Instrumentation of Field-Testing Sites for Dynamic Characterization of the Temperature-Dependent Stiffness of Pavements and Their Layers

Falling weight defectometer (FWD) tests are performed worldwide for assessing the health of pavement structures. Interpretation of FWD-measured surface defections turns out to be challenging because the behavior of pavement structures is temperature-dependent. In order to investigate the infuence of temperature on the overall pavement performance and on the stifness of individual layers, temperature sensors, asphalt strain gauges, and accelerometers were installed into one rigid (concrete) and two fexible (asphalt) pavement structures, mostly at layer interfaces. Tree diferent methods for installation of the strain gauges are compared. From correspondingly gained experience, it is recommended to install a steel dummy as a place-holder into the surface of hot asphalt layers, immediately after their construction and right before their compaction, and to replace the dummy with the actual sensor right before the installation of the next layer. Concerning the frst data obtained from dynamic testing at the feld-testing sites, FWD tests performed at diferent temperatures deliver, as expected, diferent surface defections. As for the rigid pavement, sledgehammer strokes onto a metal plate, transmitted to the pavement via a rubber pad, yield accelerometer readings that allow for detection of curling (=temperature-gradient-induced partial loss of contact of the concrete slab from lower layers). In the absence of curling, the here-proposed sledgehammer tests yield accelerometer readings that allow for quantifcation of the runtime of longitudinal waves through asphalt, cement-stabilized, and unbound layers, such that their stifness can be quantifed using the theory of elastic wave propagation through isotropic media.


Introduction
Roads are exposed to variable atmospheric conditions. Te corresponding changes in temperature have a signifcant infuence on the performance of rigid and fexible pavement structures: (1) So-called fexible pavements include layers of asphalt. Te stifness of bituminous asphalt materials decreases with increasing temperature [1,2].
(2) So-called rigid pavements include concrete slabs. Teir temperature-gradient-induced curling (� partial loss of full-face contact along one of the layer interfaces) reduces the structural stifness of concrete roads [3,4]. (3) Many pavement structures include unbound granular layers. Teir stifness was shown to be a function of stress level and moisture content. Te latter correlates with temperature changes in the unbound layers [5,6].
Consequently, it is challenging to interpret diferent surface defections measured during nominally identical falling weight defectometer (FWD) tests performed on the same pavement structure at diferent temperatures. Te described situation provides the motivation to gain (i) more insights into the load-carrying behavior of multilayered pavement structures subjected to dynamic loading and (ii) direct access to the stifness of individual layers of interest. To this end, one rigid and two fexible pavement structures are equipped with three types of sensors: (i) Pt100 sensors in order to measure the temperature at specifc depths of the pavement structure, (ii) strain gauges in order to quantify the horizontal normal strains of asphalt in radial directions relative to the center of the falling weight during FWD testing, and (iii) accelerometers in order to analyse how dynamic loads propagate through pavement structures.
FWD testing is a worldwide popular nondestructive method for the assessment of the health of pavement structures. FWD tests consist of dropping a standardized weight onto a damped spring system placed over a circular load plate that transmits the dynamic load to the pavement structure. Te force history is measured using an integrated load cell. Several displacement sensors measure the vertical defection history at specifc distances from the center of the falling weight [7,8]. Such sensors include accelerometers, which consist of a mass and spring system and convert the acceleration of the ground into electrical signals [9], geophones, which consist of a coil suspended on a permanent magnet and convert the ground velocity into an electric signal [10], and seismometers, which measure any seismic activity and may convert ground displacement, velocity, or acceleration into electrical signals [8,11]. Displacement values may be obtained from all types of sensors by integrating the acceleration and velocity signals.
Surface defections measured during FWD tests are usually evaluated using one of two popular conceptual approaches. Te frst one refers to the quantifcation and interpretation of defection basin parameters such as the surface curvature index (SCI) or the AREA parameter, see [7,[12][13][14][15][16]. Similar indexes have been developed for quantifcation of asymmetric behavior of concrete slabs subjected to central FWD testing [17][18][19]. Te second approach for evaluation of FWD tests refers to back-calculation of properties of the tested pavement structure in order to minimize the diference between measured and simulated defections. Two types of structural simulation models are frequently used: multilayered elastic half-space models and "dense-liquid" models. Te former models explicitly resolve the individual layers of pavement structures [20]. Backcalculation is commonly aimed at quantifying the stifness of the individual layers. Diferent commercial programs frequently produce diferent results even when fed with the same input data [21] because diferent combinations of layer moduli and thicknesses produce (virtually) the same defections [8,22]. Te second type of structural models ("dense-liquid models") idealizes pavement structures as an elastic plate resting on a Winkler foundation [23,24]. Backcalculation is aimed at quantifying the bending stifness of the plate and the modulus of subgrade reaction of the Winkler foundation. Analytical formulae facilitate the backcalculation procedure, see [3,[25][26][27]. Notably, backcalculations were also carried out in the context of dynamic analyses with the aid of fnite element (FE) simulations [28][29][30][31], also with the aim of assessing the infuence of longitudinal cracks on FWD test results [32].
Te present study builds on experience from FWD research approaches which consisted of equipping road sections and pavement testing facilities with diferent types of measurement sensors. Common sensors installed for in situ pavement monitoring [33] include horizontal and vertical strain gauges [34,35], fber optic sensors [36], LVDTs [37], accelerometers [38], temperature [39], moisture [40], and pressure sensors [41]. Tey have been proven to be useful in the context of validation of FWD back-calculations [42], in particular to assess the diference between laboratory and back-calculated asphalt moduli [43] and for separating transient and permanent deformations [44]. A variety of strain gauges, pressure cells, defection, and temperature sensors were installed into fexible pavement, and recommendations were given regarding their selection and use [45]. Te readings of embedded strain gauges and pressure cells were compared with back-calculated stresses [46][47][48] and strains [49][50][51]. Optical fber-based sensors were embedded for pavement health monitoring and damage detection [52]. Multidepth defectometer sensors have been installed to improve the interpretation of FWD data with respect to base and subgrade damage [53]. Moisture and groundwater sensors allowed for assessing the infuence of moisture content and depth of the groundwater table on FWD defections and back-calculated stifness of the unbound layers [5]. Te recorded time history of the defections measured during an FWD test, together with strain sensor readings, was exploited to study the cross-anisotropic viscoelastic properties of asphalt concrete [54]. MEMS accelerometers have been installed (i) to compute displacement histories either by means of double time-integration or by constrained least-squares estimation and (ii) to compare the resulting data with surface displacements measured by FWD geophones [55].
Te present study serves two main purposes: (i) experience with instrumentation of rigid and fexible pavements during their new construction will be gained and shared. Tis concerns particularly the installation of the strain gauges into asphalt layers. Tree diferent methods will be compared: "method A: cut, install, and cover after asphalt placement," "method B: installation in a fxation tool before asphalt placement," and "method C: use of steel dummy place-holders for the real sensors." (ii) First data from dynamic testing at the innovatively equipped feld-testing sites will be presented and discussed. Tis includes data from both FWD tests and a newly proposed "sledgehammer test." It consists of sledgehammer strokes onto a metal plate, transmitted to the pavement via a rubber pad. Te sledgehammer test is performed in order to obtain accelerometer readings that allow for quantifcation of the runtime of longitudinal waves through asphalt, cement-stabilized, and unbound layers, such that their stifness can be quantifed using the theory of elastic wave propagation through 2 Structural Control and Health Monitoring isotropic media. To this end, acceleration sensors are installed both at the top and the bottom of layers of interest. Te present manuscript is organized as follows. Section 2 describes the three specifc feld-testing sites together with the corresponding instrumentation layouts. Section 3 refers to the accelerometers: it presents the theoretical foundations for quantifcation of layer stifness, the criteria for sensor selection, and the experience gained from the installation of the sensors. Section 4 refers to the asphalt strain gauges: it presents the criteria for sensor selection, three diferent approaches for the installation of the sensors, and the experience gained with them. Section 5 discusses exemplary data from FWD testing of rigid and fexible pavements, together with results from the sledgehammer tests. Section 6 closes the paper with conclusions drawn from the presented results.

