Role of conductive spreader layer in reducing surface temperature of HMA pavements

Abstract

Rutting is a common type of shear failure-related deformation in asphalt (hot mix asphalt, HMA) pavements. It occurs over time as a result of slow, repeated heavy loads from vehicles moving along the wheel path. This problem is most noticeable when the pavement is at a high temperature and its stiffness is the lowest. Previous investigations have concluded that flowing water in pipes embedded in the pavement leads to a reduction in surface temperature, and consequently rutting. However, the thermophysical properties of HMA limit the cooling effect to a small region immediately around the pipe. It is proposed that the area of cooling be enhanced by adding a highly conductive spreader layer below the pavement in conjunction with the pipe. A theoretical design optimisation has been carried out by exploring different aspects of the spreader layer–pipe spacing (W), depth of the pipe–spreader (D), spreader thickness (t _s), thermal conductivity (k _s) and variation in the boundary conditions. Finite element modelling predicts that a properly designed, highly conductive spreader layer will lead to a significant reduction in surface temperature with a minimal piping network leading to an extended functional life of the HMA pavement.

Keywords:

• sustainable infrastructure
• sustainable transport engineering
• life-cycle engineering
• rutting
• asphalt

Introduction

Prolonged daytime exposure to sunlight coupled with poor thermal conductivity and high heat capacity leads to a rapid rise in surface temperature of hot mix asphalt (HMA) pavements (Wahhab and Balghunaim 1994). The problem is exacerbated in warm climates with the pavement temperature reaching as high as 70°C in places like Houston, TX (NCHRP 2004). Being a viscoelastic material, HMA exhibits lower stiffness at higher temperature and at longer times of loading (slow speed; Monismith et al. 1994, Lu and Wright 2000). Consequently, investigations have shown that the amount of damage by rutting, a form of shear failure-related deformation in HMA pavements, is directly related to the temperature (Brown and Cross 1992, Bonaquist and Mogawer 1997, Pomerantz et al. 2000, White et al. 2002). It follows that, compared with a pavement with low temperature, a pavement with a higher temperature will require more maintenance and/or have a shorter life (in terms of rutting failure). One study (Mallick et al. 2009) found that the service life of a pavement with a maximum temperature of 53°C is more than double that of a pavement with a maximum temperature of 70°C; 20 years compared to 8 years using a mechanistic empirical pavement design guide/software (NCHRP 2004). Estimated costs for milling and resurfacing a rural road vary between $120,000 (FLDoT 2001) and $415,000 (MODoT 2010) per mile. Therefore, it would be beneficial to lower HMA pavement temperature in hot conditions, particularly in areas of heavy use.

Various strategies have been adopted to reduce the pavement surface temperature. Reflective coating has been a popular approach. Synnefa et al. (2009) showed that lighter pavement colours were able to reduce average surface temperature by 5°C compared with black asphalt. Wan and Ping (2009) tested road paint in Singapore designed to reflect infrared rays. Their experiment showed a reduction in pavement temperature from 65 to 38°C, a change of 27°C. However, this approach has the downside of reflecting the energy on adjoining buildings that may exacerbate urban heat island effects. In addition, changing the reflectivity and colour of the pavement surface can lead to problematic issues to driver vision. Paint coatings may also negatively affect pavement surface mix stiffness which can lead to surface cracks, wearing properties and skid resistance.

Kubo and Kido (2006) extended the study by comparing the surface temperature characteristics and rutting capabilities of various types of asphalt: water-retaining, porous and porous with reflective paint in Tokyo. Although their study concluded that there was no clear advantage for rutting resistance, they found that the water-retaining pavement achieved a surface temperature of up to 16°C less than conventional pavements, while the temperature of the ambient air dropped by roughly 1°C. Extracting heat from asphalt pavements by convection using flowing water through pipes has been developed by Ooms Avenhorn (the Road Energy System®). This has been used as a means of reducing pavement temperature during the summer (reducing rutting), then storing liquid in large underground systems and re-circulating during the winter to de-ice the road surface (Loomans et al. 2003, Srivastava 2009). Structural integrity of the pavement and lowering of stress concentration were achieved by modification of the asphalt binder and adding grids to hold pipes in place. Large-scale testing on roads has been implemented in parking lots, using the low-grade extracted heat with heat pumps for heating and cooling nearby buildings (van Bijsterveld 2000, van Bijsterveld and de Bondt 2002). Earlier studies carried out at Worcester Polytechnic Institute have shown the ability of water flowing through pipes below HMA pavements to reduce surface temperature by 13°C. However, poor thermal conductivity of the pavement required close pipe spacing or serpentine piping for lowering the surface temperature uniformly (Chen 2008, Chen et al. 2009).

