The experimental set-up ( Figure 1 ) was built in the workshops of the Department of Mechanics and Energy of the Institut National Polytechnique Félix Houphouët Boigny in Yamoussoukro, the capital city of Côte d’Ivoire, situated in Sub-Saharan Africa between 5˚ and 11˚ north. In order to improve the thermal performance of the sheet metal roof, two test buildings are constructed side by side with different roof configurations ( Figure 2 ) without attic. Roof 1 is single corrugated roof and Roof 2 is a corrugated heat recovery roof.

Test building 1 has a simple corrugated metal roof (Roof 1) of dimensions (135 cm × 80 cm) installed on two rafters (6 cm × 5 cm). While test Building 2 is covered with a Corrugated heat recovery roof consisting of from top to bottom with a roof ( Figure 3 ) consisting of a corrugated metal sheet (135 cm × 80 cm), 12 mm thick of serpentine copper tube in which water is circulated, a sheet of aluminium foil acting as a heat reflector, with long wave emissivity of 0.03 and short wave absorptivity of 0.10, 4 cm thick of polystyrene panel (120 cm × 75 cm), 1.5 cm thick of plywood (130 cm × 78 cm) acting as the interior ceiling and two rafters (6 cm × 5 cm). The walls of the test buildings are constructed with 15 cm thick hollow cement bricks and covered with 1.5 cm thick cement plaster. The dimensions of the walls of the test buildings are as follows: length (100 cm), width (72 cm), upper height (105 cm), lower height (80 cm). In addition, windows (30 cm × 20 cm) were made on the walls opposite the doors. Both roofs are inclined at 10˚ to the horizontal. Hot water is produced by recovering some of the heat from under the sheet metal through the circulation of low-flow water in the serpentine copper tubes housed in the corrugations of the corrugated sheet metal. This water is continuously discharged into a barrel insulated with 3 cm thick polystyrene. Table 1 gives the different thermal characteristics of the different materials in the test buildings.

Two configurations are considered for the experiment. For configuration 1, test buildings 1 and 2 are without wall insulation with door closed and window closed. For configuration 2, the east and west walls of the test buildings are insulated with 3 cm thick polystyrene with door closed and window closed.

Exposed to solar radiation, the different external walls of the test buildings convert part of the received radiant energy into heat. This heat is transmitted to the internal walls of the test buildings by conduction in different ways depending on the thermal characteristics of the different materials making up the walls. Subsequently, the internal walls of the test buildings transfer the heat to the indoor air by convection and radiation. Compared to a roof covered with simple sheet metal, the combination of polystyrene, aluminium foil and tubes in which water circulates without a loop at a reduced flow rate under the corrugated sheet allows the heat recovered from the sheet metal to be transferred to the interior of the building. In this way, the solution of constructing the second roof in sheet metal reduces the heat inside the building in hot weather.

The experimental study consisted of measuring the total incident solar radiation received by the roofs of the test buildings and the temperatures of the different external and internal walls of the test buildings. In addition, data on the interior and exterior humidity of the test buildings and the inlet and outlet temperatures of the serpentine cooper tube. Moreover, inside and outlet water temperatures of the barrel were measured. The total incident solar radiation is measured using a KIPP and ZONEN pyranometer with a relative measurement error of ±2%. It is connected to a digital integrator, of the same brand, allowing the reading of the solar energy received instantaneously and of the daily irradiation. The pyranometer is placed horizontally to obtain the full solar radiation. Temperature measurements were carried out with temperature sensors (NTC sensors) with a precision of ±0.1˚C and a measuring range of [−50˚C, +110˚C]. All the temperature sensors underwent prior calibration in a metrology laboratory using a standard Testo 925 thermometer with a relative error on the measurements (k = 2) of ±0.16˚C. The absolute error on the temperature measurements is ±0.2˚C. The relative humidities were measured with thermo-hygrometer sensors type DHT22 with a temperature accuracy of ±0.5˚C with a measurement range of [−40˚C; +125˚C] and a relative humidity measurement accuracy of ±2% with a measurement range of [0%; 100%]. The flow rate of water in the serpentine tube is calculated every hour.

