudinal, radial, and tangential (L, R and T), where crack propagation systems are often denoted by two letters; the first letter represents the direction of the load application and the second indicates the direction of crack propagation. Although crack propagation can occur in six different orientations i.e., LR, RL, TL, LT, RT, and TR planes in three orthogonal directions, cracks in timber typically occur in RL and TL directions due to the inherent gaps between fibres along the longitudinal direction (L) (Danielsson 2013). Strain energy release rate and crack initiation load in RL and TL directions are almost identical, and hence, it is a common practice not to differentiate between RL and TL directions in timber fracture. Most of ETPs are manufactured using glue, and typical timber joints and connections are also often strengthened using glue. Therefore, fracture behaviour of timber and timber-adhesive bonds is very important for assessing overall structural response. During the last few decades, various methods have been developed to evaluate fracture energy (strain energy release rate in Mode-II, GII) varying from simple analytical techniques to complex compliance-based beam method (CBBM). Several test configurations for Mode-II fracture have been used for timber, composites, and metal structures including end-notched flexure (ENF), tapered end notched flexure (TENF), compact shear (CS), and four-point end notched flexure (4ENF) (Shivakumar et al. 2015) (Xavier et al. 2011) (Xu et al. 2016) (Mohamadzadeh et al. 2018). ENF has been the most widely used testing method for adhesive bonded composites and metal structures because of its simple test set-up (Xu et al. 2016). Timber is an orthotropic natural fibrous material which has similarities to fibre reinforced composite material, and hence, ASTM-D7905 (ASTM 2019) (Standard for Mode-II fracture of unidirectional fibre reinforced polymer). ENF test method has been considered in the current study.

ENF Mode-II fracture can be challenging due to unstable crack propagation with barely noticeable crack formation (Rzeczkowski 2021; Sebera et al. 2021). Therefore, compliance-based beam method (CBBM), which is based on an equivalent crack method, was proposed instead of using classical crack monitoring techniques. Standard test methods for fracture are based on actual crack length monitoring technique and depend on accurate measurement of crack length as well as on precise definition of the crack tip. Several classical data reduction methods, such as compliance calibration (CC), direct beam theory (DBT) and corrected beam theory (CBT), that are based on actual crack monitoring, have been used to analyse the strain release rate from ENF test. Due to its relative simplicity, CBBM has been used in several studies where equivalent crack can be used to investigate fracture properties.

Digital Image Correlation (DIC) technique is a useful tool that can significantly contribute to classical crack method to monitor crack tip opening along the crack line (Sutton et al. 2009). Even though DIC has been extensively used in composite materials and wood structures, further study is needed to discover more on crack mouth opening and accurate crack length monitoring for reliable assessment of timber and timber-adhesive bond fracture. The current study utilises DIC measurements to compare classical crack measurement methods and equivalent crack methods to assess their relative advantages and limitations when applied to timber. Most of the recent research investigations were performed with varying geometric configurations without considering the effect of initial crack width on crack propagation and fracture energy (De Moura et al. 2008; De Moura et al. 2006; Xavier et al. 2011). Thin initial notch width may be useful for stable and consistent crack propagation, but an artificial notch created by using a cutting tool in timber-adhesive bond could incorporate microcracks along the fibre gaps. Microcracks may propagate through pre-existing fibre gaps, leading to crack propagation within the timber rather than along the glue line. Therefore, it is important to conduct meticulous experimental study on the effect of initial notch width and notch preparation techniques used for ENF testing.

In the present work, DIC has been extensively used to experimentally evaluate Mode-II fracture energy release rate of timber and timber-adhesive bond. R curves were generated utilising both classical data reduction approaches and CBBM. Initial notch width was varied to achieve pure glue delamination (cohesive or adhesive failure) in timber-adhesive bond specimens. Relative advantages and limitations of two distinct analysis methods i.e., classical crack and equivalent crack method were carefully investigated to provide useful guidelines for future experimental research in the field.

2 Material properties

All experiments conducted on timber specimens were made from Australian softwood species radiata pine (Pinus radiata). The test specimens were cut from defect free sawn timber planks with average density of 580(± 34) kg/m3. The specimens were prepared from three separate plank sizes of 90 × 30 × 2400 mm and were stored in dry environment with a temperature of 20 °C and relative humidity of 65% to ensure achieving an equilibrium moisture content (EMC). The timber planks were taken from the same batch and the same strength grade of machine-graded pine (MGP8), supplied by AKD Softwoods. Material properties of MGP8 timber, as listed in Australian Standard AS 1720 (AS1720 2010), are shown in Table 1. As part of the current study, the moisture content of each specimen was tested before the experiment and the measured average moisture content was 10.8 (± 0.94) %.