Objective

Since the poor pavement thermal conductivity requires pipes to be closely spaced to cool a large pavement surface area uniformly, a theoretical alternative is explored in this study. A thermally conductive spreader layer (that would be placed in the same way as a geotextile layer) coupled to pipes with flowing water will increase the effective area of the pipes so that the pipes need not be placed directly below the wheel path. This study theoretically investigates the geometric and thermophysical property variations of the spreader and the placement of the pipe–spreader combination that allows for a practical solution with maximum pavement surface cooling.

Modelling and analysis

The basic concept of the pipe–spreader heat exchanger is shown in Figure 1. The goal is to place the spreader along the wheel path below the pavement to conduct the heat away from the most vulnerable region of the pavement (ruttingwise) to the pipes where the heat is removed convectively by a flowing fluid. The pipes would ideally be placed in the shoulder region to minimise stress concentration in the pavement. However, under certain conditions, the pipes may be placed in the median region of the pavement, or in other regions where the loading is less severe. Under all circumstances, the pipes would not fall directly below the wheel path.

Figure 1 Sketch of asphalt solar collector embedded in highway lane. Note: Pipes spaced away from the wheel path.

The design variables and their ranges are given in Table 1. They include pipe spacing (W) and depth (D); spreader properties such as thickness (t _s), conductivity (k _s) and spreader shape. In addition, the range of the two variable boundary conditions, solar flux absorbed (Q _s) and fluid temperature (T _f), are also specified. Table 2 includes the default material properties used. The pipe was assumed to be made of copper having a diameter of 6 mm.

Table 1 Design variables and their respective ranges.

Variables	Range
General dimensions
Pipe spacing (W)	0.1–2.0 m
Depth (D)	25–100 mm
Spreader layer properties
Conductivity (k s)	1.8–1000 W/m K
Thickness (t s)	0–12.8 mm
Shape	Sloped, lipped shapes
Boundary conditions/inputs
Absorbed solar flux (Q s)	500–1100 W/m^2
Cooling fluid (T f)	12–30°C

Table 2 Default material properties used, unless otherwise noted.

Material	Thermal conductivity (W/m K)	Heat capacity (J/kg K)	Density (kg/m^3)
Pavement	1.8	1050	2350
Spreader	160	900	2700
Note: All bodies assumed to have an initial temperature of 26.8°C.

Modelling was carried out using finite element technique, with the help of COMSOL™ (COMSOL 2009) software (general heat transfer module). The governing equations used by this software for this application are listed in Table 3. Each model was meshed using COMSOL's ‘extremely fine’ mesh setting (triangle elements) having a maximum element edge length of 0.0015 m. The grid size corresponds to the dimensions of the model; for example W = 40 cm represents a grid size of roughly 160 × 100. COMSOL's PARDISO time-dependent solver was used to simulate 8 h of heating, using adaptive time-stepping with a maximum increment of 25 s.

Table 3 Governing equations and boundary conditions used (unless otherwise noted).

Location and mechanism	Governing equation	Constants
Subdomain – pavement, spreader conduction	∂T/∂t = kΔT/ρC p	k – thermal conductivity
		ρ – density
		C p – heat capacity
		t – time
Top surface boundary condition convection with ambient air	q = h(T sur − T amb)	h – convection coefficient h = 6 W/m^2 K T amb  = 26.8°C
Top surface boundary condition radiation to ambient objects	q = ϵσ(T sur ^4 − T amb ^4)	ϵ – surface emissivity – ϵ = 0.9 σ – Stefan–Boltzmann const. – 5.67 × 10^− 8 J m^− 2 s^− 1 K^− 4
Top surface boundary condition absorbed solar flux	q = q s	q s = 700 W/m^2
Pipe wall boundary condition isothermal	T = T f	T f = 16.8°C
All other boundary conditions insulated	q = 0

Results

First, the effects of pipe and pipe–spreader combination in lowering the surface temperature are presented, followed by a detailed exploration of the geometric and thermophysical properties. In this section, temperature plots are first shown for the top surface and the cross-section of the pavement to describe the temperature distribution in a pavement under a design constraint. The plots are followed by graphs showing the maximum temperature (T _max – obtained somewhere on the surface) as a function of a particular parameter. Finally, the roles of various boundary conditions in modulating the pavement surface temperature are explored. Unless otherwise stated, all simulations are reported after 8 h of continuous heating, reflecting the worst-case scenario.

