Along with the development of new technology, pavement evaluation and design techniques began to move into the era of a Mechanistic-Empirical approach. The Empirical method refers to the results of laboratory or experimental observations, as well as engineering expertise, or a combination of both, represented by equations or formulas to describe the natural ambiance and the correlations between design inputs, including loads, materials, layer configurations, and environment, and pavement failure [1-6]. In contrast, the Mechanistic approach is performed by modeling the pavement as a multilayer structure that responds to repeated loads, aiming to calculate the critical stresses and strains in the pavement [7]. Fundamentally, the Mechanistic-Empirical is a combination of those two methodologies. Some guidelines have currently applied this approach to evaluating and designing flexible pavements; one of them is the Mechanistic-Empirical Pavement Design Guide (MEPDG) developed by the American Association of State Highway and Transportation Officials (AASHTO). The MEPDG establishes a direct tie between pavement distresses and numerous design inputs [8].

The 2015 MEPDG is substantially a road pavement design module that enhances the previous method, the 1993 AASHTO, which uses an integrated analysis technique to estimate pavement deterioration over time [9]. In addition, this guide considers a variety of input parameters that affect pavement performance, such as traffic, climate, pavement structure, and material qualities, and the application of engineering mechanics concepts to forecast essential pavement responses [10]. Hence, several adjustments are required regarding the Road Damage model, including local and global calibration variables. These allow the adaptation to different circumstances, considering each location has its local calibration that can produce a more accurate MEPDG outcome [11]. For example, one of the American states that executed the 2015 MEPDG methodology is Arizona, located in the southern part of America. In fact, a city in Arizona, Phoenix, has a freezing index of 0 (zero), the same as Indonesia [12, 13].

Eventually, applying the MEPDG procedure by calibrating the Indonesia conditions is a fascinating topic to be discussed, especially regarding the damage model used. Therefore, this research aims to evaluate the damage model of the 2015 MEPDG and analyze the thickness of the flexible pavement overlay. Subsequently, the process that should be complied with in adopting the 2015 MEPDG technique to Indonesian cases is also proposed.

The remainder of the paper is structured in the following manner. Section 2 describes the methodology and defines the object of the study. Section 3 contains data analysis, including the existing pavement, traffic performance, climate, and road deflection. Section 4 discusses the mechanistic analysis using the KENPAVE software and the variations of model parameters. Section 5 introduces the 2015 MEPDG design and damage model, as well as the suggested performance criteria. Section 6 elaborates on damage models, such as performance deformation, fatigue cracking, and the International Roughness Index (IRI) model. Section 7 describes the 2015 MEPDG approach to Indonesian conditions and a discussion in Section 8. Finally, conclusions and some recommendations for applying the Mechanistic-Empirical Pavement Design Guide in the case of Indonesia are given in Section 9.

There were four main stages conducted in this research. First, secondary data was collected from the National Road Implementation Center (under the Indonesian Ministry of Public Works and Housing) as well as from the Indonesian Meteorology, Climatology, and Geophysics Agency. These data include the existing pavement thickness, traffic volume, climate, and Falling Weight Deflectometer (FWD) deflection data. Furthermore, this first step was also undertaken to analyze the current state and the back-calculation process using the BAKFAA software by the Federal Aviation Administration (FAA). This program is deemed very convenient and user-friendly; also, the desired modulus value can be locked to determine the Root Mean Square (RMS) of the typical one [14]. In general, the back-calculation procedures are as follows:

1. Input the measured deflection of the FWD device, including the load and distance from each detector.

2. Input layer thickness and properties.

3. Input seed modulus, which is used to determine surface deflection in software. Typically, this modulus is computed based on user experience or other formulae.

4. Software-based computation of deflection.

5. Examine the errors resulting from a comparison of the measured and estimated basins.

6. New modulus search. Some back-calculation software has employed various approaches to compute the new modulus of the existing error, resulting in an acceptable difference between the measured and calculated deflection basins.

7. Control the modulus range. In certain back-calculation software, the modulus range (minimum and maximum) is set or calculated to prevent the algorithm from convergent to an unreasonably large modulus.

Second, a mechanistic analysis was conducted using the KENPAVE software by inputting the results from the previous step. This structural condition analysis aims to receive the structural responses from several trials of overlay thicknesses, including stress, strain, and deformation that occur in the flexible pavement structure at the critical points [15, 16]. Fundamentally, KENPAVE is a finite element analysis program developed by Y. H. Huang (1993) [17], which is utilized for designing either flexible or rigid pavement and analyzing the structural response of a pavement model.

The third stage consisted of calculating the damage model according to the thickness of the overlay to obtain the total damage based on the Arizona calibration on the MEPDG approach, with varying overlay thickness. It also produced some additional layers that meet the damage criteria in the 2015 MEPDG damage model. Finally, some conclusions were drawn as consideration for offering recommendations and modifications to the design results, along with some valuable information for further research improvements.

