3.1. Wave Spectrum Validation
During wave propagation, in the transition zone from deep to shallow water, there are significant changes in water depth. Wave breaking happens in the transition zone. After breaking at the transition zone, the wave continues to propagate into the shallow water in front of the structure. The purpose of creating the transition zone is to force the wave to break many times, creating incident wave conditions in front of the structure with a similar wave energy spectrum to that nearshore of the Mekong Delta.
The changes in the wave spectrum through the transition zone are shown in
Figure 7. From the deep-water zone (wave gauge WG1), the wave spectrum had a sharp peak with an energy value of 0.023 m
2/Hz (
Figure 7a), and through the transition zone, the spectral peak energy was significantly reduced to 0.011 m
2/Hz at wave gauge WG5. After the wave broke, the spectrum had transformed to a flattened shape with many low peaks (
Figure 7b). Thus, the energy decreased significantly due to the breaking of the wave.
The wave spectrum generated in the model after passing through the transition zone was also compared to the actual pattern observed in the field at Ganh Hao coast, Bac Lieu province. In the typical shallow water conditions of the Mekong Delta, the similarity of the wave spectrums between the field observations and physical model is shown in
Figure 8. The wave spectrums in both locations were flattened and scattered to multi-peaked values. In the case of the field observations, high-energy peaks were observed in the middle, while small peaks were observed on either side in the frequency bands. The wave spectrum from the physical model confirms that it has simulated the field wave conditions in order to increase the reliability of the study of waves’ interactions with the breakwater.
3.2. Experimental Analysis
Incident waves, water level, crest freeboard, and crest widths with a total of 280 experimental scenarios have been tested. A typical sketch of the experiment regarding the effects of wave propagation on the PRBW is shown in
Figure 9.
The structure of the PRBW is a porous media vertical seawall construction. The distance between the centrifugal piles and the porosity of the filling rocks affect the wave energy dissipation and reflection ability of the structure. According to the law of the conservation of energy, the energy can be expressed mathematically using the energy balance formula devised by Van der Meer and Daemen [
22]:
where
Ei,
Et,
Er, and
Ed are the energies of the incoming wave, the transmitted wave, the reflected wave, and the dissipated wave, respectively. From this, the energy balance function can be rewritten as:
where
—the wave transmission coefficient is determined by the ratio of the wave height behind the breakwater (Hm0,t) and the wave height in front of the breakwater (Hm0,i);
—the reflected wave coefficient is determined by the ratio of the reflected wave height in front of the breakwater (Hm0,r) and the wave height in front of the breakwater (Hm0,i);
Kd is determined based on the transformation from Formula (3):
The front wave disturbance coefficient in front of the breakwater is determined by the difference of the wave height at position WG5 with and without the structure:
3.2.1. The Effect of Crest Width on the Wave Transmission Coefficient
The PRBW operates under three states: the submerged (
Rc < 0), the transition (
Rc = 0), and the emerged states (
Rc > 0). The correlation between the relative crest freeboard (
Rc/Hm0,i) and the wave transmission coefficient (
Kt) through the PRBW is shown in
Figure 10. It can be seen that
Kt ranges from 0.2 to 0.7 and is inversely proportional to the crest width (
B) of the PRBW. However, the curve tends to converge when the breakwater operates under the submerged state (
Rc/Hm0,i << 0). This demonstrates that the effect of the crest width decreases when the PRBW operates under the submerged state. When the PRBW operates under the emerged state (
Rc/Hm0,i > 0), the wave transmission coefficient decreases gradually when increasing
Rc, and when
Rc/Hm0,i > 1.5, the reduction of
Kt is not significant. On the other hand, the crest width of the PRBW increases linearly with different widths from 24 cm to 38 cm and 52 cm, while the wave transmission coefficient decreases non-linearly, corresponding to the increase in crest width.
Figure 10 shows that the wave transmission coefficient is at its largest with
Kt = 0.4–0.75 for the crest width of
B = 24. Meanwhile, the wave transmission coefficients for
B = 38 cm and
B = 52 cm are similar, with
Kt = 0.2–0.65. This shows that when the crest width of the PRBW increases from
B = 24 cm to
B = 38 cm, the
Kt reduces significantly from
Kt = 0.4 to
Kt = 0.2, corresponding to the emerged state; when the crest width continues to increase
B ≥ 38 cm corresponding to the actual width of the PRBW ≥ 1.7 m, then the wave transmission coefficient changes insubstantially.