Comparison of temperature distribution between a conventional pavement, a pavement with a pipe and a pavement with pipe and spreader

Figure 2 clearly demonstrates the role of a pipe alone (2(b)) and a pipe–spreader combination (2(c)) when compared with a pavement alone (2(a)). A pipe is able to reduce the surface temperature significantly by over 20°C; however, that effect is confined to the vicinity of the pipe for about 5–10 cm (entire width of each section is 20 cm). The effect of cooling can be enhanced spatially in all directions by the addition of a spreader layer. The spreader layer also allows further lowering of the temperature. Figure 2, therefore, establishes the comparative benefit of a pipe alone and a pipe and spreader, respectively.

Figure 2 Top and side temperature distribution of (a) unmodified pavement, (b) pavement with pipe, (c) pavement with pipe and spreader layer. Note W = 40 cm, t = 8 h; Q _s = 700 W/m².

Effect of pipe spacing (no spreader)

First, the temperature distribution of the pavement was plotted as a function of pipe spacing in Figure 3. In general, the contour plots demonstrate that the maximum cooling is obtained around the pipe and the effect of cooling rapidly dissipates within 10 cm. Therefore, for larger spacing, a substantial part of the road remains hot. A spacing of 0.5 m (1.6 ft) and higher leads to minimal surface cooling and substantial hot spots. The maximum temperature on the pavement and within starts to drop only after the pipes are brought closer.

Figure 3 Top and side-cut temperature distribution with varied pipe spacing (W). Note: t = 8 h.

As summarised in Figure 4, the maximum surface temperature drops by 10°C or higher at W ≤ 0.5 m. The exponential part of the curve indicating significant lowering of maximum pavement temperature is prominent at W ≤ 0.5 m. Therefore, without a spreader, a significant cooling of pavement will require pipes to be spaced within 0.5 m.

Figure 4 Effects of pipe spacing (W) on maximum temperature (T _max). Note: t = 8 h, no spreader layer present.

Effect of pipe and spreader depth (D)

The simulations given in the previous sections were carried out for a constant pipe depth (D) of 25 mm (1 in.). However, it may be prudent to put the pipes deeper inside the pavement to avoid maintenance issues. Figure 5 shows the effect of increasing D with the spreader. As D increases, the pavement surface retains a higher temperature. At 75 mm (3 in.), the effect of the pipe–spreader system is largely diminished to an insignificant one. Furthermore, the spatial temperature distribution that one observes with cooler conditions near the pipe and warmer conditions further away also seems to be less. The maximum surface temperature increases over 20°C with an increase in D from 25 to 75 mm.

Figure 5 Top and side temperature distribution of pavement with spreader layer depth (a) D = 25 mm, (b) D = 50 mm, (c) D = 75 mm. Note: W = 40 cm, t = 8 h.

These effects are further summarised in Figure 6. As shown clearly, a pipe alone has no effect on the surface temperature when it is placed 25 mm or further below. This is because the effect of the pipe placement alone is minimal. The maximum temperature becomes sensitive to the placement depth only when the spreader layer is included. As one installs the pipe–spreader system deeper, the influence of cooling of the system diminishes at the surface of the pavement. For example, to maintain the maximum surface temperature below 40°C, the system should be placed not below 50 mm from the surface of the pavement.

Figure 6 Effects of pipe depth (D) on maximum temperature (T _max), with and without spreader layer (t _s = 6.4 mm). Note: W = 40 cm, t = 8 h.

