Evaluating Mechanical Properties and Suitability of Aspen (Populus tremula L.) Load Bearing Replacements in Historical Constructions

Abstract

The replacement of historical load-bearing wooden elements made from not commonly used species such as Populus tremula L. presents significant challenges. As these species are seldomly used in modern construction, a knowledge gap exists regarding their implementation in accordance with current building codes. This study investigates the mechanical properties of European aspen (Populus tremula L.) from central Slovakia as a potential replacement for historical structures. Notably, poplar species, including European aspen, have historically been utilized for construction across various landscapes in Europe. We conducted experimental testing on visually graded aspen timber to determine the dynamic modulus of elasticity (MOE_dyn,ultr), modulus of elasticity (MOE), modulus of rupture (MOR), and density. The results were analyzed and compared to established standards for structural timber. Notably, the 5th percentileof the strength distribution (f_0.05) was determined to be 28.78 MPa, while the characteristic strength (f_k) was 26.23 MPa, and the modulus of elasticity (E_g12) was 13.60 MPa. The correlation between MOR and dynamic MOE facilitated the determination of MOR by non-destructive testing (NDT) using the Sylvatest Duo^®. This simple linear model could grade 49% of boards into the higher strength class C30. The additional parameters and their interactions in multiregresssion models improved the predictability of the bending strength of aspen. The advanced model graded 68% of boards into C30. These characteristics, along with aspen’s growth potential, make it a promising candidate for replacing damaged structural elements in historical constructions. Our findings contribute to the understanding of the potential of European aspen as a structural timber, highlighting its viability as a fast-growing hardwood species.

1. Introduction

Poplar wood played a certain role in human construction and material culture throughout history, recognized for its versatility and widespread availability. Its utilization spans continents and eras, from ancient Chinese architecture to European and North American constructions. While not as strong as other hardwoods, poplar’s low density and workability made it a preferred choice for various applications, including building structures for agricultural uses. This historical journey of poplar wood’s usage provides valuable insights into the adaptive practices of past societies in response to the availability of local resources and the technological advances in construction methods.

The historical use of poplar wood dates to ancient times in China, where it was highly valued for its accessibility and ease of use [1]. With its lower density and ease of shaping with hand tools, poplar was a suitable option for traditional wooden architecture. Its local availability and ease of transport made it ideal for crafting both large structural components and smaller elements in many regions.

Recent studies have shown that poplar was used as a structural element in an ancient boat, estimated to be 1000–1200 years old, and found in the Moldavian region of Romania [2,3]. In Europe, round wood or sawn timber from European black poplar has been used for centuries. In the Mediterranean region of France, poplar was mainly used for rafters, stringers, and temporary construction platforms to support workers, equipment, and materials [4]. In the United States, poplar was similarly used for rafters, stringers, studding, and sheathing, thanks to the fact that poplar species grown in North America have strength properties comparable to other common construction woods such as spruce, pine, and fir [5].

In central and southern Slovakia, particularly in the Detva and Hont regions, locally available poplar timber was widely used in agricultural buildings, such as barns and haylofts, as well as in the roof structures of peasant houses [6].

Unfortunately, the utilization of various wood species in construction significantly decreased compared to historical practices in Slovakia [6]. From 2005 to 2014, less than 6% of pioneer wood species were used, with rowan at 2.5%, birch at 2.4%, aspen at 0.1%, and alder at 0.3%. Other woods accounted for 2%. Presently, the representation of aspen in Slovak forests is less than 0.1%. Aspen is primarily found alongside more dominant tree species. While the use of poplar decreased, its potential in modern construction is noteworthy. Poplar (Populus spp.), including European aspen (Populus tremula L.), demonstrates significant potential due to its fast growth (10–20 years to harvest) and adaptability for engineered wood products such as plywood, LVL, and CLT. Technological advancements, such as thermal and chemical modifications, address durability concerns, enhancing its suitability for structural applications. Moreover, extensive poplar plantations across Europe provide a renewable and reliable resource [7].