Overview of the Three Field-Testing Sites
One rigid and two fexible pavement structures were instrumented to become feld-testing sites for FWD experiments. Tey were equipped with temperature sensors, accelerometers (Section 3), and asphalt strain gauges (Section 4), see Figure 1.
As for temperature measurements, platinum-based detectors with an electrical resistance of 100 Ω, see the Pt100 in Figure 1, were installed in asphalt, concrete, cementstabilized, and unbound layers. Te sensors were connected by means of 4 wires running through Perfuoroalkoxy cables to a LEMO connector. Te cables were protected by a Ø5 × 40 mm stainless steel sleeve, a 300 mm heat shrink sleeve, and a waterproof-corrugated plastic tube.
Signals of all installed sensors were recorded during dynamic testing by means of the mobile data acquisition systems DEWE43-A (strain gauges and accelerometers) and KRYPTONi-8 × RTD (temperature sensors) by DEWESoft, see Figure 1. Tese systems are suitable for measurements at multiple unsheltered feld-testing sites because they are water-, dust-, and shockproof, and they are capable of operation within the temperature range from −40°C to 85°C. Te required electricity was provided by a portable power bank Novoo 230 Wh, see Figure 1. Between successive measurement days, the connector-ends of all sensor cables were stored inside a stainless steel box near the verge.
Tree diferent types of pavement structures, frequently used on the Austrian motor-and expressways, were instrumented: (i) a rigid pavement consisting of concrete slabs with dowel bars placed along transverse joints and tie bars along longitudinal joints, (ii) a fexible pavement consisting of asphalt concrete layers placed over a cementstabilized granular layer, and (iii) a fexible pavement consisting of asphalt concrete layers over two unbound granular layers. All three feld-testing sites were installed in the course of major rehabilitation treatments.

Field-Testing
Site #1 on Motorway A10. Field-testing site #1 is a concrete slab on the motorway A10, south of Salzburg. A slab of the emergency lane was selected, see Figure 2(a). It has the same pavement structure as the trafc lanes, see Figure 2(c), and it can be closed for FWD measurements without interrupting the trafc. Te width and the length of the slab amount to 3.50 m and 5.00 m, respectively, see Figure 2(b). Te measurement sensors were installed during the rehabilitation of the motorway. It included the following steps. Te existing asphalt and concrete layers were removed such that the subgrade was covered by two unbound granular layers only. Te upper unbound granular layer was mixed in-place with cement and water, followed by compaction, in order to transform it into a cement-stabilized layer (L3). A separation layer made of asphalt (L4) was installed. Two concrete layers (L5 and L6) were laid in one pass using a train of two slipform pavers, which are the standard concrete pavement construction techniques in Austria. Both layers are designed according to their function in the pavement structure in order to optimize functional performance (e.g., skid resistance and noise), service life, and costs [56]. Tis completed new pavement structure is depicted in Figure 2(c).
Sensors were embedded at several depths. Six temperature sensors (T2-T7) were placed at all interfaces between neighboring layers as well as in the middle of the lower concrete layer, see Figure 2(c). Tree accelerometers (B2-B4) were installed along a vertical axis running through the central measurement point (MP) for FWD testing, at the interfaces between the subgrade (L1), the unbound layer (L2), the cement-stabilized layer (L3), and the asphalt layer (L4), see Figures 2(b) and 2(c). Four strain gauges (A1-A4) were installed at the bottom of the asphalt layer, see Figure 2(d) for the symmetric crosswise arrangement relative to the measuring point.
Regarding the sequence of sensor installation, sensors B2 and T2 were put in place before the cement-stabilization of L3. Immediately after the stabilization, a trench was cut with an excavator down to the interface between layers L2 and L3. Tere, the sensors B3 and T3 were installed, the trench was reflled with the excavation material, and the layer was compacted. Tese works were completed before the start of the main phase of the cement hardening process. Sensors B4 and T4 were installed shortly before the placement of the asphalt concrete. Te strain gauges and the temperature sensors T5, T6, and T7 were installed immediately after the construction of the asphalt layer and the lower concrete layer, respectively. Details regarding the selection of the accelerometers and the strain gauges, the design of their positions, and the used installation procedures are described in Sections 3 and 4, respectively.
Te thicknesses and mass densities of the layers were determined as follows. A precision laser was used to measure the actual layer thicknesses, see Table 1. Tey slightly deviate from the nominal values of Figure 2 because of execution tolerances. Te mass densities of the two granular layers L2 and L3 were quantifed in situ using the water replacement method. After digging a hole into the layer of interest, the excavated mass (measured by means of a portable digital scale) was divided by the volume of the hole. Te volume was determined by laying a thin sheet of plastic into the hole, pouring water into it until the hole was Structural Control and Health Monitoring 3 flled, and then measuring the volume of the water with a graduated cylinder. For both layers L2 and L3, the water replacement method was performed three times each, resulting in the mean values of the mass densities listed in Table 1. As for the layers L4, L5, and L6, samples of the three materials were collected during construction, see Table 1 for the mean values of the mass densities determined in the laboratory.