Effect of spreader layer thickness (t_s)

Figure 7 shows the spatial variation of temperature with varying spreader layer thickness (t _s). A very thin spreader layer does not have the desired effect, as seen in part (a) with a noticeable hot spot. It does not effectively draw heat towards the pipe due to the limited conductive benefit that it provides. This has been confirmed experimentally (Chen 2008). The purpose of the spreader layer is to conduct heat towards the pipe; with too little material to achieve this purpose, the conduction flux becomes ‘bottlenecked’ and results in an increased temperature gradient across the spreader. A thicker layer of material does not encounter the same bottleneck. Therefore, the thickest spreader layer offers the best possible cooling effect, in terms of lowest temperature and largest area of coverage.

Figure 7 Top and side temperature distribution of pavement with spreader layer thickness, (a) t _s = 1.6 mm, (b) t _s = 3.2 mm, (c) t _s = 6.4 mm. Note: W = 40 cm, t = 8 h.

As a design optimisation, for various pipe spacing (W) and spreader layer thickness (t _s), the maximum temperature was predicted. Figure 8 represents the results of the simulation using an isothermal pipe of 17°C. At closer pipe spacing below 0.5 m, with or without the spreader layer, the temperature curves begin to converge. However, for pipe spacing >0.5 m the curve rises exponentially with increasing W. The effect of increased spreader thickness (t _s) is to lower the maximum surface temperature. This effect is more prominent above a W of 0.5 m. In addition, as the spreader becomes thicker, the maximum temperature curve becomes flatter. For widely spaced pipes, (W>1.5 m), t _s values need to be >12.8 mm (0.5 in.) to obtain substantial results. Therefore, an optimisation of pipe spacing and spreader thickness is required depending on the allowable temperature maxima of the pavement.

Figure 8 Effects of spreader layer thickness (t _s) on maximum temperature (T _max), varying pipe spacing (W). Note: t = 8 h.

Sloped and lipped spreader geometries

Although it is clear that increasing t _s will lead to further lowering of surface temperature and improvement in rutting resistance, it is impractical and unwieldy to put thick spreaders (exceeding an inch) below the pavement. However, if the pipes are placed near the edge of the road and in shoulder regions, it is possible to make the spreader thicker at the points immediately surrounding the pipe. Two designs, as shown in Figure 9, (a) a sloped design and (b) a lipped design, were investigated. The first design introduces a gentle slope that was varied from 6.4 to 9.6 mm. The lipped design allows a further curvature near the pipe for better heat transfer.

Figure 9 Sketch of (a) ‘sloped’ and (b) ‘lipped’ spreader layer designs.

The rationale for using sloped or lipped spreaders is demonstrated in Figure 10. Part (a) shows the bottlenecking that occurs with a 6.4 mm (1/4 in.) spreader layer. The maximum heat flux is at the pipe–spreader interface. Various factors come into play: the thickness and the conductivity of the spreader layer, as well as the pipe spacing, contribute significantly to the formation of the bottleneck. The cumulative effect of the surface heat flux causes a flux concentration near the pipe. The larger amount of heat passing through the spreader layer as it nears the pipe induces a temperature gradient across a shorter length (causing the area further from the pipe to heat up much faster). Part (b) correspondingly shows that a modest variation in spreader layer shape can easily reduce the effect of this flux bottleneck.

Figure 10 Heat flux distribution of pavement with spreader layer shape: (a) flat, t _s = 6.4 mm, (b) sloped, (c) lipped. Note: W = 40 cm, t = 8 h.

Figure 11 shows the effect of the sloped (c) and lipped (d) designs on temperature distribution and compares them to flat spreader layers (a) and (b), again using a pipe spacing of 40 cm. The sloped and lipped shapes do not drastically change the temperature variations along the length. The ‘Sloped’ design shows a reduction in maximum temperature by roughly 2.9°C whereas the ‘Lipped’ design shows a reduction of 4.1°C compared with the 6.4 mm spreader (Figure 11(a)), signifying that the modified designs have a benefit if varied geometry is a possibility.

Figure 11 Top and side temperature distribution of pavement with spreader layer shape: (a) flat, t _s = 6.4 mm, (b) flat, t _s = 9.6 mm, (c) sloped, (d) lipped. Note: W = 40 cm, t = 8 h.