Currently, there are no established standards for visual grading of aspen wood, nor any recognized guidelines for machine grading. While aspen (Populus tremula L.) was initially classified under conifers in STN EN 338, it no longer holds this classification in recent versions of the standard due to the high variability of its properties. Despite the decreased usage of aspen and other non-typical woods in contemporary construction, historical evidence suggests their widespread use [8,9].

To maintain the vast historical heritage, preserving traditional wooden architecture often requires replacing structural elements or their components. This process involves removing damaged sections and substituting load-bearing elements with materials that match with the original species and age. However, this practice is problematic due to the lack of mechanical properties or strength grading knowledge of wood species used in the past [10,11]. Reliable grading classes of uncommonly used woods do not exist. In the case of poplar, only Germany and France assigned a strength class to visually graded poplar species that is usable for Populus nigra and Populus × euramericana, respectively, harvested in the regions of these countries.

To understand the properties of wood, it is necessary to conduct experiments either for in situ use or in an ex situ setting using a specialized testing apparatus. Evaluating mechanical properties usually requires destructive testing methods, but can also be supplemented by non-destructive methods for closer classification [12]. Visual assessment, a commonly used non-destructive technique, is limited in reliability and objectivity, focusing on wood defect detection rather than directly measuring strength, elasticity, or density [13,14].

In the realm of predicting bending strength modulus of rupture (MOR), the key lies in understanding bending stiffness, specifically the dynamic modulus of elasticity (MOE). This can be assessed non-destructively by introducing an acoustic wave longitudinally into a board and measuring its velocity or resonance frequency to derive the dynamic MOE. This dynamic MOE, in turn, serves as a crucial factor in anticipating MOR, the ultimate bending strength. There are many non-destructive methods for determining mechanical properties. All of them are meant to be for examinations of various parameters (density, moisture, decay, deformations, etc.) [15]. Various industrial strength grading systems employ methods such as machine vision, X-rays, and mechanical bending to determine the static MOE [16]. For our research and application simplicity, we choose ultrasound methods using a Sylvatest Duo^® device (CBS-CBT, Paris, France). Preserving historical architecture for cultural identity and promoting the utilization of less common wood types in modern construction requires experimental tests to establish the mechanical properties of aspen and similar woods.

The article aims to assess aspen wood (Populus tremula L.) for its suitability as a load-bearing element in timber structures for replacement purposes of historical load-bearing construction elements. The study seeks to bridge the significant knowledge gap between the historical utilization of aspen wood and the advances in contemporary woodworking technologies. To establish aspen wood as a viable construction material, it is essential to characterize its properties, including strength, elasticity, and density. A developed ultrasound test-based model allows the evaluation of aspen bending strength nondestructively and thus advances the strength grading process. Another improved model narrows down the grading characteristics that influence the aspen bending strength the most and improve the reliability of strength grading. Furthermore commonly used multiregression models [17,18,19,20], that used individual indicative properties, we added to the model’s interaction of these properties.

2. Materials and Methods

2.1. Test Specimens

Beams of European aspen (Populus tremula L.) with dimensions of 3100 × 160 × 60 mm (length × width × thickness) were subjected to mechanical testing. The material underwent a comprehensive preparation process, including cutting, drying, and visual sorting. The wood was sourced from the University Forest Enterprise of the Technical University in Zvolen. Harvested wood was cut into logs on site and subsequently transported to the outdoor log yard. A total of 13 logs were prepared for further processing. Upon external inspection, the wood exhibited no visible damage and was free from decay caused by insects or fungi. The logs were sawn into beams using a flat-sawn pattern, yielding 80 samples. These boards were then dried in a conventional drier to achieve a moisture content of 12%. A standard drying schedule for poplar wood was selected. The drying temperature was 70 °C, the drying time was 2 weeks, and the humidity was continuously adjusted based on moisture content readings of five boards within a batch. Dried boards were kept in laboratory conditions at the temperature of 20 °C and tested within a week after drying. The moisture content of wood was measured before and after testing.

