The materials used in this study, as well as their origin and composition, are presented in Table 1.

During mix design, mixing begins with a pre-wetting of dried sand in oven to obtain an initial water content of 3.5% (in order to better control the parameters during formulation) to have a homogeneous mixture. A rest period of 30 minutes is observed to ensure homogeneous water absorption. Parallel to this step, the mixture of emulsion, water and adhesion agent is carried out in a container. After the rest period, the cement is added to the sand before mixing manually for 40 seconds with the liquid part (emulsion, water and adhesion agent). For the manufacture of recycled micro-surfacing, a mix design with 50% RAP has been determined, and glass fibers were added to improve the cohesion between recycled binder and aggregates.

In order to validate the formulation of the micro-surfacing, several trials have been carried out, and the mixtures were evaluated with the cohesion test, the abrasion test, the mini rutting test and the Hilt Cohesion Test [8] [9] .

For the Hot mix asphalt, a classical 10 mm surface course mix is used in this study, a standard bitumen of PG 70 - 28 was used. The maximum density of the mix is 2.903 with a binder content of 5.1.

The rutting phenomenon is influenced by several parameters: temperature, the duration of exposure to stresses, and the binder film thickness [10] .

In this study, 50 mm thick HMA slabs were compacted with the LPC slab compactor to obtain an air voids content of 5%. The HMA were subjected to rutting cycles before being rehabilitated with micro-surfacing in order to simulate site’s conditions. Once the micro-surfacing was installed, the rehabilitated HMA slabs were fast cured in an oven for 24 hours at 50˚C, in order to evacuate water and volatile fractions. Afterwards, the slabs are left at room temperature to stabilize, and they are weighted to measure the loss of mass (loss of water). Generally, a constant mass is reached after the first day.

After this step, the rehabilitated slabs were placed in a draft oven at 85˚C, for a period ranging from 1 to 5 days in order to simulate the evolution of the thermal aging. The thermal aging protocol was inspired by work carried out in the IRC laboratory [9] . The curing and aging temperature were determined on the basis of previous publications [11] .

The rutting test with the LPC rut tester (Figure 1) is a simulation test which consists in applying a 5 kN load with a wheel inflated to 0.6 MPa on the sample at 60˚C, for 30,000 cycles at a frequency of 1 Hz [10] .

Repeated wheel passage leads to the formation of a rut whose depth is the average of the rut depth measured at 15 points on the surface. During the tests, three types of structures were tested: the first one was HMA plate alone. The second one is the HMA slabs rehabilitated by a layer of micro-surfacing formulated with virgin aggregates, and the third one is the HMA rehabilitated with a layer of micro-surfacing formulated with RAP.

The complex (HMA and Micro-surfacing) plate has a total thickness of 50 mm with a micro-surfacing thickness of about 17 mm ± 1 mm. The small variations

in the thicknesses of the micro-surfacing are due to the differences in the rut depths of the HMA slabs before rehabilitation. In order to have 50 mm thick slabs, the HMA rutted slab were saw in their thicknesses to around 33 mm before being covered with micro-surfacing.

The evolution of the rut depths are monitored and measured as a percentage of the initial thickness of the material in function of the number of loading cycles. Initially, rutting tests were carried out on three series of HMA slabs (Table 2). All this slabs of HMA were used as reference. Rutting tests were carried out on these slabs at 60˚C.

The second part of tests consisted of casting micro-surfacing on the hot-mix asphalt slabs and then carrying out the same process of curing/aging on the complex (HMA + virgin micro-surfacing) (Figure 2). Then, the rutting tests were carried out on these at 60˚C (Figure 2). The same process is applied for the third type of structure HMA + recycled micro-surfacing.

As expected, the analysis of the impact of aging on the rutting resistance of the asphalt mixes shown in Figure 3 shows that the longer the aging, the higher the rutting resistance is. Indeed, the aging carried out at 85˚C did results in an accelerated oxidation of the binder. These phenomena are the same that are observed in the PAV and RTFOT aging processes [10] . However, there is no great difference between the resistance to rutting of the hot mix asphalt plates aged for 2 days and 5 days.