Variation in thermal conductivity of spreader layer (k_s)

The thermal conductivity of the spreader layer (k _s) will affect the temperature distribution within the pavement. Since the spreader material will be made of a conductive geotextile, the sensitivity of temperature distribution was assessed while varying k _s. In this way, it would become easier to assess theoretically a range of materials as candidate for the conductive spreader material. As shown in Figure 12, increasing k _s by one order of magnitude from the pavement conductivity (part (a)) allowed a noticeable lowering of surface temperature, but the effect did not spread across the entire width as seen by the reddish hot spots away from the pipe. However, as shown in part (b), changes of two orders of magnitude in k _s (160 W/m K) clearly led to more uniform cooling of the pavement surface. Further enhancement of k _s led to minimal improvement in surface temperature distribution.

Figure 12 Top and side temperature distribution of pavement with spreader layer conductivity k _s  =  (a) 10 W/m K, (b) 160 W/m K, (c) 500 W/m K. Note: W = 40 cm, t = 8 h, t _s = 6.4 mm.

These effects are summarised in Figure 13 which plots T _max as a function of varying conductive spreader layer conductivity. In addition, k _sx and k _sy were varied to determine whether there was a directional dominance of conductivity in the system. Results indicate that the maximum surface temperature dropped exponentially with increasing k _s. Increasing k _s beyond 200 W/m K provided diminishing returns of temperature reduction, indicating that spreader conductivity ∼200 W/m K can give the most improved performance (given W and t _s). When the spreader conductivity was directionally analysed (as k _sx and k _sy ), it was found that the vertical conductivity (k _sy ) of the spreader layer is insignificant compared with the horizontal component (k _sx ) in regard to temperature reduction.

Figure 13 Effects of thermal conductivity of spreader layer (k _s) on maximum surface temperature (T _max), comparing x-component (k _sx ) to y-component (k _sy ) of anisotropic materials. Note: W = 40 cm, t = 8 h, t _s = 6.4 mm.

Effect of boundary conditions

Although the design of the pipe–spreader system (W, D, t _s, k _s) is the most crucial aspect in reducing the maximum pavement surface temperature and, therefore, preventing the rutting potential, boundary constraints play a practical role in modulating the temperature distribution in the pavement. So far, a constant solar flux Q _s = 700 W/m² and cooling fluid temperature T _f = 16.8°C have been assumed. Next, Q _s was varied from 500 to 1100 W/m² and T _f from 12 to 30°C.

Figure 14 shows the variation in T _max for W = 40 cm and various incoming solar radiation. Results are shown for no spreader (t _s = 0) and various t _s. The maximum pavement temperature expectantly increases with increasing Q _s. The temperature is highest with no spreader and decreases with increasing t _s. With increasing t _s, the increase in maximum pavement temperature was damped as illustrated by a progressively shallower slope of each curve.

Figure 14 Effects of absorbed solar flux (Q _s) on maximum surface temperature (T _max), varying spreader thickness (t _s). Note: W = 40 cm, t = 8 h.

Figure 15 shows the variation in T _max for W = 40 cm and various T _f. As the incoming T _f increases, the heat extracted from the pavement decreases and, therefore, the pavement surface temperature increases. Increasing t _s helps in lowering the maximum pavement temperature, but increases the sensitivity to T _f as indicated by the steeper slope of the curve for progressively increasing t _s.

Figure 15 Effect of coolant temperature (T _f) on maximum temperature (T _max), varying spreader layer thickness (t _s). Note: W = 40 cm, t = 8 h.

Daily cycles

Finally, a more practical model was constructed to include sinusoidal variations in solar flux and ambient temperature, more accurately representing the conditions occurring on the road surface. Both the solar flux and ambient temperature were assumed to behave as a half sine wave with a half-period of 14 h. The values started at an initial value at the beginning of the day (Q _s = 0 W/m² and T _amb = 26.8°C, respectively), rising to some peak value (Q _s = 1000 W/m² and T _amb = 36.8°C, respectively) halfway through the day and returning to the initial value at sunset. During this time, the fluid in the pipe was assumed to be flowing sufficiently to provide an isothermal pipe wall condition of 16.8°C. This ‘daytime’ period of 14 h was followed by a ‘nighttime’ period of 10 h in which the solar flux and ambient temperature were held constant (Q _s = 0 W/m² and T _amb = 26.8°C, respectively) and the fluid flow in the pipe was assumed to be off. Therefore, an insulated pipe wall condition was used. This simulation process was repeated over a total of 7 days.