2.2. Classification and Visual Inspection of the Material

Visual grading was accomplished according to DIN 4074-5 (2003) [21]. Out of the 80 dried beams, 41 beams were suitable to be used for the four-point bending experiment. The other beams (39) were deemed unsuitable after drying due to the numerous occurrences of knot areas, but mainly due to out-of-standard curvature threshold. The selected set of suitable beams met the criteria for inclusion in the at least LS10 class. For further analysis, the knot dimensions on four sides of a board were determined and each board was characterized by maximum knot diameter (D_max) and maximum knot to height/width ratio (A) according to standard DIN 4074-5. The ring width was not determined because aspen belongs to the diffuse-porous species class and confirmed that the ring width (growth rate) plays a weak correlation with the density of wood and consequently with the mechanical properties of aspen wood [22].

2.3. Acoustic—Non-Destructive Method

For non-destructive testing, the Sylvatest Duo^® (CBS-CBT, Paris, France) device was used. Measurement with this device is based on ultrasound waves to evaluate dynamic mechanical properties of test samples (Figure 1). This instrument is equipped with dual probes, one serving as a transmitter and the other as a receiver. It precisely measures the time required for an ultrasonic wave to traverse the specified material from the transmitter to the receiver. The temporal data was used for the calculation of the speed of sound propagation (c) within the wood. When coupled with density, this information provides insights into the dynamic modulus of elasticity. To conduct the assessment, the probes are carefully inserted into pre-drilled holes on the specimen faces. The dynamic modulus of elasticity, E_dyn,_ultr was calculated according to the following equation:

$$E_{dyn}=c^2\times \rho$$

(1)

where:

E_dyn dynamic modulus of elasticity [MPa]

c speed of sound [m·s⁻¹]

ρ density at w ≈ 12% [kg·m⁻³].

2.4. Mechanical Tests According to the EN 408

As part of testing the quality of structural wood, the following properties were determined:

• Global modulus of elasticity in bending E_g12 at 12% of moisture content,
• bending strength f_m,
• the density of the sample taken for gravimetric measurement ρ₁₂.

The bending tests were carried out according to EN 408 [23] (Figure 2) using the computer-controlled universal testing machine LABTEST 6.250 (LABORTECH, Prague, Czech Republic) (Figure 3). The loading process was executed at a controlled speed of v = 0.45 mm·s⁻¹. The linear variable differential transformer sensor FWA T150 (Ahlborn, Holzkirchen, Germany), with a precision of 0.02 mm, mounted at the center of a jig, measured the global deflection of the neutral axis between supports.

Wood moisture was determined in two ways. Indicatively, it was measured with a Hydromette HT 85 T hygrometer, and the exact moisture was determined by the gravimetric method. Before the destructive test, the moisture was measured on the boards. After destructive testing, a defect-free sample containing the entire cross-section was taken as close as possible to the rupture zone according to EN 408 [23]. The specimens were dried at a temperature of 103 ± 2 °C. Subsequently, oven-dried specimens were weighed, and their density was determined. The results of both E_g and ρ were adjusted to moisture content w = 12% according to EN 384 [24].

2.5. Determination of Global Modulus of Elasticity in Bending E_g

The global modulus of elasticity E_g was calculated according to EN 408 [23] using Formula (2). However, an exception allowed by EN 408 [23] has been applied in this case: when the shear modulus G is unknown, it is assumed to be infinite.

$$E_g=\frac{3{al}^2-4a^3}{2b{h}^3\left(2\frac{w_2-w_1}{F_2-F_1}\right)}$$

(2)

where:

E_g global modulus of elasticity [MPa],

a distance between the load point and the nearest support [mm],

l length [mm],

b cross-section width [mm],

h cross-section width [mm],

F₂− F₁ load increment from the linear part of the stress–strain curve,

w₂− w₁ deformation increment corresponding to F₂ − F_1.

2.6. Determination of Bending Strength

After the experimental testing of poplar beams and subsequent data collection, a comprehensive set of descriptive statistics was applied to summarize means and variability within the dataset (

Table 1. Selected properties of structural size aspen (Populus tremula L.) wood.