Field-Testing
Site #2 on Motorway A3. Field-testing site #2 is a fexible pavement on the motorway A3, south of Vienna. Two nominally identical FWD measuring points were instrumented at a distance of approximately 70 m from each other, see Figure 3(b). Both measuring points are located in the middle of the hard shoulder which has the same design as the rest of the carriageway. Te road section was rehabilitated as follows. Te existing asphalt layers were removed. Te existing  cement-stabilized layer (L3) was relaxed using a guillotine-type breaker followed by recompaction with a roller. Tis two-step "crack and seat" treatment reduced the potential for refective cracking of three newly constructed asphalt layers. Tey are referred to as "base course L4," "binder course L5," and "surface course L6." Sensors were embedded at several depths. Five temperature sensors (T2-T6) were placed at all interfaces between neighboring layers, see Figure 3(c). Four accelerometers (B2-B4 and B6) were installed along a vertical axis running through the measurement points (MP) for FWD testing, at the interfaces between the subgrade (L1), the unbound layer (L2), the cement-stabilized layer (L3), and the base course (L4), as well as between the binder course (L5) and the surface course (L6), see were installed at the bottom of the base course, see Figure 3(d) for their symmetric arrangement relative to the measurement points. Two rather than four strain gauges were installed in order to reduce the number of embedded inhomogeneities, accounting for the fact that the asphalt layers have a signifcant load-carrying function in the pavement structure.
For the installation of sensors B2, T2, B3, and T3, rectangular areas with dimensions of 2.50 × 2.50 × 0.25 m were excavated from the cement-stabilized layer L3, symmetrically with respect to the FWD measuring points, down to the surface of the unbound layer L2, see Figure 4. A smaller trench was excavated from the unbound layer L2 down to the surface of the subgrade L1, so that B2 and T2 could be installed. Te material excavated from L2 was then reinserted and compacted with a vibrating plate compactor. Te sensors B3 and  Structural Control and Health Monitoring 5 T3 were installed. Te volume excavated from L3 was flled with lean concrete rather than with the excavated material because it was impossible to reinsert the excavated material in a way that would have led to properties similar to those of the relaxed cement-stabilized layer. Sensors B4 and T4, as well as the strain gauges, were installed shortly before the placement of the base course L4. Te temperature sensor T5 was installed immediately after the construction of the base course L4. Te sensors T6 and B6 followed immediately after the construction of the binder course L5. Details regarding the selection of the accelerometers and the strain gauges, the design of their positions, and the used installation procedures are described in Sections 3 and 4, respectively. Te thicknesses of the layers and the mass-density of the unbound layer L2, see Table 2, were determined using the same methods as described for feld-testing site #1. In situ cast cubes of lean concrete (L3) and fresh-mix samples of all three asphalt layers (L4-L6) were collected for laboratory testing. Te mass densities of these materials are listed in Table 2.

Field-Testing
Site #3 on Expressway S31. Field-testing site #3 is a fexible pavement on the expressway S31 in the federal state Burgenland. In the absence of a hard shoulder, the FWD measuring point was installed in the middle of the right lane (TL1), see Figure 5. Tere, predominantly heavyvehicle trafc is expected. In the longitudinal direction, the instrumented site is located after a breakdown bay. Tis position (i) facilitated the installation of sensors and (ii) renders in situ testing convenient and safe for the experimenters. Te rehabilitation treatment comprised (i) the renewal of the asphalt pavement and (ii) widening of the carriageways including the introduction of a central reserve. Te instrumented pavement structure consists of three asphalt layers resting on two unbound layers which are so similar that the interface between them could not be identifed.
Sensors were embedded at several depths. Tree temperature sensors (T2, T4, and T6) were placed at the interfaces between (i) the subgrade (L1) and the lower unbound layer (L2), (ii) the upper unbound layer (L3) and the base course (L4), as well as (iii) the binder course (L5) and the surface course (L6), see Figures 5(b) and 5(c). Temperature sensor T5 was installed at the interface between the base and binder courses. Tree accelerometers (B2, B4, and B6) were installed along a vertical axis running through the central measurement point (MP) for FWD testing, at the same interfaces as T2, T4, and T6. Two asphalt strain gauges (A1 and A3) were installed into the base course, see Figure 5(d) for their symmetric arrangement relative to the measurement point. Again, two rather than four strain gauges were installed in order to keep the number of embedded inhomogeneities at a reasonable minimum.
Te sequence of installation of the sensors B2-B4, T2, and T4-T6 was the same as in the other feld-testing sites. Te asphalt strain gauges, however, were installed at the top of the base course. Tis was part of an installation strategy which was specifcally aimed at ensuring the position accuracy of the strain gauges. Details regarding the selection of the accelerometers and the strain gauges, the design of their positions, and the used installation procedures are described in Sections 3 and 4, respectively.
Te thicknesses of the layers and the mass density of the unbound layers L2 and L3 were determined using the same methods as described for feld-testing site #1. Samples from all three asphalt layers (L4-L6, fresh mix) were collected for laboratory testing. Te mass densities of these materials are listed in Table 3.