Table 1 Material properties of MGP-8 grade timber (AS1720 2010; Ross 2010; Walford 2003)
Full size table

Polyurethane (PUR) based adhesive HB S309 PURBOND (Henkel-PURBOND 2013) was used for timber-adhesive bond specimens. PUR adhesives typically require moisture for curing (polyaddition) and are one component solution that sets and hardens utilising the water in wood structure (Konnerth et al. 2016). PUR provides more ductile and better adhesion solution that is suitable for a wide range of softwood timber, including finger joint, lap joints and load bearing and non-bearing constructions (Künniger et al. 2019; Obersriebnig et al. 2013). Other adhesives, such as Phenol Resorcinol Formaldehyde (PRF), Melamine urea–formaldehyde (MUF), and Melamine–formaldehyde (MF), are more brittle and stiffer than PUR. The current research used PUR as this is the most widely used adhesive in timber lamination in different ETPs such as CLT, glulam etc.

3 Testing geometry and configurations

ENF-Mode-II fracture tests were performed on both timber and timber-adhesive bond specimens using relevant guidelines proposed in ASTM-D7905 (ASTM 2019), which is typically used for unidirectional fibre reinforced polymer. A series of trial tests, in conjunction with nonlinear finite element models, was conducted on various cross-sections (15, 20 and 30 mm squares) with 300, 500 and 610 mm lengths to select the most suitable geometry for ENF testing of timber. Preliminary trial geometric dimensions were selected from previous studies (De Moura et al. 2006); (Majano-Majano et al. 2020; Mohamadzadeh et al. 2018; Sebera et al. 2021; Xavier et al. 2011). Conducted trials showed that 500 mm long specimens with 20 mm square cross-section produced stable Mode II fracture for both timber and timber-adhesive bond specimens. The schematics of test specimens with detailed configurations used for sawn timber and timber-adhesive bond are shown in Figs. 1 and 2, respectively. All timber specimens used for ENF had 20 × 20 mm cross-sections and were 500 mm long offering a span length of 460 mm (De Moura et al. 2006). Initial crack length (a0) was considered to be larger than 0.7L to achieve stable crack propagation (Carlsson et al. 1986), where L is half of the span length of the specimen (= 230 mm). Timber-adhesive bond specimens were produced by gluing two separate timber planks with 20 × 10 mm cross-section. The bond curing time, moisture content and relative humidity requirements were followed in accordance with the glue specifications (Henkel-PURBOND 2013). The current study only focussed on crack propagation along longitudinal (L) direction and the glue line. The width of the initial crack was varied between 0.1 and 3.0 mm to investigate its effect on fracture behaviour. The first set of sawn timber specimens was produced with 1 mm wide initial crack for the whole 182 mm length, whilst for the second set of specimens, the crack width was reduced to 0.5 mm for the first 20 mm (near the crack tip) length followed by 1.0 mm wide crack in the remaining 162 mm. In the case of timber-adhesive specimens, three different crack widths i.e., 3.0 mm, 1.0 mm and 0.1 mm were considered as shown in Fig. 3d, e and Table 2. Test specimens with 3.0- and 1.0-mm initial crack widths along the glue line were cut by using a saw blade, whereas a polyethylene tape was used for the 3rd set of specimens with 0.1 mm crack width along the full crack length of 182 mm. Use of polyethylene tape discarded the use of any mechanical cutting tool and hence kept timber fibres intact. Details of cutting tools, the cutting process and specimen preparation techniques are shown in Fig. 3.

Fig. 1
figure 1

Schematic showing test arrangement for ENF-Mode-II sawn timber fracture test

Full size image
Fig. 2
figure 2

Schematic showing test arrangement for ENF-Mode-II timber-adhesive bond

Full size image
Fig. 3
figure 3

Different initial crack widths used in ENF samples for sawn timber and timber-adhesive bond specimens

Full size image
Table 2 Specimens used for ENF Mode-II tests
Full size table

The ENF Mode-II fracture tests were conducted using an INSTRON 10 KN load frame using displacement control at a rate of 5 mm/min. All fracture tests were recorded with DIC technique to identify and monitor crack length propagation. A uniform spackle pattern using small black dots was used on timber specimens for effective DIC analysis. In the current test, a MERCURYRT (MERCURY-DIC 2017) system was used to monitor crack propagation length and for measuring crack tip opening displacement. The DIC system included five cameras, calibration plate, LED lights, black screen, and processing computer as shown in Fig. 4. The camera was calibrated with calibration plate through a series of 32 images with negligible error of 0.01 pixel and the coordinate system was also setup by the calibration plate. The area of interest (AOI) was set for 150 mm length and 40 mm wide, which was large enough to capture the crack propagation length. The DIC camera was fixed at 16 pixel/mm and 60 image frames per second (fps) to achieve better accuracy in data analysis. DIC system was then verified with the known displacement from the Instron loading head and synchronised with a frequency of 1 Hz. This synchronisation is important to couple the data from two different machines. The testing procedures were meticulously recorded using a DIC camera, and the recorded data were subsequently subjected to post-processing with high precision in specifically designated regions of the samples.