The crest widths affected the wave transmission coefficient (
Kt) in all three working states (
Figure 11), in which the wave transmission coefficient in the submerged state changes slightly with wide ranges of the ratio of
B/Hi (i.e., the blue curve has a small slope). The
B/Hi ratio is inversely proportional to the wave transmission coefficient. A higher ratio of
B/Hi corresponds to a lower transmission coefficient. In the emerging state, the wave transmission coefficient ranges from 0.2 to 0.45. In the transition state, the range of
Kt varies from 0.35 to 0.6, and for the submerged state,
Kt ranges from 0.55 to 0.75.
3.2.2. The Effect of Crest Width on the Wave Reflection Coefficient
Figure 12 shows the relationship between the relative crest freeboard (
Rc/Hm0,i) and wave reflection coefficient (
Kr) for different crest widths. The wave reflection coefficients range from 0.15 to 0.45 in all cases of crest width. The wave reflection coefficient is proportional to the relative crest freeboard. However, the wave reflection coefficient fluctuates inconsiderably from 0.15–0.25 when the PRBW operates under a submerged state (
Rc/Hm0,I < 0), but
Kr alters remarkably from 0.2–0.45 when the PRBW operates under emerged state (
Rc/Hm0,i > 0). When increasing the crest width from
B = 24 cm to
B = 38 cm and
B = 52 cm under the same conditions (i.e., the waves, water level, and type of rock used for the structure), the wave reflection coefficient between the three cases changes slightly. This demonstrates that the wave reflection coefficient of the PRBW is not significantly affected by the crest widths.
The combination of the reflected waves in front of the PRBW and the incoming waves through the process of interference and resonance creates a change in the wave propagation on the PRBW.
Figure 13 shows the increase in the spectral peak energy in front of the PRBW.
Kf is the front wave disturbance coefficient, which is the ratio of the wave height right in front of the breakwater to the incident wave height. The increase in spectral peak energy leads to a change in the wave height in front of the structure, which can be increased by 1.4 times compared to the case without the structure, as described in
Figure 14, which is consistent with the study of Le Xuan et al. [
12].
The incremental wave height in front of the PRBW is proportional to the relative crest freeboard (
Rc/Hm0,i) (
Figure 14). When the PRBW is still operating in the submerged state (
Rc/Hm0,i < 0), the front wave disturbance coefficient is approximately one and increases gradually when the PRBW operates under the transition state to the emerged state. The increment of the incoming wave disturbance coefficient (
Kf) is due to the reflected wave in front of the structure, as such, the
Kf is proportional to the wave reflection coefficient (
Figure 15).
3.2.3. The Effect of Crest Width on Energy Dissipation
For the PRBW structure, waves are dissipated by the process of passing through the piles of columns, the layer of rocks, and by friction with the rocks. The results show that the percentage of wave energy dissipated by the PRBW structure ranges from 40% to more than 80% of the incident wave energy. More wave energy is dissipated due to the larger the crest width; the maximum energy dissipation corresponds to the largest crest width (
B = 52 cm) (
Figure 16). In the submerged state, when the relative crest freeboard is less than 0, the percentage of energy dissipation gradually increases from 40% to 70% for the three crest widths. In the emerged state, when the relative crest freeboard is greater than 1, the dissipated wave energy tends to slightly increase and starts decreasing when
Rc/Hm0,i ≥ 1.5 for all cases. That means the dissipated energy only depends on the width and porosity of the PRBW.
Figure 17 shows the average percentage of wave dissipation energy with the working states of the PRBW (i.e., submerged, transition, and emerged states) with different crest widths. The average percentages of wave dissipation energy in the transition and emerged states are superior to that of the submerged state.
The largest wave energy dissipation under the emerged state is approximately 70.35–81.47%, and the lowest dissipation energy in the submerged state is about 53.27–65.39%. However, in the case of the submerged state, the wave dissipation energy also increases with the expansion of crest width, with average values of about 53.27%, 60.64%, and 65.39% corresponding to the widths B = 24, B = 38, and B = 52 cm, respectively. When the crest width of the PRBW increases by 14 cm, the wave dissipation energy of the structure increases by about 5% and obtains the maximum value of about 65.39%, corresponding to the maximum crest width under the submerged state.