State-of-the-Art Applications.
Two types of accelerometers have been mainly used in past pavement applications: integrated electronics piezoelectric (IEPE) and microelectromechanical system (MEMS) accelerometers. IEPE sensors are mainly used to capture dynamic events, exhibiting a frequency range of some 0.3 Hz-10 kHz. Tey consist of a fxed mass, a piezoelectric material (e.g., quartz or a piezoceramic), and an integrated signal amplifer to reduce noise [57,58]. If acceleration is imposed on the sensor, the mass will be pressed against the piezoelectric element. Te generated electric charge can be measured and correlated to the acceleration. MEMS sensors, in turn, are used for dynamic and low-frequency measurements and are also applicable for frequencies smaller than 0.3 Hz. MEMS accelerometers are either based on measurements of changes in electrical capacitance (capacitive sensors), or they use strain gauges (piezoresistive sensors), see [57,59]. Accelerometers were mounted in several studies to the surface of pavement structures in order to measure acceleration histories either caused by real trafc or by load simulators. Te measured acceleration histories were used as reference values for the optimization of dynamic simulations of the behavior of pavement structures [38,60]. Measured acceleration histories were also converted, by means of double time-integration, into   [14,55,[61][62][63][64]. Te defections were then used for back-calculation of layer properties. Accelerometers positioned on the pavement surface have also been used in the context of surface wave testing, see [65][66][67][68][69][70].
Depending on the used technique, surface wave pavement testing allows either for direct determination of the modulus of the top paving layer or for the estimation of the modulus of each layer through a back-calculation analysis.
In this study, accelerometers are used to determine the time of fight of longitudinal waves propagating vertically through these layers. Tis way, the theory of elastic wave propagation through isotropic elastic media can be used for the direct quantifcation of layer moduli [71,72,73].

Teoretical Fundamentals.
Tree types of stress waves are generated when hitting a pavement structure vertically at its surface. Longitudinal waves mainly propagate vertically downwards. Tey are also referred to as compression or Pwaves. Particle displacements are aligned with the direction of wave propagation. Transversal waves mainly propagate diagonally downwards. Tey are also referred to as shear or Swaves. Particle displacements are normal to the direction of wave propagation. Longitudinal and transversal waves are partly refected and refracted at the interfaces between diferent layers [74]. Rayleigh waves travel along the surface. Tey are also referred to as surface or R-waves. Herein, the focus rests on longitudinal waves.
Depending on the amplitude of stress waves, they may be either elastic, provided that they induce reversible deformation only, or inelastic, provided that also irreversible deformations take place. Te wave propagation front, in turn, always refers to an elastic wave because it is faster than inelastic waves [75]. Te present study takes advantage of this property.
Herein, accelerometers are installed both at the top and at the bottom of layers of interest in order to determine the time of fight, Δt, of the front of a longitudinal wave which propagates vertically through these layers. Based on measured values of Δt, the velocity of the longitudinal wave, ] L , can be quantifed as where h denotes the thickness of the layer of interest. Quantifcation of the elastic stifness of the layer based on its longitudinal wave velocity is facilitated by two realistic assumptions: (i) the layer is idealized as a macrohomogeneous material and (ii) the longitudinal wave is considered to be a bulk (rather than a bar) wave, meaning that the lateral deformation is prevented (rather than free). Under these premises, the theory of elastic waves propagating through isotropic media delivers the following relation between the component C 1111 of the elastic stifness tensor, the mass density ρ of the material, and ] L of equation (1): Assuming the layer material to be isotropic and its Poisson's ratio ] to be known, the following standard relation of isotropic elasticity allows for quantifcation of the modulus of elasticity E: 3.3. Sensor Selection. IEPE accelerometers 602D61 and HT602D61 by PCB with stainless steel housing and a ceramic sensing element, see Figure 1, comply with the requirements of the present study. Tey are suitable in terms of measurement range and resolution, they are reasonably small, and they have a sufcient resistance against (i) mechanical impact during compaction and (ii) high temperatures during construction of asphalt layers. Te side exit of the sensors together with a 3 m armored jacketed sleeve provide the high level of protection for the connector and the cable, which is required in the present application.