	$\overline{\mathbf{x}}$	Max	Min	CV %
fm [MPa]	52.40	79.14	26.74	28
Eg12 [MPa]	13,601	17,581	9446	14
Edyn,ultr [MPa]	16,286	19,880	12,625	13
ρ12 [kg·m−3]	510	612	451	9
c [m/s]	5645	6105	5093	6
A [%]	20.1	49.6	5.2	55
Dmax [mm]	16	34	8 *	57

* The lowest non-zero value.

).

Bending strength f_m is calculated according to EN 408 [23] using Formula (3). F is the highest force used in destructive testing for each sample and parameters b and h are the width and height of a cross-section of the beam.

$$f_m=\frac{3Fa}{{bh}^2}$$

(3)

Subsequently, the values for the 5th percentile were determined using the methodology outlined in EN 14358:2016 [25], which begins by sorting the data in ascending order and calculating the position of the 5th percentile. The position was calculated using Formula (4), where P is the position and N is the sample size.

$$P=\frac{5}{100}\times (N+1)$$

(4)

Using a sample size of 41, we arrive at P = 2.1. Since P is not an integer, we must interpolate between the two closest ranks.

Following this, characteristic values were computed to represent the lower bounds of the material’s strength properties, ensuring that 95% of the sample exceeded these thresholds. The characteristic MOR value can be obtained from the results of the 5th percentile according to EN 14358:2016 using the non-parametric calculation Formula (5) [25].

$$f_k=f_{05}\left(1-\frac{k_{0.5,0.75}V}{\sqrt{n}}\right)$$

(5)

where:

n is the number of test values,

f₀₅ is the 5th percentile for bending strength from test data,

V is the coefficient of variation,

k_0.5,0.75 is a multiplier to give the 5-percent lower tolerance limit with 75% confidence:

$$k_{\mathrm{0.5,0.75}}=\frac{0.49n+17}{0.28n+7.1}$$

(6)

Models predicting characteristic values of strength were proposed. The first model took into consideration the well-known linear relationship between dynamic modulus and bending strength. The second multiparameter linear regression model used other material characteristics (E_dyn, A, c, D_max, and ρ) and their interactions (E_dyn × A, E_dyn × D_max) for the prediction of bending strength variance. The remaining interactions (E_dyn × A, E_dyn × c, E_dyn × v, A × D_max, A × ρ, A × c, c × D_max, c × ρ, and D_max × ρ) were shown to be less important and are not mentioned in this study.

3. Results

The MOR of our data obtained according to EN 408 [23] is shown in Table 1. Consequently, the value of the fifth percentile of the MOR was calculated as f₀₅ = 28.78 MPa. The characteristic MOR (f_k) value was obtained from the results of the fifth percentile. The non-parametric calculation for f_k was selected based on the findings of Weidenhiller et al. (2023) that show no significant difference between parametric and non-parametric methods for calculating characteristic strength f_k according to EN 14358:2016 [25,26].

Statistical analyses were performed using STATISTICA 14 software (TIBCO Software Inc., Palo Alto, CA, USA). The results of basic descriptive statistics are presented in

Table 2. Basic descriptive statistics.

N	Mean LS10	Mean LS13	SD 1	SD 2
39	46.32	57.38	12.21	14.38

.

During the visual sorting of wood, attention was given to the following factors that can influence wood quality: knots, fiber deviation, shape distortion, wood staining, rot, insect damage, reaction wood, and the presence of heartwood. Our study specifically focused on evaluating the presence of knots and their impact on mechanical properties, particularly strength. Visual classification assigned the samples to LS7, LS10, and LS13 categories in a ratio of 2:14:25. Since only two measurements fell into the LS7 class, we decided not to include these samples in statistical analyses, as such a limited sample size would not provide reliable results for inductive statistics methods.

Based on the test results of a t-test presented in

Table 3. Two sample t-test for the difference in strength means.

F-Test	p	t-Test	SV	p
1.39	0.547	−2.42	37	0.020

, the null hypothesis about the equality of average strengths in classes LS10 and LS13 was rejected. At the selected significance level of the test α = 5%, we observe a significant difference in strength (p = 0.020).

