Experimental study of wood shaving addition in mortar and statistical modeling on selected effects

1 Introduction

A lot of research has been performed on the utilization of agricultural or industrial waste in concrete. Due to the fact that concrete is widely used and has a long service life, wastes used in it are removed from the waste stream for a long period. As the amount of aggregates required in construction industry is large, the environmental benefits of natural aggregates replacement by waste are not only related to its safe disposal, but also to the mitigation of environmental impacts arising from the extraction of aggregates, i.e. the visual intrusion and the loss of countryside. Studies [1], [2], [3], [4], [5], [6], [7], [8], [9], [10], [11] have been conducted towards the assessment of physico-mechanical properties of concrete containing wood shavings or sawdust as aggregates. Wood shavings and sawdust is waste coming from the wood industries, generated from cutting, milling and drilling operations during wood products preparing processes. Wood shavings and sawdust properties may vary significantly depending on factors as geographical origin of the timber, type of timber, part of the tree, type of industrial process which produces the shavings, etc. As in most cases of lightweight aggregates, the substitution of conventional aggregates by wood shavings or sawdust is mostly made based on the criterion of “by volume” substitution. Conventional coarse or fine aggregates replacement by the same volume of wood shavings or sawdust [1], [6], is usually expressed as percentage (%).

Due to (a) the variability of replacement materials, (b) their significant differences from natural aggregates and (c) the variability of parameters affecting the properties of concrete or mortar itself, data obtained from tests on mechanical properties of concrete or mortar specimens containing wood shavings are based on a multifactorial process. When these results come from totally different laboratory processes, their metrological traceability is of great significance in order to achieve inter-comparability. A standardized protocol for the experimental design and referencing of all essential relative data is needed (as proposed by [12] for conventional aggregate replacement by plastic) in order to facilitate any attempt to compose the results from studies when wood shavings of different origin and characteristics are used and the percentage of aggregate replacement varies. Different reporting on the total experimental procedure arises difficulties in comparing results coming from different laboratories and the statistical inference on the effect of natural aggregates replacement by wood shavings.

The present research deals with the study of wood shavings use as part of the conventional aggregates in mortar, and especially on the establishment of statistical models to predict mechanical properties of mortar containing wood shavings as a partial replacement of conventional fine aggregates. The outcome is standardized so anyone using this protocol would produce results that will be comparable to other similar studies.

2 Materials and methods

Cement type IV/B (P-W) 32.5 N and crushed limestone aggregates with a 4.5 mm maximum size were used in all mixtures. Bulk density of fine aggregates was 1740 kg/m³ (2.7% standard uncertainty, based only on the standard error of the mean). Wood shavings used in this study were generated in a factory by mechanical processing of two types of raw wood, ayous (Figure 1A) and beech (Figure 1B). Beech is a widely used wood in furniture industry. Ayous was chosen as a totally different, more lightweight type of wood. Bulk density of ayous shavings was 43±1 kg/m³ and bulk density of beech shavings was 64±2 kg/m³. Bulk density measurement procedure was repeated 10 times, providing the precision of the method under repeatability conditions [13]. This type A standard uncertainty was found to be representative of all contributing uncertainty parameters; it was compared to a result of a type B relative standard uncertainty based on both the resolution of the volumetric tube and the verification scale interval (e) of the used weighing scale (all terms defined in JCGM 200:2012 [14]). The observed uncertainty for the measured bulk density is attributed to a strong relationship of that characteristic to the wood processing method used to produce the shavings. This is expected to be an intrinsic characteristic of this material. If wood shavings are intended to be used as a construction material, the attribute of bulk density should be strongly incorporated in any relevant study. The superplasticizer used was a second generation polycarboxylic ether superplasticizer.

Figure 1:

Wood shavings used in the study: (A) Ayous, (B) Beech.