For the instant, automatic and continuous reading of the data every 2 minutes, a data acquisition card is made. The different probes and humidity sensors of each test building were mounted on an Arduino electronic board. The data read by the Arduino board is retrieved, sent and stored in a database by Wi-Fi on a computer installed 30 metres from the test buildings. By recovering the heat from the sheet metal, hot water was produced continuously from 7:30 a.m. to 6:30 p.m. The tubes in the roof were continuously supplied with low flow tap water.

Assessment of the thermal performance of the two types of corrugated roofs will consist on the one hand of the comparison of the temperatures of the external and internal faces of the roofs versus time ( Figure 4 ). The maximum and minimum temperatures reached by the inner and outer sides of the roofs are compared, as well as the hours at which these maximum and minimum values are reached. In addition, the decrement factor f and the time lag Φ of roof and walls are calculated. In fact, during the propagation of a heat wave through a building wall, the time difference between the time at which the outer wall reaches its maximum temperature and the time at which the inner wall reaches its maximum temperature is called time lag [24] - [26] . And the decreasing ratio of the amplitudes of the outer and inner roof or walls is called the decrement factor. These two parameters are calculated according to the formulas below. Indeed, a lower decrement factor with a larger time lag indicates that the roof has a good thermal performance.

$f=\frac{T_{i,\max }-T_{i,\min }}{T_{e,\max }-T_{e,\min }}$ (1)

$\Phi =t_{T_{i,\max }}-t_{T_{e,\max }}$ (2)

where $T_{i,\max }$is the maximum temperature of the inner roof or wall side of test building and $T_{i,\min }$is the minimum temperature of the inner roof or wall side of test building and t refers to the time. $T_{e,\max }$is the maximum temperature of the exterior roof or wall side of test building and $T_{e,\min }$is the minimum temperature of the exterior roof or wall side of test building and t refers to the time.

The measurements of this experimental study were carried out under sunny weather conditions (configuration 1) and under cloudy weather conditions (configuration 2) in Yamoussoukro at latitude 6.58˚N. During the sunny day, instantaneous solar incident flux reached a maximum of 943 W/m² at 11:30 a.m. with a daily irradiance of 5705 Wh/m² and during the cloudy day, instantaneous solar incident flux reached a maximum of 1024 W/m² at 12:00 a.m. with a daily irradiance of 4730 Wh/m².

The maximum and minimum temperatures recorded for the different roofs and different walls sides of the buildings in configurations 1 and 2 are presented in Table 2 and Table 3 , respectively. We notice that the temperature curves of the inner side of the proposed roof (roof 2) as a function of time for both configurations are smooth compared to those of the roof constructed with a single sheet of metal (roof 1) which records very variable temperatures as a function of time due to its high thermal conductivity. This means that the proposed roof is less sensitive to variations in external weather conditions compared to the roof constructed with single sheet metal.

Figure 7 and Figure 8 show that the inside face of the roof 1 for the two experimental configurations is significantly warmer throughout the day with differences of more than on average above 10˚C from 10:00 a.m. to 2:00 p.m. and slightly cooler at night than the other internal walls of the building. This means that the roof 1 transmits heat throughout the day by radiation to all other internal walls. Indeed, according to Cardoso et al. [25] , in an enclosure, the warmest walls transmit heat by radiation to the lowest walls. Thus, the roof would be primarily responsible for the heat gain inside the SCR test building throughout the day as indicated in the studies of Rawat et al. [5] and Kabore et al. [1] . However, Figure 9 and Figure 10 show that the temperature of the inside face of the roof 2 for the two experimental configurations is lower than the temperatures of the inner walls of the door, window and east wall during the day. Moreover, it is higher than the temperatures of the inner walls of the west and north walls with maximum deviations of less than 3˚C and higher than the floor temperature with a maximum deviation of 6.2˚C. At night, the inside walls of the west, north walls, and the floor become warmer than the internal face of the roof 2, while the internal walls of the door, window and east wall are less warm with deviations of less than 2˚C. This means that roof 2 for the two experimental configurations receives radiant heat from part of the internal walls of the test building during the day and night unlike the case of the roof 1 with transmits radiant heat to the other internal walls all day. Therefore, the walls building 2 contribute to the increase of the indoor air temperature of the building with roof 2. Apart roofs, the internal and external faces of the walls, doors and windows of the two test buildings without east and west wall insulation are practically identical. In the case of walls insulation, the external faces of the walls, doors and windows have practically the same temperatures. However, the internal faces of the insulated walls of the test building with the proposed roof record relatively low temperatures compared with those of the insulated internal walls of the test building with the single sheet roof. This could be due to the heat gain from the inner surfaces of test building 1 through the roof, as opposed to test building 2.