Fig. 4
figure 4

Full test setup with DIC system

Full size image

4 Data reduction theory

Data reduction methods available for Mode-II strain energy release rate can be classified into two distinct groups, such as classical method and equivalent crack method. The basic formula of strain energy release rate is established from Irwin–Kies equation (Kanninen et al. 1986) as shown in Eq. 1

$${G}_{II}=\frac{{P}^{2}}{2B}\frac{dC}{da}$$
(1)

where, P is the applied load, B is the specimen width, C is compliance (C = \(\delta /P\)) to be determined from least square linear regression, and a is the crack length. Several data reduction schemes are adopted to formulate generalised equations to calculate Mode-II strain energy release rate GII (Gliszczynski et al. 2021). ASTM-D7905 (ASTM 2019) suggests using compliance calibration (CC) method based on actual crack monitoring. CC method is simple and mostly used in ENF testing as this technique relies on three easily obtainable parameters such as load, displacement, and crack length (P-\(\delta\)-a). In CC method, Mode-II strain energy release rate is determined using Eq. 2.

$${G}_{II}=\frac{3{ma}^{2}{P}^{2}}{2B}$$
(2)

where, m is the coefficient of CC obtained from the slope of compliance (C) vs cubic of crack length (a3); B is the specimen width.

In direct beam theory (DBT), the crosshead displacement \(\delta\) is directly used in the fracture energy calculation. DBT is also based on actual crack monitoring and can be expressed as follows.

$${G}_{II}=\frac{9{a}^{2}P\delta }{2B(2{L}^{3}+3{a}^{3})}$$
(3)

It is important to mention that the crosshead displacements obtained from the ENF test are usually greater than those obtained from other methods, such as compact shear (CS) test, tapered end notched flexure (TENF) (De Moura et al. 2006). Additionally, DBT does not utilise any data reduction technique implemented in other evaluation methods, and the fracture energy values determined using DBT are typically greater than the other available methods.

DBT was modified as the corrected beam theory (CBT) to include a flexure modulus, which directly depends on initial compliance. CBT (Gliszczynski et al. 2021) is another classical method that utilises actual crack monitoring for calculating GII, as shown in Eq. 4.

$${G}_{II}=\frac{9{a}^{2}{P}^{2}}{2{B}^{2}{h}^{3}{E}_{f}}$$
(4)

where, Ef is the flexural modulus that can be calculated using Eq. 5, and the initial compliance (C0) can be determined from the ratio of displacement and load when a = a0. It can be noted that a0 is the initial crack length as shown in Figs. 1 and 2.

$${E}_{f}=\frac{{L}^{3}}{4B{h}^{3}{C}_{0}}$$
(5)

The compliance-based beam method (CBBM) (De Moura et al. 2008), which is based on Timoshenko beam theory, can predict GII using the load–displacement data (Gliszczynski et al. 2021) without actual crack measurements. CBBM determines \({G}_{II}\) following Eqs. 6–10.

$${G}_{IIc}=\frac{9{P}^{2}{{a}_{eq}}^{2}}{16{B}^{2}{h}^{3}{E}_{f}}$$
(6)
$$C=\frac{3{a}^{3}+2{L}^{3}}{8B{h}^{3}{E}_{f(CBBM)}}+\frac{3L}{10Bh{G}_{LR}}$$
(7)

where, \({E}_{f(CBBM)}\) is the flexural modulus; I = Bh3 /12 is the second moment of area and \({G}_{LR}\) is the shear modulus in longitudinal-radial plane. Initial elastic compliance C0 and initial crack length a0 are used to calculate corrected flexure modulus \({E}_{f(CBBM)}\) using Eq. 8

$${E}_{f(CBBM)}=\frac{3{a}_{0}^{3}+2{L}^{3}}{12I}{\left({C}_{0}-\frac{3L}{10{G}_{LR}Bh}\right)}^{-1}$$
(8)

To get an equivalent crack length at each point, the experimental compliance C should be replaced by a corrected compliance CC. An equivalent crack length aeq and the corrected compliance CC can be determined using Eq. 9:

$${a}_{eq}={\left[\frac{{C}_{C}}{{C}_{0C}}{a}_{0}^{3}+\frac{2}{3}\left(\frac{{C}_{C}}{{C}_{0C}}-1\right){L}^{3}\right]}^{1/3}$$
(9)

where, \({C}_{C}=C-\frac{3L}{5A{G}_{LR}}\) and \({C}_{0C}={C}_{0}-\frac{3L}{5A{G}_{LR}}\)

Once \({a}_{eq}\) is determined, Eq. 9 can be modified as follows.