Usually, particle distribution of materials is estimated by sieve analysis. In the case of shavings a significant question emerged upon which is the actual dimension that corresponds to the nominal sieve size. To asses that, the sample that was taken in order to be analyzed using sieve analysis was also initially measured by a totally different method. By this method, approximately all shavings having a length of more than 3 mm (actually able to be optically discriminated) were measured using a high accuracy digital caliper. Two dimensions were measured on each shaving: length, which was taken to be the maximum dimension, and width, which was taken to be the dimension of the shaving on an axis perpendicular to the length’s. As it is shown in Figure 2 the shaving width is not statistically related to its length (Ayous: r=0.04, Beech: r=0.20). Shavings smaller than 3 mm were chosen not to be measured using the caliper because (a) their measurement was not practicable due to their very small size and very large population and (b) it was observed that in such shaving sizes there was no significant discrimination between the length and the width of the shaving. The essential question was whether the shaving passes the sieve according to its length or its width which is obviously redundant for such small shavings. Following, on the same samples that were partly measured using a caliper, sieve analysis was performed as in fine aggregates. Results from the particle size distribution analysis of the two types of wood shavings and fine aggregates are presented in Figure 3. As it is obvious by this figure, almost the total of shavings pass through 5 mm sieve. Since, within both samples, a large population was measured to have a length of more than 5 mm, it is deduced that during the sieve analysis the critical dimension of the shaving is width and not length. This is also verified (Figure 2) by the fact that only a small proportion of the biggest shavings were measured to have a width of more than 5 mm which denotes that the shaving width distribution could be said to be strongly related to the sieve analysis result. It should be also denoted that this sieve analysis result should be used only as a criterion for a qualitative assessment for the shaving material before mixing [2] as there is no proof that this shaving geometry holds intact, also, after this material is added to the mixture.

Figure 2:

Width versus length for the two types of wood shavings.

Figure 3:

Results of sieve analysis.

At first, a reference mix was made with an aggregate to cement ratio equal to 3, water to cement ratio equal to 0.5 and 1% by mass of cement superplasticizer. Then, three levels of fine aggregate replacement were used, 30, 50 and 70% by volume. Unit weight (density) tests were performed after the mixing and before mortar was poured into the molds. Unit weight (D) was determined by measuring the mass of mortar (m_u ) contained in a known volume (V_u ) of a sample of fresh mortar as described in ASTM C138:

A calculation was used to illustrate the changes of cement proportion in the mixture and whether this stays practically constant. This calculation was used, also to assess the impact of wood shavings degree of compaction as the air within the initial amount of this “fluffy” material (before mixing) was displaced by all the other constituents of the mixture (during mixing). This is needed especially in the case of wood shavings, as this material is a lightweight material with a typical curved shape (Figure 1), different from the conventional aggregates. After the fresh mortar unit weight was measured, the ratios of the initial masses of the constituents being mixed were used in order to assess the mix proportion of each constituent. A sharing of the measured unit weight to the contributing constituents was calculated based on the reasonable assumption that the final mixture was homogeneous throughout its volume. Mass ratio for each constituent is equal to the initially decided for the fresh mortar prepared by mixing (aggregate to cement ratio 3, water to cement ratio 0.5 and 1% by mass of cement superplasticizer) (Eq. 2).

where m_i,init is the mass of constituent i initially decided before mixing and m_i is the mass of constituent i within any sample (part) of fresh mortar. In any case of a sample of fresh mortar, m_u corresponds to ∑imi as in Eq. 1. The mixture proportion of constituent i (MP_i ) is defined as:

The calculated mixture proportions are shown in Table 1. After regression for curve fitting for the values of cement mix proportion change (CMPC) as a function of fine aggregates replacement according to Eq. (4), it was found that for ayous wood h₁=0.49±0.11 and h₂=0.0053±0.0017 (R²=0.9997) and for beech wood h₁=0.51±0.04 and h₂=0.00041±0.0007 (R²=0.999).

According to the calculated mix proportions, the cement mix proportion increases significantly as the percentage of by volume replacement of conventional aggregates increases (Figure 4). This outcome is expected to be more significant when the replacement material has a lower specific gravity and/or is more “fluffy”. This should be taken into account each time a lightweight and/or “fluffy” material is used to replace conventional aggregates.