In Figure 4, we graphically present 95% interval estimates for the mean strength value in classes LS10 and LS13.

Results of the calculated values for aspen wood from our dataset could potentially be assigned to class C24 according to EN 338 [9]. By comparing our findings and recalculations of the 5th percentileof strength (f₀₅) to the characteristic strength f_k = 26.23 MPa outlined in EN 14358:2016 [25], classification into the appropriate strength class can be achieved based on our observed results. Another factor determining the class is the mean modulus of elasticity E_mean and characteristic value of density ρ_k, both of which exceeded the C24 class requirements, potentially granting an even higher class. However, the main deciding factor is the characteristic strength f_k. Supporting evidence for the C24 class classification is the characterization of other poplar woods into C24 and higher classes in Germany and France according to EN 1912:2024 [27].

Due to the high correlation, the dynamic modulus of elasticity may be used as an indicative property for the strength classification of aspen wood. The lower level of the 90% prediction interval shows the limit below which 5% of the f_m/E_dyn,ultr correlation dependence values lie. This relationship can be observed in Figure 5, where the intersection between the prediction interval and the 5t^h percentile strength reduced to the level of the characteristic strength f_k is depicted. The final linear dependence (slope b = 0.0042) passing through this intersection gave the following model of the characteristic strength prediction: f_k,model = −39.19 + 0.0042 × E_dyn,ultr. The selected multiparameter regression model results are listed in

Table 4. Regression models for significant predictions of the bending strength f_m. The general prediction function is f_m = p₀ + p₁ × E_dyn + p₂ × A + p₃ × c + p₄ × D_max + p₅ × ρ + p₆ × E_dyn × A + p₇ × E_dyn × D_max + Σp_i × NSM_i.

fm = F(Independent Variables)	Variance Explained, %	p-Value of a Weak Model
F(ρ)	15.19	0.00
F(Dmax) **	16.74	0.00
F(c) *	21.10	0.00
F(A) **	23.02	0.00
F(Edyn)	35.57	0.00
F(c + ρ) **	36.15	0.00
F(Edyn + Dmax)	47.43	0.00
F(Edyn + A)	48.66	0.00
F(Edyn + A + c)	48.67	0.00
F(Edyn + A + Edyn × A)	52.13	0.00
F(Edyn + Dmax + Edyn × Dmax) *	52.50	0.00

Significance of absolute parameter p₀: (**) at p-value < 0.05, (*) at p-value < 0.1, (none) at p-value > 0.1, bold format of an independent variable stands for a significant parameter of the variable at p-value < 0.05, normal format of an independent variable stands for a significant parameter of the variable at p-value < 0.1, and a red format of an independent variable stands for a parameter that is not a significant NSM is a non-significant two-variable multiplication.

. We discarded models with more than three variables and models with more than two nonsignificant parameters. The single independent variable models started with variance explained at 15% for density and finished with already mentioned 36% for dynamic modulus. The most advanced model was able to predict up to 53% of bending strength variance while keeping both the model and parameters of the model statistically reliable for prediction of bending strength: f_m = −62.3 + 7.46 × E_dyn + 3.01 × D_max − 0.213 × E_dyn × D_max.

4. Discussion

Due to the limited research on the mechanical properties of aspen wood, this comparative analysis includes Populus tremula L. as well as other poplar species. The findings from various studies on poplar wood are summarized in

Table 5. Comparison of the properties of poplar wood with other studies.

	[16] *	[29] *			[19] **			This Study **
Wood species	Populus tremula L.	P. trichocarpa × P. deltoides			Populus × euramericana			Populus tremula L.
		Set 1	Set 2	Set 3	Set 1	Set 2	Set 3
Cross section [mm]	20 × 20	20 × 20	20 × 20	20 × 20	150 × 50	150 × 80	200 × 100	150 × 50
ρw,12% [kg·m−3]	458	423	446	391	436	403	385	509.6
MOE [MPa]	10,900	7857	7783	6510	7053	7432	7659	13,601
MOR [MPa]	97.3	62	69	74	39.9	37	34.59	52.4
Edyn,ultr [MPa]					9528	8394	7699	16,286

* Clear wood samples, ** structural size samples.