On the other hand, a maximum difference of only 1.1˚C and 0.9˚C ( Figure 5 and Figure 6 ) between the temperature of the inside face of roof 2 and the air temperature in the center of the test building at the warmest hours respectively for test configuration 1 and test configuration 2 is recorded. Furthermore, a maximum difference of only 6.2˚C and 2.8˚C ( Figure 9 and Figure 10 ) between the temperature of the inside face of roof 2 and floor temperature at the warmest hours respectively for test configuration 1 and test configuration 2 is recorded.

Moreover, a maximum difference of 13.14˚C and 13.9˚C ( Figure 5 and Figure 6 ) between the temperature of the inside face of roof 1 and the air temperature in the center of the test building at the warmest hours respectively for test configuration 1 and test configuration 2 is recorded. Furthermore , a maximum difference of 19.9˚C and 16.8˚C ( Figure 7 and Figure 8 ) between the temperature of the inside face of roof 1 and floor temperature at the warmest hours respectively for test configuration 1 and test configuration 2 is recorded. So, the use of roof 2 (proposed roof) creates a pleasant atmosphere inside the building unlike the roof made of simple sheet metal. In Fact, the large vertical variations in temperature from the inner roof surface to the floor within building test 1 imply a more accelerated and larger heat exchange process between the roof and the indoor environment [25] . This causes discomfort due to high radiant heat exchange regardless of the physical environment.

The values of the time lag and decrement factor of the different roofs for the two considered configurations are summarised in Table 4 . The calculated values show that for the considered configurations the proposed roof has a time lag of 3.5 hours with relatively low decrement factors of 0.51 and 0.25 for configurations 1 and 2 respectively. On the other hand, the roof constructed with a simple metal sheet records a zero time lag when the east and west walls of the building are insulated and when the walls are not insulated. In addition, the insulation of the east and west walls results in smaller decrement factors. The values of the time lag and decrement factors for roof 2 show that the proposed roofing system can significantly delay the transmission of heat received through the roof to the interior of the building. It is therefore suitable for hot climates where climatic variations are often critical.

Figure 11 and Figure 12 show that the relative humidity within the RCR test building is almost uniform compared to the relative humidity within the SCR test building. However, the relative humidity in the RCR test building remains the highest compared to the SCR test building throughout the experiments as reported in the study by Ponni et al. [20] .

During the day, an average amount of about 60 L of hot water was produced and discharged regularly into the barrel from 7:00 a.m. to 6:30 p.m. Following the heating of the water in the 30 meters long supply pipe upstream of the roof, the water temperature at the collector inlet varied from 23.42˚C to 35.52˚C, the temperature of the water leaving the collector after heat recovery varied from 23.46˚C to 42.5˚C and the tank water varied form 25.89˚C to 39.35˚C in sunny weather day ( Figure 13 ). The water temperature at the collector inlet varied from 24.15˚C to 35.9˚C and the temperature of the water leaving the collector after heat recovery varied from 24.55˚C to 38.88˚C and the tank water varied from 26.80˚C to 38.53˚C in cloudy weather day ( Figure 14 ).