Data Acquisition Rate and System.
A suitable data acquisition rate had to be selected in order to ensure that the time of fight of a longitudinal wave propagating through a layer of interest can be determined with acceptable accuracy. In order to achieve a measurement accuracy of at least 10%, it is necessary to record at least 10 acceleration values while a longitudinal wave travels from the accelerometer at the top of a layer to the accelerometer at the bottom of the same layer. Tis calls for the data acquisition rate which is by factor of at least 10 larger than the inverse of the estimated time of fight of the elastic wave through the layer of interest. In order to estimate times of fight a priori, equations (1)-(3) were evaluated for the designed thicknesses of the layers, see Tables 1-3, and values of the elastic modulus, Poisson's ratio, and mass density of each layer of interest were taken from the literature [76][77][78][79][80], see Table 4. For asphalt and unbound materials, upper and lower bounds of the elastic modulus, representative for winter and summer conditions, respectively, were taken into account. Te stifness properties of the concrete and cement-stabilized layers were assumed to remain virtually constant within the temperature range investigated [81]. Te estimated times of fight are listed in Table 4.
(1) At the feld-testing site #1 on the A10, three accelerometers were installed: at the bottom of the unbound layer L2, at the interface between L2 and the cement-stabilized layer L3, and at the top of L3, see Figure 2. Tis allows for in situ stifness characterization of layers L2 and L3. Te required data acquisition rate required is equal to 74 kHz, see Table 4 (as for quantifcation of the stifness properties of the Structural Control and Health Monitoring concrete and asphalt layers, samples were taken for laboratory testing). (2) At the feld-testing site #2 on the A3, four accelerometers were installed: at the bottom of the unbound layer L2, at the interface between L2 and the lean concrete layer L3, at the interface between L3 and the asphalt base course L4, and at the top of the asphalt binder course L5, see Figure 3. Tis allows for in situ stifness characterization of layers L2 and L3 as well as of the sandwich layer consisting of the asphalt base and binder courses (L4 and L5). Te required data acquisition rate required is related to the high stifness of asphalt exposed to winter temperatures. It is equal to 169 kHz, see Table 4. (3) At the feld-testing site #3 on the S31, three accelerometers were installed: at the bottom of the lower unbound layer L2, at the interface between the upper unbound layer L3 and the asphalt base course L4, and at the top of the asphalt binder course L5, see Figure 5. Tis allows for in situ stifness characterization of two sandwich layers. Tey consist of the unbound materials (L2 and L3) and of the asphalt base and binder courses (L4 and L5), respectively. Te required data acquisition rate is equal to 154 kHz, see Table 4.
Te USB data acquisition system DEWE-43-A, see Figure 1, with a data acquisition rate of 200 kHz and eight fully synchronized channels complies with the requirements of all three feld-testing sites. DEWE-43-A supports voltage and full-bridge signals without additional adapters, as well as IEPE, charge, thermocouples, half-bridge, quarter bridge, RTD, current, resistance, and LVDT signals with the help of DSI adapters. Tus, the DEWE-43-A is capable of simultaneously acquiring data from four strain gauges and four accelerometers. DSI-ACC adapters were required for powering the IEPE accelerometers with the required directed current. As for the ACC, a BNC adapter RS 124-2521 was used.

Sensor Installation.
Installing accelerometers at multiple depths during the construction of the pavement represented a novel challenge, difering from the installation at a single depth and after construction as described in [14,38,62,64]. As for the present study, the following two requirements had to be fulflled: (1) Te sensors must remain in place and deliver a reliable signal after the installation and compaction of subsequent layers of the pavement structures (2) Te sensors must be aligned along a vertical axis, such that the longitudinal wave, produced by hitting the surface of the completed pavement right above the sensors, propagates downwards along this axis Te installation of the accelerometers in unbound granular layers (B2 and B3) was conducted as follows, see also Figure 4. Te starting level was that of the unbound layers, given that they remained in place during the rehabilitation works on all three feld-testing sites. For the installation of the lowest sensors B2 and T2, an excavator was used to reach the boundary between the unbound layer L2 and the subgrade L1. Layer-wise excavation allowed for separating diferent materials in order to refll the trenches later with the right excavation materials. In a frst step, the excavation reached the boundary between L2 and L3, see Figure 4(a). In a second step, a smaller trench was created reaching the boundary between L2 and L1, see Figure 4(b). At this level, sensors B2 and T2 were embedded using a quick-setting cement mortar that served two goals: (i) to prevent a displacement of the sensor or a disconnection of the cable and (ii) to protect the sensor from direct contact with (tips of ) large aggregates, since concentrated loads could damage the sensors, see Figure 4(c). Te exact horizontal and vertical position of all sensors was determined using a measuring tape and a precision laser, respectively. Two independent permanent reference points were used. Te cables from both sensors were placed in a protection tube starting approximately 30 cm away from the sensors. In the next step, the excavation material from layer L2 was reinserted in the smaller trench, and the layer was compacted using a vibrating plate, see Figure 4(d). At the interface between L2 and L3, sensors B3 and T3 were installed similarly, see Figure 4(e). Finally, the remaining trench was closed with the excavation material from L3 and compacted. Only in the case of the second feld-testing site on the A3 motorway, as described in Section 2, a new material was inserted, see Figure 4(f ).
High-temperature accelerometers were installed at the interfaces between asphalt layers and the cement-stabilized or lean concrete layers (B4 and B6). Te procedure is illustrated in Figure 6 and described next. First, openings for the sensors and grooves for the cables were cut in the cement-stabilized or lean concrete layers, see Figure 6(a). Ten, the sensors were fxed and the cables were covered using quick-setting mortar, see Figure 6(b). Right before paving, the sensors were frst covered with loose asphalt mixture, followed by careful compaction with a hand tamper, see Figure 6(c). Tis provided protection to the sensor against high compaction forces and slightly reduced the temperature of the material in contact with the sensor. Special care was taken so that delivery trucks and, especially, the track chain of the paver did not drive over the positions of the sensors. Te plastic tubes protecting the cables were used only outside of the bound layers (starting from the edge of the pavement) in order to minimize potential weak spots and cavities.
For the installation of the accelerometers between two asphalt layers (B6), the opening for the sensor and the groove for the cable were produced during the construction of the lower asphalt layer. Tis was achieved by pushing a steel dummy of the sensor and a steel pipe as placeholders for the sensor and the cable into the surface of the freshly placed asphalt, followed by regular compaction with rollers. Later, the dummy and the pipe were removed, and the sensor together with its cable could be simply inserted into the opening and the groove, see Figures 6(d) and 6(e). Before the construction of the next layer, a small quantity of loose asphalt was sieved using a standard sieve to obtain a material with a grain size smaller than 8 mm. Te sieved material was used to fll remaining cavities and to cover the accelerometer with approximately 1 cm of asphalt as a protective measure, see Figure 6(f ). Ten, the unsieved material was piled on the sensors and compacted manually, as shown in Figure 6(c). Te new asphalt layer was afterwards paved as usual. All but two of the installed accelerometers deliver reliable signals under dynamic loads. Te exact reason why sensors B2 on the A10 and B6 of the S31 do not work remains unknown. It is speculated that either the sensor, and/ or the sensor-cable connection, and/or the cable was/were mechanically damaged during construction.