. Study [16] was identified as the most relevant for comparison, as it specifically investigated aspen wood (Populus tremula L.). However, this research was conducted on small, clear wood specimens. As expected, their results show higher strength and modulus of elasticity due to the size effect [28].

A key distinction between our study and others is their focus on poplar hybrids. For instance, De Boever et al. [29] examined Populus trichocarpa × Populus deltoides using clear wood specimens and bending tests according to EN 408, which yielded significantly lower values than those reported in aspen. Casado et al. [19] investigated Populus × euramericana using structural-sized samples of varying dimensions. Their study, which also included tests on the same cross-section, reported lower bending strength than the MOR found in this study (Table 5). The observed 24% decrease is largely attributed to the use of a different poplar hybrid and corresponds to the 14% lower density of Populus × euramericana compared to Populus tremula L.

Differences in grading systems further contribute to the reported variations in MOR, particularly reflected in the disparity between characteristic strengths (26.2 MPa vs. 18.2 MPa in [19]).

The comparison between the two grading classes confirmed that knot presence, a key grading parameter in our analysis, significantly influences the ultimate strength of the boards. A 19% difference in mean modulus of rupture (MOR) between Slovakian aspen boards graded that LS10 and LS13 was statistically significant. This finding aligns with the 18% characteristic strength difference observed in visually graded Populus nigra in Germany [27], suggesting a comparable distribution of bending strength between these species.

Similarly to Oliveira et al. (2001), parameters of a linear model predicting the bending strength based on dynamic modulus of elasticity were determined. The model was further expanded to predict characteristic strength that provides a reliable prediction of aspen wood’s bending characteristic strength [17]. A 35.6% bending strength variance could be explained by our linear model, developed by Cheng and Hu (2011), compared with 18.23% reported on a 48-sample size of poplar wood (Populus tomentosa Carr.) [20]. This model offers a dependable approach for classifying boards into strength classes using nondestructive testing. Furthermore, the model can be refined as additional sample data become available. Model predictions indicate that half of the measured samples (49%) could be classified into strength class C30.

This standard model could be improved by adding more variables that affect the strength of wood including interaction of the variables. Hanhijävri and Ranta-Maunus (2008) reported density and knots dimensions to be an effective combination for spruce and especially for pine [18]. Our best model for aspen considers the dynamic modulus as the main component and maximum diameter of a knot D_max and the interaction between the modulus and the diameter (E_dyn × D_max). In this case, 68% of boards could be classified into strength class C30. The models based on other variable combinations showed lower prediction reliability. Additional research with a larger sample size is necessary to further validate and enhance the model’s accuracy.

5. Conclusions

Based on the results of this research, we conclude that European aspen (Populus tremula L.) from forests in the central region of Slovakia demonstrates significant potential for use in construction. With its documented mechanical properties, aspen wood could serve as a suitable replacement material for historical load-bearing structures originally constructed with aspen. Key findings include:

• Static modulus of elasticity measured at E_g12 = 13,601 MPa, and the bending strength at f_m = 52.40 MPa.
• Characteristic strength values were f₀₅ = 28.78 MPa and f_k = 26.23 MPa, affirming the structural quality of the wood.
• The average dynamic modulus of elasticity for the tested beams was E_dyn,ult = 16,286 MPa.
• A set of 41 visually graded beams (LS10) can be classified into strength class C24.
• Dynamic modulus of elasticity shows strong potential as an indicator for classifying aspen wood into higher strength classes. A simple linear model based on nondestructive dynamic modulus can grade 49% into the higher strength class C30.
• The additional parameters in the multiregresssion model, such as maximum knot diameter and interaction between D_max and dynamic modulus, improve the predictability of the bending strength of aspen. The interactions of the indicative properties were shown to be an effective tool. The advanced multiregression model graded 68% into C30.