$${G}_{II}=\frac{9{P}^{2}}{16{B}^{2}{h}^{3}{E}_{f}}{\left[\frac{{C}_{C}}{{C}_{0C}}{a}_{0}^{3}+\frac{2}{3}\left(\frac{{C}_{C}}{{C}_{0C}}-1\right){L}^{3}\right]}^{2/3}$$
(10)

According to Eq. 8, flexure modulus may be calculated from the initial elastic compliance C0 and the initial crack length a0. CBBM depends on the shear modulus in the longitudinal-radial plane \({G}_{LR}\), which may show considerable variations within the same graded samples due to moisture, humidity, natural defects, etc. Previous research showed that GLR (AS1720 2010; Ross 2010) may vary significantly between 350 and 690 MPa based on early and late grown timber, and hence the effect of this variation on fracture response was investigated by conducting a sensitivity study at the onset of the current study. To understand the effect of shear modulus GLR on fracture energy, a theoretical range between 300 and 1000 MPa was considered to produce Fig. 5, in which GII values were calculated using CBBM. Obtained theoretical results clearly show GLR has insignificant effect on GII obtained using CBBM.

Fig. 5
figure 5

Influence of Shear modulus (GLR) on strain energy release rate GII obtained from CBBM (theoretical demonstration)

Full size image

5 Application of DIC in Mode-II fracture test of timber

Since there is no fracture testing standard for timber, it is important to develop specific testing criteria which include, but are not limited to, crack mouth opening, crack length monitoring and width of initial notch for accurate measurement of fracture resistance of timber structure.

5.1 Interaction of Mode I and Mode II in crack development and propagation

DIC technique was used in the current study as the crack propagation in Mode-II often does not show any significant gap at the cracking interface (Rzeczkowski 2021); a small amount of crack tip opening typically occurs due to displacement in normal direction under Mode-I fracture condition. To identify the crack initiation, propagation and to measure the crack length, two different probes such as a line probe and an area strain probe were used. Line probes act as digital extensometer and are very effective to investigate crack tip displacement and crack length monitoring, while the area strain probes are effective to visualise crack initiation and crack propagation. Digital line probes were placed along the glue line of timber-adhesive bond specimens and along the initial crack line of sawn timber specimens as shown in Fig. 6. Four line probes were placed at equal distance along the initial 69 mm of crack length. A schematic of the line probe locations is shown in Fig. 6a. Each line probe consists of two-point probes, i.e., \({\mathrm{a}}_{\mathrm{n}}\) and \({\mathrm{a}}_{\mathrm{n}}^{\mathrm{^{\prime}}}\) to record relative displacements in both X and Y directions. The location of those line probes in a real specimen are shown in Fig. 6b. Horizontal and vertical displacements of each line probe were determined from the change in coordinates of the point probes. During crack propagation, two parts would slide against each other in the horizontal direction as shown in Fig. 6c and d. All horizontal and vertical movements of line probes were recorded by DIC cameras. The horizontal movement, as recorded from tests using DIC, of all point probes in four different locations for one sawn timber and one timber-adhesive bond specimen is shown in Fig. 7.

Fig. 6
figure 6

DIC techniques used to identify crack initiation and propagation

Full size image
Fig. 7
figure 7

Horizontal slip of timber and timber adhesive bond samples in four separate locations

Full size image

Horizontal slips (X-slip) and vertical Y-displacements were calculated from the horizontal and vertical movements of the probes as shown in Table 3.

Table 3 X-slip and Y-displacement at 4 locations
Full size table

In ENF testing, the crack tip primarily exhibits horizontal displacement in Mode II (∆II), while a minor portion of vertical displacement (∆I) is also observed due to the contribution of Mode I fracture towards crack tip opening. This type of interaction between fracture Modes is not uncommon but greater ratio of interaction between modes i.e., Mode (\({\Delta }_{I}\)/\({\Delta }_{II}\)) =\(\gamma\) can result in unstable crack propagation; this inevitable interaction would require careful investigation for timber specimens (Rikards et al. 1996). Figure 8 shows the comparison of Mode-II (horizontal X direction) and Mode-I (vertical Y direction) crack mouth displacements for timber and timber-adhesive bond specimens as determined from the initial and final coordinates of the selected line probes. The average value of \(\gamma\), i.e., Mode (\({\Delta }_{I}\)/\({\Delta }_{II}\)) for sawn timber at locations A1, A2, A3 and A4 was 0.12, 0.05, 0.03 and 0.02, respectively; whilst those for timber-adhesive bond were 0.14, 0.02, 0.01 and 0.003. A1 was located at the crack initiation point, and therefore, Mode–I vertical opening was considerably greater than other locations resulting in substantially high \(\gamma\) value at the point. However, at all other locations, Mode I displacements were insignificant and clearly showed that ENF testing method used herein can be used as a reliable technique for generating stable fracture for both timber and timber-adhesive bonds.