A pipe half-spacing of W = 61 cm was chosen. This was selected as it is one-third of the width of a US interstate traffic lane (12 ft), such that the pipes can be placed away from the wheel path as shown in Figure 1. An unmodified control simulation, featuring pavement material only, was created. Another simulation was done with a pipe (D = 25.4 mm) below the surface to present the effects of the pipe itself. Finally, a model was created with a spreader layer (t _s = 6.4 mm, k _s = 1000 W/m K) and pipe (D = 25.4 mm) below the surface to show its full effect in comparison.

Figure 16 shows the variations in maximum surface temperature over the course of 7 days. Unmodified pavement clearly shows the highest maximum temperature, which increases over the first 3 days, representing the cumulative effects of energy storage in the material. The sample with the pipe shows a very small reduction in the maximum temperature of the pavement. Although the pipe itself provides a means of removing the energy from the pavement, the poor conductivity of the pavement material limits the effective pipe area. Regions near the pipe are kept cool, whereas the area far away from the pipe is barely affected and reaches a consistent high maximum temperature. The sample with the spreader layer shows a dramatic reduction in temperature. The high conductivity of the spreader layer bolsters heat conduction through the pavement, allowing energy to be transported into the pipe and out of the pavement, therefore, preventing a significant temperature rise.

Figure 16 Maximum temperature (T _max) of samples subjected to sinusoidal solar flux and varied ambient temperature over 7 days. Note: W = 61 cm.

The boundary conditions from the previous simulation were applied to a 3D extension of the model. T _f was varied linearly along the direction of the pipe from 16.8 to 26.8°C to account for the rise in fluid temperature (representing a constant heat flux case – the worst case scenario). Figure 17 shows the temperature distribution of the aforementioned configuration with the spreader layer. With this configuration, the maximum temperature around the wheel path is lowered by 35–40°C. This is a substantial improvement over the case without a spreader. In summary, the temperature of the pavement surface is reduced, increasing the stiffness of the asphalt mixture and its resistance to permanent deformation.

Figure 17 Maximum temperature distribution of pavement with spreader layer, using sinusoidal solar flux and ambient temperature. Note: W = 61 cm, pipes lie outside wheel path (position of pipes projected onto surface with dashed lines, T _f varies linearly from 16.8 to 26.8°C (front-to-back).

Discussion

Rutting is a serious problem in HMA pavements in many regions of the world (Brown and Cross 1992, Bonaquist and Mogawer 1997, Pomerantz et al. 2000, White et al. 2002). The subtropical and tropical regions of the globe are densely populated; in these regions roads are generally trafficked by heavily loaded (quite often overloaded) vehicles. A combination of high temperature and load leads to severe rutting damage that causes over ten million dollars in maintenance expenditure in the USA alone. Rutting can be reduced by either modifying the asphalt to make it stiffer at higher temperature or reducing the pavement temperature. Polymer-modified binders have been used to increase stiffness of pavements at higher temperatures, but their cost is double that of asphalt (Srivastava 2009). The approach of lowering the pavement temperature using reflective coating does not provide the optimal solution (Pomerantz et al. 2000, Kubo and Kido 2006, Synnefa et al. 2009, Wan and Ping 2009). Therefore, alternatives must be explored as a means of reducing rutting-related damage.

The proposed approach is a modification of the approach developed by the Dutch group for energy extraction from HMA pavements (Loomans et al. 2003, Srivastava 2009). Their design used a closely spaced piping network below the pavement with flowing water that extracted the heat. Since it would be impractical to have such closely spaced pipes below active roads, we proposed a modified design. This design incorporates a conductive spreader layer below the wheel path at depths of 25 mm or greater connected to the embedded pipes that are located away from the wheel paths (Figure 1). The road below the wheel path is the region prone to rutting failure and, therefore, cooling that region is a priority. Numerical analysis of the design is able to establish temperatures below 45°C along the wheel path with only two pipes placed within the active road and the rest in the median/shoulder region (Figure 17). This is a substantial improvement in lowering the temperature from a peak of 75°C on hot sunny days as obtained from simulations (Figure 16). Therefore, there is a substantial practical advantage of using a spreader layer in conjunction with pipes below the pavement.

The reduction in rutting will increase the service life of a given roadway, allowing for a much longer interval between resurfacing. However, resurfacing may be inevitable during the existence of the road. In order to prevent damages to the piping system, other than developing a new system for resurfacing, milling should be limited to the pavement layer above the pipes. During major rehabilitation work (which has a significantly higher interval than maintenance work), a re-laying of the system can be done.

