1 Introduction

In modern industrial practices, composites play a crucial role in various sectors. Combining materials with distinct properties facilitates the optimisation of load-carrying components and the customisation of their strength to meet expected loads. Extensive literature outlines many material configurations, with one relatively less explored being the composite blend of wood and carbon fibre reinforced polymers (CFRP). Enhancing the structural integrity of existing elements by adhering CFRP to their external surface increases the stiffness and strength of the setup. However, subjecting the CFRP composite to high temperatures can trigger its degradation, potentially decreasing the safety of the entire reinforced structure. To mitigate the risk of heat exposure, thick layers of insulating sheathing materials are typically employed, inevitably escalating the overall technological expenses. A practical solution is placing the CFRP tape internally within the wooden cross-sectional area during the structural element’s fabrication phase. Wood serves then as a natural insulator, characterised by unique properties as charring when exposed to elevated temperatures (around 300 °C). The resulting charred layer acts as a safeguard for the inner section, maintaining a constant inside temperature at the level of 20–40 °C. Positioning the CFRP tape within this wooden cover may assure its high performance throughout the desired duration.

Many contemporary works cover the static investigation of wood and CFRP combinations. These contain existing structures’ reinforcement and the newly manufactured load-carrying components’ evaluation. A comprehensive overview of strengthening methods was presented by Franke et al. (2015) and Schober et al. (2015). The acknowledged approach is to position CFRP bars near the surface using epoxy glue (Raftery and Whelan 2014; Bergner et al. 2018; Wdowiak-Postulak 2023), which can be preloaded (Yang et al. 2016a). CFRP mats, tapes, or strips are a substitute solution. These can be mounted as L or U shapes (Brunetti et al. 2019; Vahedian et al. 2019; Zhang et al. 2024) and preloaded (Halicka and Ślósarz 2021, 2022) or put into the pre-cut incisions (Nowak et al. 2013; Morales-Conde et al. 2015). An alternative set is combining wood with CFRP during the production stage. CFRP tapes are then glued between wooden parts in a variety of locations (Yang et al. 2016b; He et al. 2022) and under specified pressure to obtain a strong adhesive. Experimental, analytical, and numerical investigations on such load-carrying structural members were presented in the earlier author’s works (Kawecki and Podgórski 2020; Kawecki 2021a, b, 2023; Kawecki et al. 2023).

Only singular studies on wood-CFRP systems, in extreme temperature conditions, exist. Testing wood-CFRP products at high temperatures should be of top priority due to the low glass transitioning temperature. One of the noteworthy experiments can be exposure to high temperatures caused by fire. Despite technological progress, researchers still face significant obstacles in conducting structural-sized fire tests. The explanation is high temperature, which obligates to secure pricey equipment, such as testing machines, force actuators, or displacement sensors, from damage. To minimise the fire exposure range, investigators try to burn specimens only in the places where the highest forces or moments occur. This approach facilitates reducing the fire source while retaining the mechanical behaviour of the member in fire. The opening group of tests provides the load being applied before ignition and being preserved constant during the experiment. Such tests were carried out on different wood species, both on small- and large-scale elements (Firmanti et al. 2004, 2006; Qin et al. 2021). Furthermore, glued-laminated timber was examined in this way (Lineham et al. 2016; Fahrni et al. 2019). The subsequent group of tests, commonly known as post-fire tests, includes subjecting samples to fire before load application (Yang et al. 2009; Quiquero et al. 2018; Chorlton and Gales 2020; Bai et al. 2021). However, the structural behaviour replication may be affected as the strategy does not consider real-time burning of load-carrying members. The first tests on burning wood-CFRP products were performed over twenty years ago (Martin and Tingley 2000; Ogawa 2000) and are starting to be under interest again (Kawecki et al. 2023; Wdowiak-Postulak et al. 2024; Zhang et al. 2024). Nevertheless, not enough knowledge on the behaviour of composites in high temperatures is still provided.