Fig. 8
figure 8

Mode-I and Mode-II displacements recorded at four different points in timber and timber-adhesive bond specimens

Full size image

5.2 Initiation of crack and monitoring of crack length using DIC

After confirming the prevalence of horizontal slip in crack formation during ENF testing, the present study expanded the use of DIC to accurately monitor and measure the actual crack length in both sawn timber and timber-adhesive bonds. Digital line probes were placed at every millimetre length along the glue line of timber-adhesive bond samples and along the initial crack line of sawn timber specimens, as schematically shown in Fig. 9a. As the line probes were placed at 1 mm interval, they also represented the actual location of the probe from the crack tip, i.e., A4 represented crack length at 4 mm away from the tip of the initial crack.

Fig. 9
figure 9

Schematic showing determination of crack length in ENF specimens

Full size image

Once the loading was applied, the relative horizontal displacements occurring at point probes made line probes diagonal as shown in Fig. 9b. The magnitudes of horizontal slip resulting in crack initiation were carefully assessed by analysing line probe data to determine a specific displacement criterion that can be used to identify crack initiation. As part of the current study, four specific values for horizontal displacements (Xcr) of 0.10, 0.15, 0.25 and 0.35 mm were considered as the critical threshold displacement (Xcr) to identify crack initiation. When the horizontal extension of a particular line probe exceeded the considered crack initiation slip, that location was considered as cracked. This process was continued to capture the progressive crack growth from 1 to 69 mm, and the corresponding load as well as the crosshead displacements in vertical direction were also recorded.

Bond specimen number 10 is shown as an example of a representative ENF sample to demonstrate fracture detection and crack length monitoring. Figure 10a shows horizontal extension from several probes showing a rapid and significant rise in horizontal slip at the tip of the crack, however the rate of increase in horizontal displacement attained a constant upward trend after approximately 765 s. The image of a specimen and the relevant strain field at 765 s are shown in Fig. 10b. After analysing the DIC image and the area strain contour, it was observed that the horizonal slip varied between 0.10 and 0.35 mm within 765 s that was considered for analysing crack propagation.

Fig. 10
figure 10

Extension of line probes in X direction from timber adhesive bond-10

Full size image

To establish a suitable value for Xcr, a sensitivity study was performed based on 0.1, 0.15, 0.25 and 0.35 mm of horizontal displacements captured from line probes. Figure 11a shows corresponding Xcr values in the load–displacement curve of an ENF test sample. At first, total crack lengths were measured using 0.1, 0.15, 0.25- and 0.35-mm horizontal extension as the threshold for crack initiation. Load vs crack length was plotted for all considered Xcr values, as shown in Fig. 9b. The line representing crack length formation with Xcr = 0.1 mm, originated at 920 N load, and continued until the crack developed up to 34 mm, but the load suddenly dropped to around 583 N as the crack length went beyond that point. When the crack initiates, it is usual for the load to experience a sudden fall. In the current experimental study, cracks were noticeable when the applied load reached 600 ± 30 N, as shown in Fig. 11a. Additionally, all Xcr values were used to form R curves which are shown in Fig. 11c. The analysis indicates that adopting Xcr = 0.10 mm led to a higher prediction of fracture energy, owing to the unusually higher load values recorded for crack initiation. Hence, it was concluded that 0.10 mm horizontal extension was not a suitable threshold for timber. Among the other considered Xcr values, 0.35 mm also produced some overprediction of loading, but the other two Xcr values, 0.15 mm and 0.25 mm, both produced reliable load and fracture energy along the crack length. As all specimens used in the present study had the same cross-sectional dimensions and were fabricated using identical timber species and adhesive, a horizontal slip of 0.15 mm was chosen as the threshold for detecting crack initiation in both sawn timber and timber-adhesive specimens.

Fig. 11
figure 11

Crack sensitivity analysis using different horizontal slip thresholds

Full size image

6 Effect of initial crack width on sawn timber and timber-adhesive fracture

Width of the initial notch used in ENF testing can have a significant effect on observed fracture behaviour. Sawn timber and timber-adhesive bond specimens were tested with different initial crack widths to determine its effect on fracture characteristics. Figure 12a and b show load–displacement curves obtained from Mode-II ENF testing with varying notch widths. Figure 12a illustrates that applying varying notch widths (NW), such as 0.5 mm and 1 mm introduced using saw blades, did not significantly impact the sawn timber fracture, except for a slight reduction in the initial peak load. However, relatively complex interactions between timber and glue lines were observed in timber-adhesive bond specimens when NW was varied. Figure 12b shows that the load–displacement curves for timber-adhesive bond specimens with 1 mm and 3 mm NW were almost identical, but specimens with 0.1 mm NW (shown in Fig. 12c) produced significantly greater fracture resistance, strength and load carrying capacity. It is worth noting that both 1- and 3-mm notches were cut using saw blades, but 0.1 mm notch was introduced by using a polyethylene tape that prohibited bonding between timber sections in the specified region.