State of the Art Applications.
Asphalt strain gauges have been successfully employed in the context of pavement testing [82][83][84], monitoring of instrumented sections [85,86], studies of stifness properties of layers [87,88], and in comparing vehicle loads with FWD tests [89]. Te performance of KM-100HAS and other asphalt strain gauges has been compared in full-scale experiments under controlled loading and temperature conditions [90]. Diferent installation methods have been studied in a project involving 374 strain gauges which were used to monitor pavements over a period of four years [91].
In this study, the measurements of the asphalt strain gauges are used for validation of numerical simulations of the pavement structure subjected to FWD loading. In addition, they allow for detecting anomalous FWD results, such as obtained when the slab is curling.

Sensor Selection. KM-100HAS asphalt strain gauges by
Tokyo Measuring Instruments were installed, see Figure 1. Tey have a temperature range from −20°C to +180°C, an amplitude range of ±5000 μm∕m, and a measurement length of 100 mm. Tese sensors have reinforcing bars at both ends, see Figure 1. Tey ensure a frm embedment in asphalt. Te used data acquisition system, DEWE-43-A, provides a 5 V (350 Ω) excitation for full bridge sensors. Tis is larger than the recommended voltage (2 V) but smaller than the allowable bridge excitation for KM-100HAS (10 V).

Design of the Installation Position of the Strain Gauges.
Te installation positions of the strain gauges were decided based on the results of linear elastic, static, and axisymmetric fnite element (FE) simulations of an FWD experiment on multilayered pavement structures performed with ABAQUS [92]. For each one of the three feld-testing sites, a customized simulation was performed. Te elastic properties of the layers were taken from Table 4. Te imposed FWD forces were set equal to 200 kN for feld-testing site #1 and to 150 kN otherwise. Tese forces were selected based on experience and according to the Austrian FWD standard stating that in the case of very stif pavement structures, the applied force should be increased until that the magnitude of the measured defections exceeds 0.020 mm (at r � 1.80 m) [93]. It was also considered that this requirement should be satisfed even in the winter months. Exemplary FE results for the feld-testing site #3 (winter simulation) are shown in Figure 7, whereby the used pavement and load model is depicted in Figure 7(a).
Te essential results from the FE simulations are the radial normal strains of asphalt, as a function of the distance from the center of the falling weight, in the specifc depths from the Te horizontal distance of the asphalt strain gauges from the center of the falling weight was determined based on the following four considerations: (i) the strain shall be sufciently large to obtain reliable measurements; (ii) the strain gradients at the position of the sensors should be reasonably small, such that the measurements are easy to interpret; (iii) the diference between the simulated strain during summer and winter should be sufciently large, such that the measurements will capture seasonal variations; and (iv) the sensors should not be installed too near to the load center to limit a potential infuence on the fight-time measurement. Based on trade-ofs between these considerations, it was decided to install the strain gauges in horizontal distance from the center of the falling weight amounting to 45 cm for testing site #1 on the A10, to 25 cm for testing site #2 on the A3, and to 35 cm for testing site #3 on the S31.

Installation Method A: Cut, Install, and Cover after Asphalt Placement.
Installation method A was applied at the feld-testing site #1 on the A10, see the schematic overview in Figure 8(a). Te installation method is similar to the trenchcut method described in [91], with the diference that we used the same asphalt mixture to fll the holes which were excavated in order to place the sensors (rather than a diferent material as in [91]). Te installation began immediately after the placement of the asphalt mixture and the frst roller pass, see Figures 8(a)-A and 9(a). Te position of the four strain gauges was marked, see Figure 9(b). Te openings for the sensors and grooves for the cables were excavated using a pickaxe, a geological hammer and a shovel, see Figure 9(c). Tese openings reached down to the cement-stabilized layer on top of which the asphalt layer was constructed, see Figure 8(a)-B. Te manufacturer of the strain gauges recommends covering the sensors with asphalt having a maximum aggregate size of 5 mm. Terefore, a standard sieve was used to decrease the maximum aggregate size of hot asphalt, directly taken from the auger of the paver, from 16 mm to smaller than 8 mm, see Figure 9(d). Sieving worked best when using a short wooden plank to press the material through the sieve. From the sieved material, 1 cm thick asphalt beds were produced, the strain gauges were placed on top, and their vertical position was measured, see Figures 8(a)-C and 9(e). Te sensors were covered with another 1 cm thick layer of the sieved material. Finally, the opening was closed using the regular asphalt with a maximum aggregate size of 16 mm. Compaction was started by hand with a tamper and continued with a roller, see Figure 8(a)-D. In the immediate vicinity of the installed strain gauges, the roller was operated in the static mode rather than in the vibrating mode.  Te following experience was gained with installation method A. Its main advantages are the position accuracy of the sensors and the low potential for damage during construction. Te main disadvantages are problems related to rather fast cooling of the asphalt. It made the installation process quite stressful, and it got progressively more difcult to work with the material. Tese problems manifest themselves in visible imperfections regarding the uniformity of the asphalt layer in the region of the feld-testing site, see Figure 9(f ). While these imperfections are most probably the result of delayed compaction, they are rather unproblematic in the case of feld-testing site #1, since a concrete slab was later placed on top of the asphalt layer. At the feld-testing sites #2 and #3, however, the asphalt base course serves a much more important role in the behavior of the pavement structures. Terefore, installation method B was designed and used for feld-testing site #2.