3.3. Development of Experimental Equations for the Wave Transmission Coefficient
The analysis of the parameters affecting wave propagation is the basis for building empirical formulas. Following Buckingham’s Π-theorem, we have a wave transmission coefficient Kt (a dependent variable) that is completely determined by the values of a set of several independent variables including crest freeboard (Rc), incident wave height (Hm0,i), wave length (Lm), water depth (D), crest width (Bn), and the nominal stone diameter (Dn50).
Hence, we investigated the correlation between seven variables with one dependent variable and six independent variables, as shown in
Figure 18. Looking at the first row of
Figure 18, it can be seen that the crest freeboard (
Rc) and water depth (D) are strongly dependent on the wave transmission coefficient (
Kt), with correlations of −0.86 and 0.86, respectively. Accordingly, the crest width (
B) and wave length (
Lp) have correlations with
Kt of −0.35 and 0.37, respectively.
Rc and
B are inversely proportional to
Kt, while
D and Lp are proportional to
Kt. The wave height (
Hs) has less of an influence on the
Kt, with a correlation of −0.02 (
Figure 18). Therefore, we have performed a principal dimensionless variable analysis to generate the general formula as follows:
From (6), we have evaluated the factors affecting
Kt, including
Rc/Hm0,i,
D/L,
Bn/Dn50,
Hm0,i/Dn50, and D/L, but the contributions of these variables are different, as shown in
Figure 18. Therefore, we focused on the distribution of the dimensionless variables to build the empirical formula.
In order to cover a wide range of factors, we derive two empirical equations based on the multivariate regression analysis, including linear and non-linear equations:
where
a,
b,
c, and
d are the empirical constants characterizing the influence of the respective factors, namely, relative crest freeboard (
Rc/Hm0,i),
D/Lm, and (
B/Dn50) to
Kt.
a, b, c, and d are determined through the linear regression analysis method with the data obtained from the experimental results.
The values
a,
b,
c, and
d are defined based on the maximum correlation coefficient
R2. For this linear equation, we obtained experimental data with
R2 = 0.82, as shown in
Figure 19a. The defined constants are
a = −0.1619;
b = 0.4817,
c = −0.0161, and
d = 0.6284.
Substituting
a,
b, and
c into (7), we obtain an empirical formula that determines the wave propagation coefficient:
The range of application of the formula is:
For the non-linear equation, we have developed an equation with multiple independent variables, including crest width
B. The equation is as follows:
in which:
The values
a,
b,
c, and
d will be selected so that the correlation coefficient
R2 is highest (
R2 = 0.73), as shown in
Figure 19b, and the results of the constants are as follows:
a = 0.000363;
b = −0.088,
c = −2.641,
d = 1.161,
e = −0.033,
f = 3.019,
α = −3.70, and
β = −0.1.
The range of application of the formula is:
3.4. Comparison of Different Empirical Formulas for Pile–Rock Breakwaters
To evaluate the feasible application of different empirical formulas for pile–rock breakwaters and verify our proposed equations, we have applied two empirical formulas developed by d’Angremond et al. [
23] and Van Der Meer et al. [
24] for pile–rock breakwaters and compared the wave transmission coefficient (
Kt) obtained from our empirical formulas with the
Kt calculated from three formulas obtained from previous studies, including those of Le Xuan et al. [
12], Tuan and Luan [
25], and Tuan et al. [
26]. We applied these empirical equations without any changes or adjustments of the parameters or equations and calculated
Kt from these equations with our variables and wave conditions for the pile–rock breakwater used in this study. This was because Tuan et al. [
26] had built an empirical equation that applied the formula from [
23] for a perforated hollow breakwater to consider the surf similarity parameter or breaking wave Iribarren number
ξ0 but did not involve the crest width in the equation. Moreover, Tuan and Luan [
25] employed the equation from [
24] for bamboo fences in the Mekong delta. Le Xuan et al. [
12] applied both empirical formulas from Van Der Meer and Daemen [
22,
23] for pile–rock breakwaters but did not consider the variation of the crest width.
We have evaluated our proposed formulas, including our linear and non-linear equations, by comparing the
Kt from our study with those from [
23,
24] in terms of the statistical performance, including adjusted R-squared (
R2),
MSE, and
RMSE. From
Table 3, it can be seen that for the linear equations, the statistical indices are similar to those of the two previous formulas [
22,
23].