Fig. 12
figure 12

Load vs displacement curves obtained from Mode-II ENF tests conducted on sawn timber and timber-adhesive bond specimens with varying notch width

Full size image

Figure 13 shows timber and timber-adhesive bond specimens with crack propagation after ENF testing. Sawn timber specimens with 1 mm and 0.5 mm notch are shown in Fig. 13a and c, in which cracks propagated along the notch line. Crack mouth opening in normal direction (Mode-I) was greater in 1 mm NW sample than that in 0.50 mm NW. Use of 0.5 mm notch for sawn timber seemed to produce small hairline crack tip opening in the vertical direction and the fracture was predominantly caused by horizontal slip (Mode-II), which is the preferred mode in ENF testing. Figure 13b–e show timber-adhesive bond specimens after ENF tests with varying notch widths. Figure 13b and d indicate that the introduction of a 3 mm and 1 mm notch triggered cracks right above the glue line as the mechanical cutting process weakened the nearby timber fibres. However, in Fig. 13e, in which 0.1 mm notch was created by placing a polyethylene tape between the timber pieces, a very thin delamination crack initiated and propagated along the glue line producing an ideal fracture response for timber-adhesive bond.

Fig. 13
figure 13

Crack propagations in sawn timber and timber-adhesive bond samples under ENF testing with varying notch widths

Full size image

ENF test specimens were cut open along the fracture surface to conduct further examination of the fracture surface using a microscope as shown in Fig. 14. Remarkable differences were observed between the timber fracture surface and the timber-adhesive bond fracture surface. Timber failure occurred along the longitudinal fibre in specimens where mechanical cutting was used to create the initial notch. However, in bonded timber specimens with 0.1 mm NW (polythene tape), pure delamination of adhesive with no timber fracture was observed. Figure 15 compares microscopic images of fracture surfaces of specimens with notch cut by saw blade against those prepared using a plastic tape. In Fig. 15a, highlighted area shows damage in timber fibre caused by the cutting tool, which eventually affected fracture behaviour in ENF testing. In Fig. 15b, on the contrary, pure delamination occurred at the bond line when the initial notch was created using a plastic tape.

Fig. 14
figure 14

Microscopic image of timber and timber adhesive bond fracture surface

Full size image
Fig. 15
figure 15

Microscopic image of crack initiation surfaces of specimens with notch induced by saw blade vs plastic tape

Full size image

Introducing a notch using a cutting tool in a bond fracture test does not seem to be an ideal testing procedure based on test observations. Stress concentration can cause premature failure at the saw-cut notch, which can also cause damage to the surrounding wood fibre allowing fracture to propagate along the existing longitudinal timber fibre gaps. Timber is a bio- composite material, and hence any induced crack may coincide with existing fibre gaps and/or natural defects resulting in reduced resistance to crack growth. This phenomenon enhanced the likelihood of cracks initiated in timber rather than along the glue line. Introducing a notch by cutting timber along the bond line was challenging, and the width of the notch showed significant effect on bond strength, stiffness, and total load bearing capability. Therefore, use of thin plastic tape is recommended to conduct Mode-II ENF testing in glued timber.

7 Load–displacement behaviour of timber and glued timber under Mode II fracture

Once the specific testing criteria were decided based on sensitivity analysis presented in Sect. 5 and 6, load–displacement response of considered timber samples were carefully analysed to characterise fracture behaviour. Load and displacement values were extracted from the Instron load frame where displacement (\(\delta\)) was measured at the load application point.

Figure 16 illustrates the load–displacement curves obtained for 10 specimens of sawn timber with 0.5 mm notch and 8 specimens of timber-adhesive bond with 0.1 mm notch, respectively. It is worth noting that two specimens representing timber-adhesive bond i.e., Bond-1 and Bond-2 were excluded from the analysis as pre-existing timber defects (knots) were located adjacent to the crack line that significantly affected experimental results. Although the overall load–displacement curves for both timber and timber-adhesive bond specimens look similar, glued timber fracture was quite different from timber fracture giving greater strength, stiffness and load carrying capacity. In glued specimens, fracture load required for crack initiation was 42% greater, on average, than sawn timber fracture. The average load head displacement for timber specimens was 10.37 ± 1.2 mm, whilst that for glued timber specimens was 12.67 ± 1.4 mm showing more than 22% increase in vertical displacement before fracture initiated. Timber showed a tendency to develop slow and gradual crack propagation, whereas crack propagated relatively rapidly in glued timber specimens.

Fig. 16
figure 16

Load vs displacement curves obtained from Mode-II ENF test of sawn timber and timber adhesive bond fracture

Full size image

In Mode-II fracture, crack formation followed a linear propagation; the fracture process for both sawn timber and timber-adhesive bond are shown in Fig. 17. DIC was used to measure the crack length, and area strain field was created to visualize the propagation of cracks.