The case of this study is the National Road segment of Majalengka/Cirebon Regency (Sta 8+100) to Palimanan City (Sta 20+400). This road is 20.15 kilometers long, with an undivided two-lane configuration in each direction, a pavement width of 7.00 meters, and an average shoulder width of 1.00 meters. As indicated in Fig. (1) [18], this road section is located in an urban region, with an industrial area connected to Indonesian National Road Route 1 (North Coast) and adjacent to the Cikopo – Palimanan toll road. Heavy vehicles on these highways are typically loaded with cement, sand, and industrial materials.

Fig. (1). Map of the study area [18].

KENPAVE software was utilized to perform the mechanistic analysis, with the input values derived from the initial step. The developed model is a flexible pavement with material considered linear elastic. This model uses variations of overlay thicknesses: 50 mm, 80 mm, 100 mm, 130 mm, 150 mm, and 180 mm. With the single-axle dual-wheel scheme, the axle load of 80 kN is distributed over the four wheels so that each wheel receives a load of 20 kN. Another parameter determined in the model is a tire pressure of 750 kPa with a contact radius of 92.1 mm for each wheel, with a 165 mm distance between wheels and a 1,800 mm axle length. The critical point on the Y-axis is located directly beneath the load, in the middle of the two wheels (with a distance of 165/2 mm), and half of the center between the wheels (165/4 mm). On the Z-axis, the critical points are: at depth 0, the midpoint of each layer, the bottom of the asphalt layer, and the top of the subgrade layer. Table 4 shows the output strain from KENPAVE software for the implementation of the MEPDG damage model calculation.

In the Mechanistic-Empirical Pavement Design Guide, the empirical damage model comprises five types: Rut Depth, Load-Related Cracking, Non-Load Related Cracking-Transverse Cracking, Reflection Cracking on Hot-Mix Asphalt (HMA) Overlay, and Surface Smoothness [20]. Nonetheless, in the case of Indonesia, the sort of Non-Load Related Cracking- Transverse Cracking damage, which is caused by a freezing temperature factor below 0°C, is not relevant. As a consequence, only four empirical damage models were considered in this study. Other inputs, if necessary, will be adopted from the Indonesian specifications (AASHTO 2015 [21]) or the American standard, which will be utilized if none are available. The damage model calculation referring to the 2015 MEPDG methodology produced the damage value of each model (Table 5), which was then adjusted to the performance criteria. Whereas Table 6 shows the performance criteria recommended for arterial roads with 90% reliability in the 2015 MEPDG method.

An overlay of 130 mm is needed when applying the 2015 MEPDG method in order to prevent permanent deformation of the asphalt layer by 0.25 inches. Thus, the maximum value of the performance criteria of 0.24 inches can be fulfilled. Besides, the fatigue and crack damage analysis resulted in an overlay thickness of 180 mm. While at 150 mm thick, the output was nearly similar to the load requirements. In general, the longitudinal crack damage of the 2015 MEPDG was more dominant than alligator cracking due to the overlay thickness.

However, the results of structural damage analysis using the MEPDG method, as represented by the International Roughness Index (IRI), which represents quantitative longitudinal surface-bump data of road surfaces [ 22 ], remain dubious due to the relatively low value produced over a ten-year period. This is due to the fact that Arizona’s local calibration factor only reaches 1.2281, which is very small compared to the global calibration used in America, which reaches 40.

The adaptation of the 2015 MEPDG to the Indonesian settings presently still requires further exploration. Consequently, thorough traffic data and material properties data are mandatory to support the utilization of this method. Similarly, the global and local calibration models need vast data to generate a damage model suitable for national conditions. A flowchart of the process of implementing the 2015 MEPDG method in Indonesia is described in Fig. (2).

7.2. Damage Model Adjustment

To be applied in Indonesia, the empirical model and performance criteria of the 2015 MEPDG should be adapted to local circumstances. Hence, the first step is to modify the input of the empirical model in order to match the units used in Indonesia before calibrating the formulas and adjusting the pavement damage that exists in the area under review. Furthermore, as winter does not occur in Indonesia, the variables related to freezing temperature were omitted. Some formulas are described as follows:

Fig. (2). Flowchart of MEPDG (2015) implementation to Indonesia.

• Asphalt permanent deformation model:

(2)

With:

∆_p(HMA): permanent deformation on HMA layer (mm)

Ɛ_r(HMA): resilient or elastic strain at the center of HMA

n: number of load repetitions

T: pavement temperature (°C)

kz: depth correction, which is formulated as follows:

(3)

(4)

(5)

With:

D: depth below the surface (mm)

H_HMA: total HMA thickness (mm)

• Soil Permanent Deformation Model:

(6)

With:

∆p(soil): permanent deformation of the soil layer (mm)

n: number of load repetitions

ε₀: the intercept of permanent deformation repetitive load

tests in the laboratory (in/in)

ε_r: resilient strain of the laboratory test to determine properties ε₀, β, ρ (in/in)