The rut depth results are shown in Figure 4 (slabs rehabilitated with virgin micro-surfacing) and Figure 5 (slabs rehabilitated with recycled micro-surfacing). The first thing to note on both figures is that from 0 to 30,000 cycles, only the HMA slabs are tested. At 30,000 cycles, the slabs are rehabilitated, which is why the rut depth goes down to zero.

As it can be seen from Figure 4 and Figure 5, the rutting resistance of ESG-10 mixes without micro-surfacing increase with the curing and aging times. The same trend is observed when micro-surfacing is added on top of the ESG-10 mixes. Also, by addition of micro-surfacing, the rutting resistance of ESG-10 mixes was improved. Figure 5 also shows that the rutting resistance of ESG-10 mixes is further improved with the addition of micro-surfacing mixes which consists of 50% RAP as the aggregates in micro-surfacing mixes.

We believe that the rutting resistance is improved due to the existence of aged bitumen in both ESG-10 and micro-surfacing mixes. Also, the rutting resistance is further improved by using the 50% RAP as the aggregates in the micro-surfacing. This can be explained by the existence of high stiffness (very low penetration) of the RAP binder.

The complex structure HMA/virgin micro-surfacing and HMA/recycled micro-surfacing have a deformation memory due to the initial rutting of the asphalt plate. Hence, a higher resistance to rutting was observed for the complex structure HMA/virgin micro-surfacing and HMA/recycled micro-surfacing. Initially, the micro-surfacing fills the initial ruts of the slabs, then it brings a reinforcement of 15-mm thick allowing the whole structure to perform better. It should be noted however that it is necessary that micro-surfacing, conforming to the specifications of the ISSA, do not undergo a heat wave before being well stabilized, since it would results in a structure highly sensitive to rutting.

All tested slabs in this study meet the requirements of NF EN 13108-1 standard since all the complex HMA/virgin micro-surfacing have rutting percentages of less than 10% to 30,000 cycles.

Statistically, the validation of the model will be carried out through the adjusted coefficient of determination R a 2 and the ratio between the standard error S_e and the standard deviation S_y. These parameters are defined by Equations (1) (2) and (3):

S e = ∑ ( Y − Y ^ ) 2 x − k (1)

S y = ∑ ( Y − Y ¯ ) 2 x − 1 (2)

where x is sample size; k is the number of independent variables in the model; Y: values tested; Y ^ : predicted values and Y ¯ is the mean value of the measured rut depth. The lowest the value of S_e/S_y, the greater the prediction [12] [13] . The coefficient of determination R² is defined by Equation (3):

R a 2 = 1 − ( x − 1 ) × ( 1 − R 2 ) x − k − 1 (3)

Analysis of the correlation matrix presented in Table 3 shows that the percentage of virgin aggregate (sand), the percentage of cement and the curing time are not significant with respect to the percentage of rutting. Indeed, they have low correlation coefficients and P > 0.05. Thus, in the prediction of the percentage of rutting, these parameters will not be taken into account.

The prediction parameters considered are the number of cycles, the percentage of added water, the percentage of bituminous emulsion, the percentage of adhesion agent, filler, and fiber. Analysis of the linear regression showed that the statistical calculation can’t estimate the percentage of rutting according to the percentage of bitumen emulsion, adhesion agent, RAP, fiber and filler. Those parameters are not considerate by the prediction model. Only three variables are selected, including the number of cycles, the rutting state before rehabilitation and the percentage of added water. Hence the development of the model (Equation (4)):

Rutting ( % ) = − 1 .10 − ( 0 .000024 × Numberofcycle )     + ( 0 .420 × Water ( % ) ) + ( 0 .991 × HMArutting ( % ) ) (4)

Figure 6 shows a strong correlation between the predicted rutting values and the measured rutting values measured with the LPC rut tester. Indeed, a strong

adjusted coefficient of determination of R² = 0.65 is obtained with an average dispersion of S_e/S_y = 0.73.

The Analysis of the variance (Table 4) showed that the rutting state before rehabilitation and the amount of added water are the most significant parameters in the rutting prediction model. The analysis of the variance also shows that the variable rutting state of the pavement has a p-value more significant than the

variable percentage of added water. Thus, the parameter “added water” influences less the behaviour of the model. The cycle number is the least significant parameter of the model, its p-value is not significant. However, it remains an important parameter in the prediction of rutting.