The goal of the research is to enlarge scientific understanding in the domain of wood-CFRP composites. Especially, the influence of CFRP tape placed inside the cross-section on the behaviour of the beam exposed to fire was of prime importance. The tests were comparative between glued-laminated timber beams (B) and composite beams containing CFRP tape (BW). The factor increasing the diversity of the tests was the changing cross-section width. Preliminary tests on singular specimens were performed and described in the earlier authors’ work (Kawecki et al. 2023). More detailed analyses of 24 specimens are provided in the current paper. Experimental tests were carried out on structural-sized beams, which reduced a scale effect observed for testing small elements. A pre-ignition load was adjusted to the expected nature of the structural work of the elements, emphasising the practical issues of using CFRP tapes as a reinforcement, being intended to increase the load-carrying capacity. This made discussion on the hopes and threats of using such a composite as wood-CFRP in engineering structures possible.

2 Materials and methods

Experimental studies comprised 24 structural-sized beams, with dimensions of 160 mm height, varying width (100 mm, 120 mm, and 145 mm) and 3.0 m length, tested under three-point bending with span between supports equal to \(\:L=2.8\:\text{m}\). Each six types of cross-sections are presented in Fig. 1.

Fig. 1
figure 1

Cross-sections of the tested beams

Full size image

The first reason for the size of the specimens was the standard slenderness, which should be near \(\:L/h=18\) to provide damage because of exceeding tensile strength in the tensile zone of the bottom wooden lamella under bending. Additionally, the size allowed for a relatively easy transport, because each beam could be moved manually by two people and accurately positioned at the experimental stand. Similar cross-sections were used in experiments carried out by other scientists (Ogawa 2000; Firmanti et al. 2006; Schmid et al. 2010; Quiquero et al. 2018; Chorlton and Gales 2020). Testing for behaviour variances involved three different widths of beams, based on the market dimensions of CFRP tapes. The assumption was consistent with the work of Martin and Tingley (2000).

12 B-beams and 12 BW-beams of the corresponding dimensions were prepared. The B-series specimens were made of GL24h glued-laminated timber with four wooden lamellas adhesively bonded using Loctite HB 110 Purbond polyurethane adhesive. Wood density was equal to 465 ± 28 kg/m3, and moisture content was between 12 ÷ 15%, being the values typical for coniferous wood. The mechanical properties of the timber used in the beams design were: modulus of elasticity E0,mean=11.50 GPa, shear modulus Gmean=0.65 GPa and bending strength \(\:{f}_{m}=24\:\text{M}\text{P}\text{a}.\) The first two properties, determined experimentally in the earlier author’s works (Kawecki and Podgórski 2020; Kawecki 2021a, b) were similar to the standard values. The mean actual bending strength obtained in the tests, called the modulus of rupture (MOR), was much higher than the standard bending strength and equal to 41 MPa, however being consistent with the Wood Handbook (Ross et al. 2010). The BW-series specimens were made of the same GL24h glued-laminated timber with the only variance in adding the S&P SM C-Laminate CFRP tape above the bottom wooden lamella. CFRP tape was placed in the tensile zone, as for typical reinforcing technique used for bent members, however this was covered with the bottom wooden lamella to protect the CFRP composite against high temperatures caused by fire. The composites were adhesively bonded with the same Loctite HB 110 Purbond polyurethane adhesive and with the same technological regime during production, provided by ABIES Poland Ltd., as shown in Fig. 2.

Fig. 2
figure 2

Preparing the elements in ABIES Poland Ltd

Full size image

The choice of three-point bending static scheme was dictated by the possibility of achieving the assumed effort of the tested elements on the stand, which could have been 50% higher than in commonly used four-point bending at the same load level. The idea of the self-designed loading construction was based on the lever mechanism, as shown in Fig. 3.

Fig. 3
figure 3

View on the experimental stand before beams combustion

Full size image

Single IBC 1000 L container was placed on a steel lever and was filled with the desired amount of water, that allowed controlling the loading level. Other IBC 1000 L containers were positioned as a counterweight opposite the lever. The stand was tightly fenced against the wind. For more details, refer to the earlier work (Kawecki et al. 2023).