ε_v: average vertical strain in the layer

h_soil: soil layer thickness (mm)

(7)

(8)

(9)

(10)

With:

W_c: water content (%)

M_r: resilient modulus (psi)

a_1,9: regression constraints: a₁ = 0.15 dan a₉ = 20.0

b_1,9: regression constraints: b₁ = 0.0 dan b₉ = 0.0

• Fatigue cracking model:

(11)

with:

N_f-HMA: a number of load repetitions

ε_t: tensile strain at a critical location

E_HMA: dynamic modulus HMA (psi)

Where:

(12)

(13)

with:

V_be: effective asphalt content by volume (%)

V_a: air void percentage in the HMA mixture (%)

C_h: thickness correction (depending on the crack type)

• C_H of Alligator Cracking:

(14)

• C_H of Longitudinal Cracking:

(15)

With:

H_HMA: a total of HMA thickness (mm)

• Alligator Cracking Area:

(16)

(17)

(18)

With:

FC_bottom: alligator crack area at the bottom of HMA (%)

Dl_bottom: cumulative damage index at the bottom of HMA

• Longitudinal Cracking Area:

(19)

with:

FC_top: the length of the longitudinal crack at the top of HMA (m/km)

Dl_top: cumulative damage index at the top of HMA

The alligator cracking width in the MEPDG model is 12 ft (3.67 m) with a length of 500 ft (152.40 m). When applied to Indonesia, the area's dimension refer to the national road standards, which is 3.6 m (11,81 ft) wide and 150 m (492.13 ft) long. This measure results in a total area of 5,812.5 ft². On the other side, the length of the longitudinal cracking considered in the MEPDG is 500 ft (152.40 m). Thus, the size reviewed for the Indonesian case can be converted to 150 m (492.13 ft) long.

• The IRI model:

(20)

with:

IRI₀: IRI after overlay (m/km)

SF: site factor

FCtotal: fatigue cracking area (alligator, longitudinal, reflection) (%)

RD: average rut depth (mm)

(21)

with:

Age: pavement life (year)

PI: soil index plastic percentage

R: average annual rainfall (mm)

P₀₂: percent passing 0.02 mm

P₂₀₀: percent passing 0.075 mm

As a whole, the implementation of the 2015 MEPDG method in Indonesia still needs much more research, testing, and surveys, as well as many suitable kinds of equipment. Consequently, the budget will be affected. The simplification of the MEPDG method can be done by simplifying the mechanistic analysis process, which requires traffic, climate and various material properties. Although these data support the level of accuracy of the MEPDG method, but in order to produce output in the form of a specific level of damage which is, of course, very useful in road maintenance in Indonesia, simplification must be used. Modeling the material as linear elastic is one way to simplify the mechanistic analysis process. Although the reliability of the output will be reduced, this strategy may lower the number of input variables on the material properties. It is also possible to minimize the diversity of traffic data and the level of detail in the daily climatic data as a kind of simplification. Finally, the accuracy of the results may be enhanced by calibrating the original empirical model to Indonesian settings.

Overall, the procedure of Arizona calibration is applicable as the foundation for the execution of this method in Indonesia, which indicates that the general methodology of MEPDG may be developed in Indonesia by calibrating the original Arizona model to specific regions in Indonesia.

The results of the calculation of overlay thickness using the damage model in the 2015 MEPDG Arizona calibration resulted in an overlay thickness of 180 mm, which has a control limit of 268.789 m/km performance criteria for longitudinal cracks. The same thickness is also obtained when using the method from Indonesia (Pavement Design Manual 2017). From this, it means that the Arizona calibration can be used as the basis for the use of the MEPDG method in Indonesia. Analysis of the functional condition of the road from the calculation of the IRI value method MEPDG 2015 Arizona calibration is still questionable because it produces small damage at the end of the design life. This is because the value of the calibration factor used by Arizona is quite small in IRI models, such as at the proper rut depth calibration factor 1.2281, while at global calibration used a value of 40.

The findings of this article can be used to guide further research and exploration of MEPDG implementation in Indonesia. The calibration of the empirical model with the Arizona calibration values, which is the basis for application to road projects in Indonesia, is essentially the starting point for this strategy. Therefore, to produce a procedure with a high level of reliability, a more extensive analysis of the variables is required, as well as a more robust and diversified data set, both in terms of material properties, traffic and axle loads, as well as climate data. However, it can also be done simplification of the method, which in this research is made in a flow chart for implementation in Indonesia.

The future direction of this research topic would be to calibrate and validate the model against the local parameters to improve the accuracy of the pavement predictions. Additionally, a sensitivity analysis should also be performed to identify the MEPDG design inputs for flexible pavements. In the end, the outcomes of the pavement design using the 2015 MEPDG can be compared to the 2017 Road Pavement Design Manual, which has so far been applied in Indonesia. Moreover, the outputs of the empirical model resulting from the 2015 MEPDG will be more comprehensive when compared and assessed to some road damage data on the existing pavement in Indonesia.