Installation Method B: Installation in a Fixation
Tool, before Asphalt Placement. Installation method B was applied at both measuring points of the feld-testing site #2 on the A3, see Figure 8(b). Te installation method is an extension of the mound method described in [91]. A device was developed and produced to ensure the position stability of the strain gauges during construction, see Figure 10 the L-shaped bars ensured that an asphalt strain gauge ftted into the fxation device. Te distance between the two Lshaped bars was designed to be slightly longer than the length of the sensor, leaving room in longitudinal direction for the strain gauge to operate without constraints. Te vertical parts of the L-shaped bars had openings for the cable and the axial reinforcing bar of the sensor, respectively. One day before the asphalt paving, the device was screwed onto the underlying lean concrete layer, see Figures 8(b)-A and 10(b). Wooden planks were installed temporarily to prevent construction vehicles from accidentally driving over the device. Right before asphalt placement, a 1 cm thick bed made of sieved asphalt (see also Subsection 4.4) was placed on the device, and the strain gauge was installed on top of it, see Figure 8(b)-B. Te openings of the vertical parts of the Lshaped bars were closed with wire in order to prevent possible vertical movements of the sensors, see Figure 10(c). Tus, the fxation device prevented rigid body motions of the asphalt strain gauge, but in a way that allows the sensor to operate without constraints. Te sensors were covered, frst with sieved material, see Figures 8(b)-C and 10(d), and then with loose asphalt mixture around a larger area in order to protect the sensors from damage associated with laydown and compaction, see Figure 10(e). During the paving process, care was taken to ensure that the construction machines did not drive over the sensors, see Figure 10(f ). Roller compaction in the area near the sensors was conducted in the static mode than in the vibrating mode, see Figure 8(b)-D.
Te following experience was gained with installation of method B. Its main advantage is that the asphalt strain gauges remained in their desired positions during installation and compaction of the asphalt layer. Te main disadvantage is that despite the numerous steps taken to protect the asphalt strain gauges, only two out of four sensors work now that the construction work is fnished. Terefore, installation method C was designed and used for feld-testing site #3, with the aim to protect strain gauges from damage during compaction. Figure 8(c). Before installation, steel dummy sensors were produced, see Figure 11(a). Tey had the same dimensions as the actual strain gauges.

Installation Method C: Use of Steel Dummy Place-Holders for the Real Sensors. Installation method C resembles the installation of accelerometers in asphalt layers, see Subsection 3.5 and
Te installation method works only if the strain gauges are installed at the top of a freshly built asphalt layer, rather than at the bottom (as done at the feld-testing sites #1 and #2). Te installation process began right after the placement of the asphalt base layer, but before its compaction. Te steel dummies were hammered into the asphalt at the positions where the strain gauges should be fnally located, see Figures 8(c)-A and 11(b). Ten, the rollers compacted the asphalt in the vibrating mode, without paying special attention to the steel dummies, see Figure 8(c)-A. One day before placement of the next asphalt layer (�asphalt binder course), the steel dummies were removed, see Figures 8(c)-B and 11(c), and grooves for the cables of the sensor were cut into the surface by means of an angle grinder. Te sensors were installed, see Figure 11(e), whereby the small gaps to the surrounding asphalt were flled with cement paste in order to ensure frm bond, see Figure 11(d). Te strain gauges were covered with a sieved asphalt binder course material, followed by manual compaction, see Figures 8(c)-C and 11(f), in order to protect the sensors from damage associated with laydown and compaction.
Te following experience was gained with installation method C. Its main advantage is that the dummy placeholders allow for installing the strain gauges at the intended positions, without exposing the real sensors to overly high loads during compaction. Compaction of the asphalt layer into which the sensors are now embedded did not result in Figure 9: Installation of asphalt strain gauges at feld-testing site #1 on the A10 using method A: cut, install, and cover after asphalt placement.
loading of the sensors because the compaction loads were carried by the dummy place-holders. Compaction of the next asphalt layer on top of the installed sensors did not damage the sensors either because the sensors did not protrude from the surface onto which the next asphalt layer was constructed and because compaction of the new layer resulted in considerable strains in this layer, but not in the much cooler layer underneath, in which the sensors are embedded. Te main limitation of this method is that it can only be used, at least in the presented form, for installation of strain gauges at the top of an asphalt layer. Both strain gauges installed at feld-testing site #3 on the S31 deliver realistic measurements.