The beams were loaded up to 90% of the characteristic effort, being understood as tensile stresses occurring in the bottom wooden lamella equal to 21.6 MPa, when characteristic standard bending strength is equal to \(\:{f}_{m}=24\:\text{M}\text{P}\text{a}\). When compared to the experimentally measured bending strength \(\:MOR=41\:\text{M}\text{P}\text{a}\), the aforementioned loading level meant about 50% of the mean actual load-carrying capacity. A relative error of the force calculated using the equivalent cross-section comparing to the values measured directly on the stand was negligible, while a relative error of deflection measurement did not exceed 5%. The results confirmed appropriate accuracy in both assumptions and stand design. For the derivation of the analytical formulas based on the equivalent cross-section model, used to estimate the characteristic effort, refer to the earlier work (Kawecki 2021b; Kawecki et al. 2023). In-situ layout for measurement of the initial force \(\:\left({P}_{init}\right)\) by AXIS FB20kN force sensor and deflection \(\:\left({w}_{init}\right)\) by HBM LVDT sensor is presented in Fig. 4.

Fig. 4
figure 4

Pre-ignition measurements of elastic properties of the beams

Full size image

An initial bending moment \(\:{(M}_{init})\) applied to the beams was calculated based on Eq. 1, according to the three-point bending static scheme. Mean measurement results for B-beams and BW-beams are given in Table 1.

$$\:{M}_{init}=\frac{{P}_{init}L}{4}$$
(1)
Table 1 Mean forces applied to the beams and corresponding bending moments and deflections
Full size table

After applying the initial load \(\:\left({P}_{init}\right)\), and in consequence the initial bending moment \(\:\left({M}_{init}\right)\) to the corresponding specimen, devices used for measurements of pre-ignition force and deflection were moved to a safe place. Then, two digital cameras for simultaneous monitoring of the beam from the left and right sides, displacement HBM LVDT sensor placed on the steel lever hidden behind mineral wool, and pyrometer Benetech GM1651 measuring temperature, placed at a distance of 1.5 m and focused on the centre of the front surface of the beam, measuring the circular surface with a diameter of 42 mm, were activated to register the beam’s behaviour (see Fig. 3 for the devices and Fig. 5 for temperature measurement point). In the next step, the furnace in dimensions of 1.0 × 0.5 × 0.5 m3 filled with 80 kg of well-dried (12% moisture content) wood with 690 kg/m3 density, providing 800 °C temperature for 30 min, was burned and positioned under the beam’s centre. For more details on the furnace design, refer to Kawecki et al. (2023). The stand was set outdoors. Hence, the experiment results could have been affected by the weather. Similar test conditions were ensured by tight fencing of the stand, choosing windless days and monitoring wind speed with NAVIS WL 11X/WDS anemometer. The low daily averaged wind speed equal to 0.12 m/s was enough to state that its influence can be neglected. One hour mean wind speed was similarly low and stable and maximal gusts were equal to 1.3 m/s, however these occurred rarely. Exemplary specimen during combustion test and static scheme are presented in Fig. 5. After a beam failure, understood as a collapse with characteristic cracking sound for B-beams or a fast increase in deflection during fire to over 50 mm for BW-beams, the furnace was taken away to avoid uncontrolled combustion. Immediately after, the beams were moved to different place, and the fire was extinguished using sand. Each test was conducted identically.

Fig. 5
figure 5

Exemplary specimen during combustion test

Full size image

The temperature curves obtained in the experiments were compared to the commonly used Standard Time-Temperature Curve ISO-834 (Eq. 2) and External Fire Curve (Eq. 3) for each specimen which made the results easier to be interpreted. The formulas can be found, among others, in Cantor et al. (2022):

$$\:{{\Theta\:}}_{\text{I}\text{S}\text{O}-834}=20+345{\text{l}\text{o}\text{g}}_{10}\left(8t+1\right)$$
(2)
$$\:{{\Theta\:}}_{\text{E}\text{x}\text{t}\text{e}\text{r}\text{n}\text{a}\text{l}}=20+660\left(1-0.687{e}^{-0.32t}-0.313{e}^{-3.8t}\right)$$
(3)

where: \(\:t\) – time (min).