The success of this design is dependent on the spreader material. As seen in this study, a 6.4 or 12.8 mm thick spreader (0.25 or 0.5 in.) with a k _s value above 100 W/m K will be a good spreader material (Figure 13). Although several metallic sheets such as copper and aluminium are available with these specifications, they are cost prohibitive and will most likely lead to shear failure due to low adhesion to asphalt. Geosynthetics are a durable and flexible man-made material that has adequate friction with asphalt mix layer; therefore, they are compatible for long-term maintenance and are the material of choice for use as a heat spreader. Geosynthetics already serve a multitude of tasks in pavement structures such as drainage, separation and reinforcement (Joint Departments of the Army and Air Force 1995, Koerner 1998, Button and Lytton 2007). They have also been used as reinforcement to prevent rutting. However, they are made of polypropylene or polyethylene terephthalate, which are normally poor conductors and would be inadequate spreaders. The option is to use carbon fibres that have high conductivity and can be woven in the form of mats. Already, polyacrylonitrile based woven composite fibres exist which can be used as spreader material. One of the features of the current design is that a high k _sx is adequate for the spreader material. This allows composite laminate sheets that have high lateral conductivity (k _x ) but poor transversal conductivity (k _y ) to be used for spreader material. The use of flexible sheets that can be easily replaced and coupled to the pipes without difficulty during repaving of pavement surfaces makes composite laminate sheets an attractive option as a spreader layer.

Although the current study theoretically investigated the feasibility of putting a spreader–pipe network below the pavement, the study was limited to a thermal analysis only. Structural analysis is required on the pavement–pipe–spreader system before large-scale implementation is possible. Previous studies have conducted stress analysis on pavements with embedded pipes only (van Bijsterveld 2000, van Bijsterveld and de Bondt 2002). They have shown that although stress concentration takes place at the pipe–pavement interface, they are below the critical failure point. Some of the stress concentration (for pipes located directly under or close to wheel path) can be mitigated by having the pipes deeper in the pavement.Since the proposed system has pipes that are further away from the wheel path, this problem can be minimised. Similar stress analysis under cyclic compressive loading needs to be conducted. A comprehensive analysis that incorporates modulus values of piping–spreader materials is required.

Finally, the current design presents new opportunities that have not been fully explored for HMA pavements. The pipes have the potential of delivering a circulating water/fluid at temperatures above 55°C on a regular basis in warm climates. The hot water/fluid can be used for various applications. In addition to serving hot water to a community, the water may be used in absorption chiller or refrigeration systems, air conditioning systems or as a heat source to an organic Rankine cycle for power generation or simply as process heat for industrial operations (Goswami et al. 2000). Preliminary economic analysis provides favourable payback in regions of high solar insolation (Mallick et al. 2011). Although further design analysis and optimisation will be required for scaling up the system, the current design provides the first step towards sustainable HMA pavements.

Conclusion

In the current study a theoretical thermal analysis has been conducted to determine the major factors affecting the surface temperature distribution of a HMA pavement. A pipe with fluid circulating within it is coupled to a conducting spreader layer and the entire system has been embedded under the pavement. Sensitivity analysis of the temperature distribution as well as T_max was conducted as a function of pipe spacing (W), depth of the pipe–spreader system (D), thickness (t_s) of spreader layer (and other variable thickness) and thermal conductivity of the spreader (k_s). It was found that pipes alone cannot cool a very large area because of poor pavement conductivity, and spreaders are effective in cooling a larger area of the pavement. Increasing t_s and k_s allows effective spreader with lowering of T_max, but material and structural limitations will determine the final properties. One option is to place the pipe–spreader deeper (D = 50 mm) inside the pavement and use higher t_s values. Deeper placement of the system will also allow design flexibility of lipped or sloped spreader layers which allows one to cool the pavement further. It may shield the system from stress concentrations. A sensitivity analysis to the two major boundary conditions (Q_s and T_f) shows the slopes of increasing T_max with higher flux and fluid temperature. Finally, a realistic simulation clearly indicates that T_max can be lowered by more than 30°C – thus significantly improving the rutting resistance of HMA pavements.