Figure 4:

Percentage of cement mix proportion change (CMPC) versus the percentage of fine aggregates replacement.

Because of the high water absorption of wood shavings, they absorb some of the water of the mixture, thus leaving insufficient water for the workability and the setting of the cement. For this reason some researches [1], [2], [4], [6], [7] use water saturated shavings or additional water. In both these cases the final and actual water to cement ratio is unknown as the excess water quantity used is not easy to be assessed on whether it stays inside the porosity of wood or it is not absorbed, for the above cases, respectively. Alternatively, within this study, it was chosen to use the wood shavings in an unprocessed form and no excess water was added to the mixture. The advantage of this option was assumed to be that even if a proportion of the water was absorbed by wood shavings during the mixing, this would be in a form of water-cement sub-mixture, which assures that the probability this water to interact with cement was maximum.

The constituents were mixed in a mixer at slow speed in order to obtain good homogenization. First, cement and aggregates mixing took place. Then the water with superplasticizer diluted in it was added. Flow test of mortar was carried out according to ASTM C 1437 [15]. Specimens from each mixture were cast with dimensions 40×40×160 mm in order to perform all tests. Non-destructive test of ultrasonic pulse velocity (UPV) was carried out at the age of 28 and 365 days by using the method described in ASTM C 597 [16], specifically using a portable ultrasonic non-destructive digital indicating tester (PUNDIT). Strength tests were carried out at 7, 14, 28 and 365 days of curing. Flexural Strength test was carried out by center-point loading as described in ASTM C 293 [17]. The end parts of the prisms that were left intact after failure in flexure, were used to carry out equivalent cube test by applying the load through square steel plates, 40 mm size. The flexural strength and the non-destructive tests results reported correspond to the average of three tested specimens. The equivalent cube compressive strength results are the average of six tested specimens.

3 Results

3.1 Fresh mortar

Generally, wood shavings absorb more water compared to conventional fine aggregates. For this reason, workability of the mix decreases as the % by volume replacement of fine aggregates increases (Table 2). The reference mixture, as well as A30Sh and A50Sh, were self-consolidating mixtures and the diameter measured was not after the 25 drops of the table as specified in ASTM C 1437 [15]. Due to the difference in the bulk density of the two wood types (ayous and beech) an equal fine aggregates by volume replacement percentage leads to different mix proportions for each wood type. This means that, when beech is used, mix proportion for wood shavings, for a specific by volume replacement percentage, has a greater value than when ayous is used. It is possible that this leads to a greater water absorption by the wood shavings and consequently to a lower workability of fresh mortar.

Table 2:

Flow test results.

	Ref.	A30Sh	A50Sh	A70Sh	B30Sh	B50Sh	B70Sh
Flow without table drop	24.8	22.3	21.8	n.s	n.s	n.s	n.s
Flow with table drop	o.f.	o.f.	o.f.	22.4	22.2	21.1	20.1

1. n.s., Non-significant; o.f., over flow.

The unit weight decreased as % by volume replacement of fine aggregates increased (Figure 5). This reduction is attributed to the fact that the wood shavings have less specific gravity than conventional aggregates.

Figure 5:

Unit weight of fresh mortar versus % by volume replacement of conventional aggregates.

3.2 Hardened mortar

The results of flexural and equivalent cube compressive strength tests are shown in Table 3. The flexural and compressive strength of mortar containing wood shavings decreased as the fine aggregates replacement increased. This decrease is attributed to the weaken bonding of mortar cement paste and wood shavings, as compared to the bonding of mortar cement paste and conventional aggregates.

Table 3:

Mechanical properties test results.