Fig. 17
figure 17

Use of DIC in tracking crack development in timber and timber-adhesive bond

Full size image

Data obtained from digital extensometers (line probes) and the relevant area strain fields for all specimens were thoroughly analysed to record load–displacement response as well as the corresponding crack length. Strain energy release rate GII primarily depends on the accurate and simultaneous measurement of crack length, the corresponding load and crack mouth opening. Figure 18 shows the relationship between these three key parameters for Mode-II fracture obtained from ENF tests conducted on sawn timber (sample number-2) and timber-adhesive bond specimens (sample number-7).

Fig. 18
figure 18

Load vs displacement and crack length measured from Mode-II ENF tests of timber and timber-adhesive bond specimens

Full size image

Obtained results clearly show that the elastic part of the load–displacement curve is linear and has a constant initial slope prior to fracture. A sudden drop in load was observed as soon as fracture initiated in both timber and glued timber specimens. The post-elastic part of the load–displacement response is highly nonlinear and does not follow the principles of linear elastic fracture mechanics (LEFM). This phenomenon clearly shows that to accurately simulate Mode-II fracture behaviour of solid timber and glued timber, advanced nonlinear fracture mechanics-based FEM analysis techniques, i.e., cohesive zone model (CZM) or continuum damage mechanics (CDM) would be required.

8 Strain energy release rates for timber and glued timber under Mode II fracture

Load vs displacement and crack length curves, as shown in Fig. 18, were used to determine fracture energy for all tested specimens. All theoretical approaches, such as CC, DBT, CBT, and CBBM were used in the current study to develop fracture energy vs crack length relationship, which is commonly known as the R-curve. Figure 19 presents variation in GII obtained for sawn timber and timber-adhesive bond specimens. Classical data reduction techniques, such as CC, DBT, and CBT showed similar trends of GII for both sawn timber and timber-adhesive bond interfaces. However, DBT predictions were greater than the other two methods since crosshead displacements are used in DBT. In contrast, CBBM, which uses equivalent crack length instead of actual measured values, showed different variation in GII, compared to other techniques, as shown in Fig. 19. In timber-adhesive specimens, the R curve gradually achieved a peak fracture energy value and then followed a declining trend, but the R-curve for sawn timber obtained using CBBM showed a somewhat upward trend with some fluctuations. During the crack development process, timber-adhesive bonded specimens sustained greater load than those for timber samples (as shown in Fig. 16) resulting in greater compliance (C0 and C), which eventually contribute to flexural modulus (Ef) and fracture energy (GII) (Eqs. 6–10). The average flexural modulus (Ef) values for sawn timber and timber-adhesive bond are shown in Table 4. It is worth noting that the fracture energy calculated using CBBM at the point of crack initiation was smaller than the values obtained from classical methods but showed a close match in terms of overall energy release rate with CC and CBT. This observation points to the fundamental issue with CBBM, which is based on load–displacement response rather than the actual crack propagation.

Fig. 19
figure 19

Fracture energy vs crack length (R-curve) for sawn timber and Timber-adhesive bond using CC, DBT, CBT and CBBM

Full size image
Table 4 Test results from sawn timber and timber adhesive bond fracture
Full size table

The R curves show that fracture energy rate at crack initiation point is low and gradually follows an upward trend as the load increased. As the crack becomes larger and gradually propagates towards the loading point, grater fracture resistance was observed due to localised compression from the applied load. This phenomenon may cause an overestimation in fracture energy. Hence, to obtain a reliably constant value of critical strain release rate (GIIc), an average fracture energy rate from 20 to 50 mm crack was considered in each test conducted in the current study.

Strain energy release rates including (GII0) and (GIIc) of sawn timber using classical methods and the equivalent crack method are presented in Table 4. Among the classical methods, CC and CBT showed consistent fracture energy release during crack propagation. The average critical fracture energy release rates obtained using CC and CBT were 1.28 N/mm and 1.37 N/mm, respectively. However, the critical fracture energy obtained from DBT was 1.82 N/mm, which is greater than those obtained from CC and CBT; this discrepancy is due to the excessive crosshead displacement used in DBT. On the contrary, the average fracture energy from equivalent crack method CBBM was the lowest; the average fracture energy from CBBM was 1.17 N/mm. Current experiments showed that the Mode-II fracture energy for radiata pine timber species varied within 1.17 N/mm to 1.37 N/mm, which is consistent with other Mode II energy levels published in the literature. Franke and Quenneville (2014) reported Mode-II fracture energy of 1.7 N/mm for New Zealand grown radiata pine by doing compact shear (CS) tests. King et al. (1999) reported Mode-II fracture energy of wet and dry radiata pine to be 1.17 N/mm using CS test. Majano-Majano et al. (2020) and Kretschmann (1995) conducted tests on Eucalyptus globulus and Sitka spruce and reported fracture energy to be 1.54 N/mm and 1.25 N/mm. is Eucalyptus globulus.