Figure 20 shows a comparison between the wave transmission coefficients (
Kt) against
Rc/Hm0,i for the case of the linear equation, non-linear equation, and the comparison with the
Kt calculated from [
23] and [
24]. It can be seen that there is good agreement between the
Kt from the empirical formula and experimental data shown in
Figure 20a,b. In addition, the trend from the linear and non-linear equations is consistent with the trend of
Kt obtained from d’Angremond et al. [
23] and Van Der Meer et al. [
24] (
Figure 20c,d), but it also has a slight difference in statistical parameters, as described in
Table 3. The best statistical performance was obtained from Van der Meer’s formula with an adjusted R-squared value,
R2 = 0.88,
MSE = 0.0028, and
RMSE = 0.053. Since this equation is in a general form, it can be applied to many types of breakwaters and explain 88% of the experimental data. The following linear equation’s statistical performance is
R2 = 0.82,
MSE = 0.0033, and
RMSE = 0.058. The next one is d’Angremond’s et al. [
23] formula with
R2 = 0.78,
MSE = 0.0041, and
RMSE = 0.064. This formula had the best fit applied for a pile–rock breakwater with a range of −0.5 <
Rc/Hm0,i < 0.3, and is not highly suitable for a range of −0.5 <
Rc/Hm0,i < 2.0. Therefore, d’Angremond’s formula, it can be applied for submerged breakwaters; however, this formula may not be used for the emerged breakwater. The non-linear equation yields a lower adjusted R-squared (
R2) value compared to the three equations mentioned above, with
R2 = 0.73,
MSE = 0.0044, and
RMSE = 0.066 (see
Table 3). However, the non-linear equation is valuable because it can fit well in a range of 0 <
Rc/Hm0,i < 2, while the linear equation is unable to fit this range corresponding to the breakwater’s working conditions in submerged or semi-submerged states (
Rc ≤ 0). This analysis confirms that our developed formulas are suitable for pile–rock breakwaters with linear and non-linear equations. This experiment also provides a data set for further studies that investigate different types of formulas for breakwaters.
In order to define the suitability of the previous formulas developed by Le Xuan et al. [
12], Tuan and Luan [
25], and Tuan et al. [
26] for PRBWs, we have calculated
Kt from these equations and evaluated the statistical indices as follows. Both formulas from [
25,
26] yield very low adjusted R-squared values and relatively high
MSE and
RMSE with values of
R2 = 0.04,
MSE = 0.0224,
RMSE = 0.150, and
R2 = 0.22, and
MSE = 0.0285 and
RMSE = 0.169, respectively (see
Table 3). This means that these formulas should not be applied for pile–rock breakwaters. Therefore, we also recommend not applying these equations for PRBWs. The
Kt obtained from the formula of Le Xuan et al. [
12] is highly similar to the experimental data in this study but not so high with respect to its
R2 = 0.61,
MSE = 0.0086, and
RMSE = 0.093 because of the same structure of the pile–rock breakwater. Nonetheless, Le Xuan et al.’s [
12] equation can explain 61% of the experimental data while our equations can explain from 73% to 82% of the experimental data.
Figure 21 shows a comparison of the correlation of the
Kt and relative crest freeboard and scatter plots between previous studies carried out by Le Xuan et al. [
12], Tuan and Luan [
25], and Tuan et al. [
26] compared to the current study. The
Kt from Tuan et al.’s study [
26] (
Figure 21a, left) decreases over time as the
Rc/Hm0,i increases, and these values are found to be larger than the measurement data as well as the values in the current study (
Figure 21a, right). The reason for this is that the porosity of the perforated hollow breakwater is very large, which leads to high wave energy transmission through the breakwater. In the study of Tuan and Luan [
25] that experimented with a bamboo fence permeable breakwater, the values of the
Kt also tend to decrease over time when
Rc/Hm0,i increases (
Figure 21b, left). However, most of the
Kt values are overestimated compared to this study and its measurement values (
Figure 21b, right). This is because the bamboo fence was built using bamboo poles and brushwood bundles, which have very low wave resistance. Therefore, their wave energy transmission capacity is very high.
Figure 21c shows that the value of
Kt is similar to those in Le Xuan et al. [
12] because the breakwater in this study has the same structural shape as that in [
12]. The centrifugal piles and rocks inside could significantly dissipate the wave energy compared to those in [
25,
26].