After the tests, beams were transferred to the laboratory and cut at two cross-sections. The result was two slices per each beam representing the burnt cross-section at the end of the combustion test. Next, the slices were cleaned from a charred layer and photographed in high resolution with millimetre paper to prepare scaled contours in a vector-based form. Exemplary beam after the combustion test, slicing manner and B-beam and BW-beam exemplary slices with corresponding vector-based forms are presented in Fig. 6.

Fig. 6
figure 6

Exemplary beam after the test, slicing manner and B-beam and BW-beam exemplary slices

Full size image

The vector-based contours enabled to measure a mean real charring depth (CD) in the moment of beams’ failure. Because charring was observed to be four sided, the mean real charring depth was calculated based on the perimeter of the slice after removing a charred layer. As the edges of a cleaned slice were not perfectly linear and rounding in the corners occurred (comp. Figure 6), the method provided relatively easy and precise determination of the mean real charring depth. The calculations were based on Eq. 4, elaborated according to Fig. 7.

$$\:CD\left(U\right)=\frac{b+h}{4}-\frac{U}{8}$$
(4)

where: \(\:CD\) – mean real charring depth, \(\:U\) – perimeter of the slice after removing a charred layer.

Fig. 7
figure 7

The method for calculating charring depth (CD)

Full size image

A mean real charring rate \(\:\left({\beta\:}_{r}\right)\) was calculated, based on the mean real charring depth (CD) and the time of failure of each beam \(\:\left({t}_{fail}\right)\), according to Eq. 5:

$$\:{\beta\:}_{r}=\frac{CD}{{t}_{fail}}$$
(5)

Based on the charring rate, mean heat flux density (HF) for each combustion test was estimated. The work by Martinka et al. (2018) provided extensive results combining \(\:{\beta\:}_{r}\left(HF\right)\) and combustion time, and samples presenting similar values were taken for further analyses. The results obtained by Martinka et al. were analysed in terms of the charring rate values at different times of burning for varying heat flux densities. The specimens were prepared from coniferous wood with density about 450 kg/m3, being consistent with the current research. Next, the results obtained by Martinka et al. (2018) were extended to cover a relatively wide range of heat flux densities (20–110 kW/m2) according to the provided power functions. Because the wood tested by Martinka et al. (2018) had w = 0% of moisture content, the values were recalculated using Eq. 6 proposed by Cachim and Franssen (2009), into w = 12% moisture content.

$$\:{\beta\:}_{\text{450,12}}=\frac{{\beta\:}_{\text{450,0}}}{{\left(\frac{1.12}{1+w}\right)}^{1.5}}$$
(6)

where: \(\:{\beta\:}_{\text{450,12}}\) – charring rate for 12% moisture content, \(\:{\beta\:}_{\text{450,0}}\) – charring rate for 0% moisture content, \(\:w\) – moisture content

The values of \(\:{\beta\:}_{r}\) for various HF and combustion times are gathered in Table 2. The dependencies between \(\:{\beta\:}_{r}\) and combustion time for various HF were approximated with exponential functions with R2 > 0.95, and then the relationships were extended to fire exposure beginning. The approximating functions and corresponding coefficients of determination are gathered in Table 3. The dependencies defined above, allowing validation of current results, are presented in further section.

Table 2 Recalculated values of \(\:{\beta\:}_{r}\) for different HF and combustion time
Full size table
Table 3 Dependencies between \(\:{\beta\:}_{r}\) and combustion time for various HF
Full size table

3 Results and discussion

Temperature was recorded in every two seconds and progressive deflection was measured continuously during the entire test, up to failure. The results are presented in figures one after another in successive series. Each graph comprises both the mean measured temperature compared to ISO-834 and External Fire Curves and its standard deviation, middle-span deflection during fire exposure and interpretation of the failure form according to two criteria – (D) as ductile failure, understood as clearly visible growth of displacements and (R) as rapid brittle failure, understood as a collapse of the specimen with a characteristic cracking sound. After applying the criterion, the failure time \(\:\left({t}_{fail}\right)\) was marked as the value used in further estimation of charring rate. For better visibility, the same scale was used for the beams with corresponding widths.

