Withdrawal resistance
The maximum nail withdrawal resistance loads were initially analyzed separating the data obtained by the radial and tangential faces. All nail models and diameters inserted into the tangential faces required higher loads to be withdrawn than those inserted into the radial faces, which can be explained by the higher number of dense parenchyma layers pierced by the nails [
25]. In this direction, there is greater interaction between the wood tissues and the nail shanks, resulting in greater withdrawal resistance and consequently greater damage to the wood´s surface [
17]. However, after applying a factorial ANOVA, no statistically significant difference was found between the mean results obtained for the radial and tangential faces (P value = 0.125). As this observation had already been made by other authors in similar experiments [
26,
27], it was decided to analyze the data jointly, as shown in
Table 2.
The scientific literature on nails indicates a positive correlation between diameter and withdrawal resistance [
7,
28,
29]. In this study, the increase in nail diameter generated an increase of 5%, 17% and 9.5% in withdrawal resistance for smooth, helical, and annular nails, respectively.
As expected, smooth-shank nails had the lowest withdrawal resistance values. In accordance with data found in literature [
19], helical nails showed a resistance increase of approximately 30 to 40% (according to nail diameter), while the annular nails presented the highest withdrawal resistance, approximately twice the values presented by smooth-shank nails of same diameter. This result was consistent with data presented by Rammer [
1] and Skulteti et al. [
12].
Despite the numerical difference for both factors, the Tukey test identified that the difference between the means of different diameter nails of the same model was not statistically significant at a 95% confidence. Therefore, the Tukey test defined three homogeneous subsets, exclusively based on the nail models.
The values determined for the coefficients of variation (C.V.) ranged between 15% and 27%. Despite being high, they are considered satisfactory for nail withdrawal tests as the manual insertion of the nails is subject to operator-caused variations. Nevertheless, the C.V. values presented lower variations than those obtained by Rammer et al. [
8] for the same three nail models: annular = 17% to 32%, helical = 12% to 41% and smooth = 22% to 48%.
In addition to the maximum load values, the behavior of the different types of nails tested can also be understood observing the curves formed by the “displacement x load” graphs (
Figure 2) and typical wood failure modes (
Figure 3).
According to Li et al. [
29], nail withdrawal causes various levels of stress at the interface between the threads and the wood components, leading to a combination of shear and tension of the fibers. Due to static friction, during the initial stage of the test, the smooth-shank nails present a linear behavior between displacement and load. After the maximum resistance is reached, the static friction is overcome, and the graph shows a sudden drop. From this moment on, the resistance of the connection depends exclusively on the dynamic friction, which declines rapidly until the nail is completely withdrawn. As the insertion of smooth-shank nails occurs exclusively by separating the wood fibers, their withdrawal occurs without causing major damage to the wood specimen (
Figure 3A).
Although helical nails tend to rotate around their axis during insertion, the rotating movement does not occur during withdrawal, and therefore, the threads increase the nail´s friction surface, ensuring prolonged resistance, which allows higher maximum loads than smooth nails [
11]. During the withdrawal tests, the helical nails also exbibit an initial linear behavior between displacement and load, but unlike the smooth nails, after reaching maximum load, they present a sudden drop and then a tendency to stabilize. This is caused by the presence of the threads, which still exert dynamic friction against the wood fibers, and consequently, the connection is still capable of resisting considerable load while the nail is withdrawn from the wood. As can be seen on
Figure 3B, the withdrawal of helical nails causes the tearing of small portions of wood fiber.
In general, annular nails exhibit initial linear elastic behavior, provided by the mechanical resistance of the wood fibers lodged between the threads. As the fibers begin to tear, the graphs assume a non-linear (inelastic) behavior until they reach the maximum load values. After the peak, the load decreases quickly, as the wood fibers are torn, pulled out and brought to the surface in the form of a column of fragments adhered to the nail shaft (
Figure 3C). From this moment on, resistance is only due to friction, being similar to the behavior shown by smooth nails [
11,
28].
Based on the wood density of
A. decandra, nail diameters and nail insertion depth, maximum loads of withdrawal resistance for the six model/diameter nail combinations were estimated using the equations found in the literature (
Table 3).
The AWC [
20] and Rammer [
8] equations overestimated the maximum withdrawal loads of the 2.8 mm and 3.5 mm smooth nails by approximately 30 and 60%, respectively, not being suitable. All the equations tested for helical nails showed results close to those obtained experimentally, especially the equation by Blass and Uibel [
19], which underestimated the maximum load by 6% for 2.8 mm nails and approximately 9% for 3.5 mm nails. To predict the maximum resistance load of annular nails, the most precise equation was that of Rammer et al. [
8] which overestimated approximately 3% and 18% for 2.8 mm and 3.5 mm nails, respectively. It is important to register that, for structural project calculus, it is more appropriate to use equations which underestimate the resistance values, increasing the project's safety margin, than the opposite. Therefore, Blass and Uibel [
19] can be considered the most suitable equation.