Samples	Flexural strength (MPa)				Compressive strength (MPa)
	7 days	14 days	28 days	365 days	7 days	14 days	28 days	365 days
Ref.	10.1±0.4	9.3±2.2	9.1±0.9	12.3±0.9	43.7±0.9	49.2±0.9	58.1±2.1	74.4±2.7
A30Sh	7.8±0.1	8.5±1.0	6.8±0.9	9.9±0.8	31.7±0.6	39.4±0.4	45.0±0.9	53.5±0.3
A50Sh	6.5±0.1	7.3±0.8	7.6±1.4	9.3±1.7	26.2±0.3	32.9±0.5	41.5±0.3	46.0±0.7
A70Sh	6.2±1.0	7.2±0.7	7.7±0.1	9.2±0.1	20.8±1.1	23.8±0.6	29.1±0.8	32.2±0.8
B30Sh	7.7±0.1	7.9±0.8	8.4±2.6	9.9±1.7	29.2±0.7	34.9±2.4	41.1±0.4	46.3±3.6
B50Sh	7.1±1.7	8.0±0.9	7.5±1.5	8.4±0.9	21.7±1.8	26.8±2.0	31.3±1.5	33.7±4.6
B70Sh	5.1±0.1	7.1±0.1	6.6±0.9	7.3±0.2	14.8±1.3	21.7±1.2	28.0±0.8	26.9±3.3

It is shown that the resulting decrease in strength of mortar containing wood shavings is not attributed only on the impact of the replacement of fine aggregates with wood shavings. The resulting significant increase in the specific proportion of cement in the final mixture is expected to have a positive impact in the value of strength. Therefore the result of strength value decrease is the combination of the simultaneous and adverse impact of the two above phenomena. It seems that a decision on fine aggregate replacement by wood shavings should not only rely on a calculation according to the volumes of these two materials as they appear before mixing. This calculation should be according to the apparent volume of each mixture constituent volume as the conditions in which this appears within the mixture.

As it is shown in Figure 6, in all cases, compressive strength of mortar containing ayous shavings was found greater than compressive strength of mortar containing beech shavings. The mean difference calculated from 12 groups of six specimens, each, having the same value for fine aggregates fraction and specimen age, was found to be 20±7%. This finding has no significant statistical relation to either the values of fine aggregate replacement fraction (Pearson r=0.255, significance p=0.423) or the values of specimen age (Pearson r=0.217, significance p=0.498).

Figure 6:

Comparison of compressive strength test results from groups of six specimens with specified fine aggregates fraction and specimen age (each group corresponding to one figure point).

According to the experimental results and following an equation based on one initially proposed by Freiesleben Hansen and Pedersen [18], compressive strength is given as a function of fine aggregate replacement fraction (W) and specimen age (t) by Eq. (5):

(5)CS(t, W)=(CS∞,1− kWn)exp[−(τ/t)a]

where CS(t, W) is the compressive strength at age t (days) when fine aggregates replacement fraction is W, CS_∞,1 is the limiting compressive strength for the reference (maximum asymptotic value of the strength for the function that fits the data), n is a shape parameter for the function of compressive strength when fine aggregates replacement fraction is W, k is compressive strength reduction parameter such that kWⁿ equals to the reduction of the specimen limiting strength due to a fine aggregate replacement equal to W, τ is a time constant and a is a shape parameter for the sigmoidal function of compressive strength with specimen age t, CS_∞,1–kWⁿ corresponds to the limiting compressive strength of a specimen with a fine aggregate replacement fraction equal to W.

This means that for a given specimen age the relation between compressive strength and fine aggregate replacement is a function of fine aggregate replacement fraction to a power n (Figure 7A). Simultaneously, for a given fine aggregates replacement fraction, compressive strength is a function of age, which corresponds to a sigmoidal curve (Figure 7B).

Figure 7:

(A) Compressive strength versus fine aggregates replacement fraction, (B) compressive strength versus specimen age.

The regression procedure using Eq. (2) based on the experimental results of the present study provided a statistically significant model (Pearson r=0.96) with parameter values: CS_∞,1=74±3 MPa, k=55±4 MPa, n=0.8±0.1, a=0.7±0.2 and τ=3.1±0.6 days.