3.1 B_100 and BW_100 beams

The first series of results for B_100 and BW_100 beams is presented in Figs. 8 and 9, respectively. The mean temperature, taking into account its standard deviation, was close to ISO-834 and External Fire Curves. Deflections during fire were comparable for all the specimens tested. The measured failure times were equal to 6 min 20 s, 6 min 40 s, 13 min 10 s and 14 min 10 s for B-beams and 6 min 10 s, 7 min 50 s, 8 min 50 s and 9 min 20 s for BW-beams, consecutively. The observation was that when CFRP tape is placed in the beam’s section, the failure form partially changes from rapid brittle (R) to the ductile (D/R).

Fig. 8
figure 8

Time-temperature-deflection relationships of B_100 specimens

Full size image
Fig. 9
figure 9

Time-temperature-deflection relationships of BW_100 specimens

Full size image

3.2 B_120 and BW_120 beams

The second series of results for B_120 and BW_120 beams is presented in Figs. 10 and 11, respectively. Specimens differed from earlier ones only in width, which was 20 mm larger. The mean temperature, taking into account its standard deviation, was close to ISO-834 and External Fire Curves. Deflections during fire were comparable for all the specimens. The measured failure times were equal to 7 min 20 s, 8 min 10 s, 11 min 50 s and 12 min 0 s for B-beams and 9 min 50 s, 11 min 30 s, 11 min 40 s and 17 min 20 s for BW-beams, consecutively. When comparing the form of failure, a full change from rapid brittle (R) to ductile (D) can be observed for wider beams. Signs of damage of BW-beams included a significant acceleration of beam deflection and no characteristic sound of brittle wood cracking was audible.

Fig. 10
figure 10

Time-temperature-deflection relationships of B_120 specimens

Full size image
Fig. 11
figure 11

Time-temperature-deflection relationships of BW_120 specimens

Full size image

3.3 B_150 and BW_150 beams

The last series of results for B_150 and BW_150 beams is presented in Figs. 12 and 13, respectively. The width of specimens was 25 mm larger compared to the previous beams. The mean temperature, taking into account its standard deviation, was close to ISO-834 and External Fire Curves. Deflections during fire were comparable for all the specimens, despite the B_150_1, where two finger joints occurred in the middle of the span (for more details, refer to the earlier work Kawecki et al. 2023). The measured failure times were equal to 11 min 0 s, 12 min 10 s, 12 min 40 s and 14 min 40 s for B-beams and 15 min 20 s, 17 min 20 s, 20 min 30 s and 29 min 50 s for BW-beams, consecutively. When comparing the form of failure, a full change from rapid brittle (R) to ductile (D), similarly to 120 mm width samples, can be observed. Signs of damage of BW-beams also included a significant acceleration of beam deflection and no characteristic sound of brittle wood cracking as for B-beams.

Fig. 12
figure 12

Time-temperature-deflection relationships of B_150 specimens

Full size image
Fig. 13
figure 13

Time-temperature-deflection relationships of BW_150 specimens

Full size image

3.4 Charring depth, charring rate and heat flux density

The next factors being determined were charring depth (CD), charring rate \(\:\left({\beta\:}_{r}\right)\) and estimated heat flux density (HF) for each specimen. The values of mean heat flux density were determined from dependencies elaborated in Sect. 2 and based on results presented by Martinka et al. (2018) and Cachim and Franssen (2009). Charring rates in dependence of failure times, obtained in current experiment, were superimposed on the graphs (Figs. 14 and 15), respectively for B-beams and BW-beams. Heat flux density value for every beam was read manually with an accuracy of 5 kW/m2 and then the results were summarised in Table 4.

Fig. 14
figure 14

B-beams heat flux density estimation

Full size image
Fig. 15
figure 15

BW-beams heat flux density estimation

Full size image
Table 4 Summary of results: charring depth, failure time, charring rate and heat flux density
Full size table

The values determined for B-beams could be treated as reference values. In the case of BW-beams, the disturbances in the failure time could happen because of the high probability of a partial beam failure caused by the temperature increase in the CFRP strips. This was indicated by the manner of failure of BW-beams, which changed from sudden collapse with characteristic cracking sound for B-beams to plastic flow for BW-beams.