Figure 20 a presents a bar chart displaying the average fracture energy values obtained using all data reduction techniques considered in the present study. The percentages of variation based on compliance calibration (CC) technique are also shown in Fig. 20. DBT clearly predicts the highest fracture energy release rate compared to other methods with 24% and 35% greater than CC for glued timber and sawn timber, respectively. The average fracture energy obtained from CBT was only 7% and 5% greater than those predicted by CC for glued timber and sawn timber, respectively. CBBM produced 8% smaller fracture energy for sawn timber but 10% greater for glued timber fracture energy values obtained by CC. Figure 20b presents a comparison of each data reduction technique for both glued timber and sawn timber. Clearly, glued timber would require more energy to fracture when compared to an equivalent sawn timber specimen.

Fig. 20
figure 20

Overview of the results from timber and timber bond fracture

Full size image

The fracture process zone (FPZ), the area immediately near the crack tip, undergoes toughening and softening processes due to numerous microcracks, cracks-branching, and fibre-bridging. These nonlinear events should be included in R-curve as they may have impact on how crack propagates through timber and timber-adhesive bond. Since the plateau value of these curves determines the critical fracture energy (Gc), the R-curve is a useful tool for assessing the critical fracture energy in the effect of the FPZ. An overall response of R-curve for all timber and timber adhesive bond specimens obtained from CC (as a representative classical method) and CBBM is presented in Figs. 21 and 22.

Fig. 21
figure 21

Mode II fracture energy vs crack length, i.e., R curves for sawn timber using CC and CBBM

Full size image
Fig. 22
figure 22

Mode-II fracture energy vs crack length, i.e., R curves for timber-adhesive bond using CC and CBBM

Full size image

In timber structures, occurrence of glue delamination is often considered as an unlikely event in softwood timber-adhesive joints, and hence adhesive material properties are often neglected in numerical and analytical simulations of timber structures. This study showed that small imperfections, such as cutting initial cracks using mechanical tools could trigger combined failure of timber and adhesive that may require significantly less amount of energy to fracture. This clearly highlights that various types of timber connections could be vulnerable to unwanted delamination causing failure, and hence material characteristics for both adhesive and timber should be considered in numerical as well as in analytical methods. Ignoring the glue delamination behaviour in different ETPs may lead to erroneous prediction of load carrying capacity of a structure. Experimental results obtained from the current study will aid in developing reliable and accurate numerical models for timber, and hybrid structural systems that include cross-laminated timber (CLT), glulam etc.

9 Conclusion

An experimental investigation of fracture initiation, propagation, and fracture energy release rate of both sawn timber and timber-adhesive bond was conducted and analysed in the current research. DIC technique was extensively used in the current study to examine relevant failure characteristics and to propose specific testing criteria for ENF Mode II testing of timber.

Initial crack width used in ENF Mode II specimens, showed significant influence on the fracture initiation load, resistance, and stiffness of timber-adhesive bond compared to those of sawn timber. Fracture response of timber specimens was almost the same for both 0.5 mm and 1 mm initial crack. On the contrary, timber-adhesive bond specimens did not produce glue delamination when saw cut initial cracks of 1 mm and 3 mm were introduced. Reliable fracture behaviour for timber-adhesive bond was obtained only when 0.1 mm initial crack was introduced by placing a thin polyethylene tape between the two surfaces. Timber is a bio-composite material, and hence, any induced crack may coincide with existing fibre gaps resulting in reduced resistance to crack growth. Instead of a bond line, the pre-cut timber-adhesive bond specimens produced fracture in timber. As a result, they were classified as wood fracture rather than bond delamination, and the presented results clearly identified the distinguishing features between timber and adhesive bond failure.

In the present research, multiple data reduction methods, such as CC, DBT, CBT, and CBBM, were employed to analyse the fracture energy of timber and timber-adhesive bond. Some key differences were identified in the fracture resistance curve (R curve) by comparing actual vs equivalent crack monitoring techniques. CBBM showed lower crack initiation loads compared to CC, DBT, and CBT; in some cases, the R-curves obtained using CBBM were not representative of real fracture behaviour. In addition, CBBM produced a flat shape in the R-curve after reaching 30–40 mm, indicating constant fracture propagation, while CC, DBT, and CBT showed a slow upward trend.

Digital image correlation (DIC) was successfully used as a reliable tool for fracture detection and crack length monitoring in timber. The use of strain field and local displacements from every millimetre of crack path was shown to be useful to capture fracture characteristics that are essential for reliable finite element modelling. The relationship between fracture energy and crack length for timber and timber-adhesive bond can be utilized for cohesive zone modelling and continuum damage based material modelling. Finite element models along with experimental results will provide a powerful solution for various complex conditions in timber joints and connections.