In Eq. (5) the parameter of wood shaving type was not investigated, although the statistical significance of that result was satisfying enough in order to be used as a general model for predicting the loss of limiting strength when any kind of wood shaving is used for fine aggregate replacement. Making one step more, the parameter of wood shaving type was introduced in Eq. (5), forming Eq. (6):

(6)CS(t, W)=[CS∞,2−(k1m1+k2m2)Wn]exp[−(τ/t)a]

where m₁, m₂ equal to one when wood shaving type is ayous or beech, respectively, otherwise each equal to zero. The combination m₁=0 and m₂=0 corresponds to the case of reference specimens (no use of wood shavings). k₁ and k₂ are shape parameters, similar with k in Eq. (5).

This equation has been tried for only one type of wood per mixture and not for two types together in the same mortar mixture. When two or more types of wood shavings are to be used simultaneously in the same mortar mixture, then the use of Eq. (5) is suggested, but it is also suggested that further studies should be done for multiple kind of wood shavings in the same mortar mixture, mainly in order to investigate the significance in which this factor contributes to the uncertainty of Eq. (5) parameters. Any combination of m₁ or m₂ other than values of 0 and 1 has not been studied and it is suggested for future investigation.

The regression procedure using Eq. (6) based on the experimental results of the present study provided a statistically significant model (Pearson r=0.976) with parameter values: CS_∞,2=74±3 MPa, k₁=48±3 MPa, k₂=60±3 MPa, n=0.76±0.07, a=0.7±0.1 and τ=3.1±0.5 days (Figure 8).

Figure 8:

Compressive strength versus fine aggregates replacement fraction (A) only for ayous wood shavings and (B) only for beech wood shavings.

The results of UPV tests are shown in Figure 9.

Figure 9:

Ultrasonic pulse velocity versus fine aggregates replacement fraction.

UPV linearly decreases as fine aggregates replacement fraction increases. This is attributed to the different properties of wood compared to the properties of conventional fine aggregates. The importance of UPV is that it is significantly correlated with mortar elastic properties. A regression model was applied to the experimental data using Eq. (7):

(7)UPV(t, W)=[UPV∞+(l1m1+l2m2)W][1−exp(−t/t0)]

where UPV_∞ is the limiting UPV for the reference which is the maximum asymptotic value of UPV for the function that fits the data, UPV_∞ ·[1−exp(−t/t₀)] is the UPV of the reference (W=0) for the specified age of curing (t), m₁ and m₂ equal to one when wood shaving type is ayous or beech, respectively, otherwise equal to zero, l₁, l₂ are shape parameters and t₀ is a time constant.

The regression procedure using Eq. (4) based on the experimental results of the present study for UPV provided a statistically significant model (Pearson r=0.981) with parameter values UPV_∞ =(5.33±0.08)·10³ m/s, l₁=(−1.73±0.17)·10³ m/s, l₂=(−2.18±0.16)·10³ m/s, t₀=11.4±0.7 days.

Observation by means of a stereoscope shows a homogeneous mixture in which wood shavings are well wrapped (Figure 10).

Figure 10:

Stereoscope images of mortar with (A) 70% by volume replacement of fine aggregates by ayous shavings and (B) 20% by volume replacement of fine aggregates by beech shavings.

4 Conclusions

Based on the results presented, the following conclusions can be drawn:

• Compressive and flexural strength decreases as the by volume replacement percentage of conventional aggregates increases, but the mix design could compensate this reduction in strength.

• The unit weight of fresh mortar containing wood shavings decreases as wood shavings content increases.

• As the mix proportion of cement increases when wood shavings are used as a by volume replacement of conventional fine aggregates, the cost of the mixture should be carefully monitored.

• It is deduced that a sigmoidal curve (model) fits very well the results for compressive strength as a function of curing age.

• A sigmoidal curve not involving the type of wood shaving used as a fine aggregate replacement is a significant predictor of compressive strength. Specific to the geographical region of anyone willing to use this curve, a further specification of the curve parameters values could be done by repeating the same experimental procedure as within the present study, using the types of wood mostly used in industrial processes in the specific region. As further research, more studies could be performed for combined results regarding mechanical properties as well as durability or thermal properties of mortar containing wood shavings and replacement of conventional aggregates by mixtures of various wood types.