3.5 Results comparisons and discussion

Varying values of heat flux densities were observed for the tests, despite many efforts made to provide the same testing conditions. However, the specimens exposed to similar heat flux densities could have been compared directly between each other. This was done between BW-beams and B-beams. Variation in the heat flux density equal to 15 kW/m2 was accepted, because of approximate determination of this property. The final analysis of the results of failure times was made based on the comparison of samples with similar heat flux density values, calculating the mean value and measurement uncertainty. To decide whether the mean value can be a representative, and to estimate the size of the error, an uncertainty of the measurement was calculated using Eqs. 7, 8 and 9. The compared failure times for the corresponding specimens and statistical analysis are collected in Table 5.

$$\:\stackrel{-}{x}=\frac{\sum\:{x}_{i}}{n}$$
(7)
$$\:{s}_{\stackrel{-}{x}}=\sqrt{\frac{\sum\:{\left({x}_{i}-\stackrel{-}{x}\right)}^{2}}{n\left(n-1\right)}}$$
(8)
$$\:err=\frac{{s}_{\stackrel{-}{x}}}{\stackrel{-}{x}}100\%$$
(9)

where: \(\:{x}_{i}\) – measurement, \(\:\stackrel{-}{x}\) – measurements mean, \(\:n\) – number of measurements, \(\:{s}_{\stackrel{-}{x}}\) – standard deviation of measurements, \(\:err\) – uncertainty.

Table 5 Comparisons between BW-beams and B-beams and statistical analysis
Full size table

The measurement uncertainty analysis showed that the failure time mean coefficient can be a representative with maximum 6% error for BW_100 beams, 11% error for BW_120 beams, and nearly 8% error for BW_150 beams. The obtained failure time increase in dependence of width of the beam is given in Fig. 16.

Fig. 16
figure 16

Comparison of the mean values for all the tested specimens

Full size image

Results may be approximated using a linear function with the coefficient of determination R2 = 0.998. Clearly visible is the fact that the wider the cross-section, the higher the positive influence of CFRP.

4 Conclusion

The objective of the study was to enhance scientific knowledge within the field of wood-CFRP composites. Specifically, the focus was on analysing the impact of CFRP tape positioned inside the cross-section on the structural performance of structural-sized beams subjected to fire. A pre-ignition load was adjusted to highlight the practical considerations associated with utilising CFRP tapes as reinforcement, aimed at increasing the load-carrying capacity of wood-CFRP composites compared to wooden members.

The paper presented experimental tests on 24 structural-sized glulam (B) and wood-CFRP composite (BW) beams with applied pre-ignition load causing 90% of the effort, burned in the middle of the span. Pre-ignition static deflection and load were measured. Subsequently, the samples were burned under similar temperature conditions. The measured factors were combustion time and deflection increase. Two cross-section slices were cut from each fractured beam after the beams were cooled with sand and vector-based contours of the remaining cross-section were determined. The charring rate and the approximate heat flux density of each tested specimen were determined, enabling a direct comparison of the results. Statistically reliable number of samples made the comparison of results for reinforced and unreinforced beams in terms of failure time and nature of the failure possible.

The results obtained for several samples per each cross-section were consistent with preliminary tests. Based on the analyses performed in the current paper, it can be concluded that:

  • Introducing CFRP tapes between the wooden parts can increase the fire resistance of the structure.

  • The thickness of the wood material covering the CFRP composite is one of the most important factors.

  • The trend leading to the conclusion that the wider the cross-section, the longer BW-beam lasting in fire is clearly visible based on combustion time.

  • The beams with CFRP tapes are characterised by a different manner of collapse, transforming from instantaneous cracking (B) to the ductile failure (BW).

The authors must indicate the nonstandard test setup and procedure. The obtained results can be treated as a preliminary insight into the behaviour of wood-CFRP composites in fire. Further studies, acceptable to the regulatory bodies should be conducted using a recognised international standard, such as ISO 834.