Results from Dynamic Field Testing on the Rigid Pavement.
Experiments at feld-testing site #1 were performed over four days from March 2021 until January 2022. Te FWD tests were performed with a maximum force of 200 kN. Te number of FWD tests performed immediately one after the other and the corresponding average values of the maximum defections measured by the geophones and of the maximum strains measured by the asphalt strain gauges are given in Table 5. Te defections measured in September, October, and January are similarly large, while those measured in March are signifcantly Figure 10: Installation of asphalt strain gauges at feld-testing site #2 on the A3 using method B: installation in a fxation tool, before asphalt placement. Figure 11: Installation of asphalt strain gauges at feld-testing site #3 on the S31 using method C: use of steel dummy placeholders for the real sensors.
larger. Tis underlines the challenges associated with the interpretation of FWD test results. Also, the measured asphalt strains underline the outstanding nature of the tests performed in March. Te strain measured in spring is equal to 16.6 × 10 −6 . Tis is signifcantly larger than the strains measured in fall and winter, which range from 4.6 × 10 −6 to 6.6 × 10 −6 . Te temperatures measured by means of a digital infrared thermometer at the surface of the slab and by means of the Pt100 sensors inside the pavement structure are listed in Table 6. Te asphalt temperature amounted to some 6°C in March, some 18°C in September, some 9°C in October, and some 0°C in January, see values of T 5 and T 4 in Table 6. It is concluded that stifness changes in the asphalt layer must have infuenced the behavior of the pavement structure. However, the temperature of asphalt alone cannot explain the signifcant diferences of FWD tests performed in March and in other months (note that the asphalt temperatures were quite similar in March and October). Tis provides the motivation to discuss, for the results obtained in March, indicators for temperature-induced curling, i.e., for uplift at the center of the slab leading to partial loss of contact along an interface between two adjacent layers of the pavement structure, Figure12(b). Tis type of slab curling is driven by the temperature diference between the top and the bottom of the slab [94], whereby the top is warmer than the bottom.
Te frst indicator for slab curling in March is provided by the temperatures measured across the depth of the concrete slab, see T surf , T 7 , T 6 , and T 5 in Table 6. In January, the temperature of the concrete slab was almost uniform. In March, September, and October, the temperature at the top of the concrete slab was larger than that at its bottom, i.e., T surf > T 5 . Te corresponding temperature diference, T surf -T 5 , amounted to 16.8°C in March, 2.9°C in September, 1.5°C in October, and −0.2°C in January, see Table 6. Tus, the temperature gradient experienced by the concrete slab in March was signifcantly larger than those in other months.
Te second indicator for slab curling in March is provided by the measured asphalt strains. In this context, it is recalled that the readings of the strain gauges were set to zero Table 5: Experimental results from FWD experiments on feld-testing site #1: average values (from n FWD tests) of the maximum defections measured by geophones at diferent distances from the center of the slab (w(r)) and of the maximum tensile strain obtained by the asphalt strain gauges (ε ASG ); note that throughout the manuscript, tension is associated with a positive mathematical sign.  Table 6: Experimental results from feld-testing site #1: temperature measured at the surface of the slab (T surf ), the top and mid-depth of the bottom concrete layer (T 7 and T 6 , respectively), the interface between concrete and asphalt (T 5 ), the interface between asphalt and the cement-stabilized layer (T 4 ), the interface between the cement-stabilized layer and the unbound layer (T 3 ), and the interface between the unbound layer and the subgrade (T 2 ), see also Figure 2.  before the frst FWD test of the day. In an FWD test on a curling-free pavement, the structure responds to the dynamic loading with the stifness of a frmly bonded multilayered half-space. Tis results in tensile strains at the positions of the asphalt strain gauges, as indicated by strains measured during the curling-free FWD tests performed in September, October, and January. An FWD test on a curled pavement structure can be subdivided into two phases. During the frst phase, the curled part of the pavement structure is pushed down until full-face contact is reestablished along all interfaces. During the second phase, the pavement structure responds to the continued dynamic loading with the stifness of a frmly bonded multilayered half-space. Let us consider that loss of contact occurred, in March, in the interface between the asphalt and the cementstabilized layer, i.e., that the asphalt was well bonded to the concrete slab and curled up together with it. During the frst phase of the FWD tests, the sandwich structure consisting of concrete and asphalt layers was pushed from the convexly curved initial confguration down to a plane state. Tis resulted in a signifcant tensile strain experienced by the asphalt strain gauges. During the second phase of the FWD tests, the tensile strain increased, similar to the situations in the curling-free FWD tests performed in September, October, and January. Te third indicator for slab curling in March is provided by the results of the sledgehammer tests, see Table 7 for the number of tests performed, corresponding results, and the derived values of the modulus of elasticity of the cementstabilized layer. Te tests performed in September, October, and January delivered stifness moduli in the overlapping intervals from 9.16 GPa ± 1.24 GPa to 7.70 GPa ± 1.62 GPa. Te accelerometer readings recorded in March, in turn, could not be evaluated because the signal arrived at both acceleration sensors virtually at the same time, see Figure 13(a). In addition, the amplitude of the signals captured in March was one order of magnitude smaller than the other months, compare Figures 13(a) and 13(c). Tese results can be explained as follows.
Te sledgehammer strokes were not strong enough to close the curling-induced gap between the asphalt and the cement-stabilized layer. Te front of the longitudinal wave propagating vertically downwards was refected at the upper free surface of the separated interface. Terefore, it did not arrive at the accelerometers below. Te wave resulting in the frst signals of the accelerometers had to travel around the separated interface. Tis wave was initially traveling diagonally downwards, away from the vertical axis through the accelerometers, towards the edge between the separated region and the contact region of the interface. Tere, the wave had to change direction and continued to propagate diagonally downwards, but this time underneath the separated region and towards the accelerometers, see Figure 12. Te efective propagation distances from the hit surface around the separated interface to the two accelerometers were similarly large. Tis explains why the wavefront arrived virtually at the same time at the two accelerometers, although they are buried at diferent depths. Te change of the traveling direction of the wavefront at the edge of the separated region, in turn, explains why the recorded accelerations were much smaller in March than that of the curling-free cases of other months.
It is very likely that full-face contact prevailed along all layer interfaces in September, October, and January. Tus, the dynamic wave created by sledgehammer strokes propagated vertically downwards, reaching accelerometer B4 frst and accelerometer B3 by some 95 μs later, see Table 7. Te diference between the wave arrival times at B4 and B3 could be measured in a straightforward fashion, see Figure 13(c). Te accuracy of the results is underlined by the standard deviations of the time of fight, ranging from 4 to 10 μs, see Table 7. Tis interval is virtually one-to two-times the resolution of the accelerometer readings which were captured with a data acquisition rate of 200 kHz, i.e., every 5 μs.
Te accelerometer readings captured during FWD testing, in turn, do not allow for reliable determination of the time of fight through the cement-stabilized layer, see Figures 13(b) and 13(d). Te reason will be explained in the next subsection where accelerometer readings are available at four rather than two diferent depths.

Results from Dynamic Field Testing on a Flexible
Pavement. Experiments at measurement point MP2 of feldtesting site #2 were performed over two days in July 2021 and April 2022, respectively. Te FWD tests were performed with a maximum force of 150 kN. Te number of FWD tests performed immediately one after other, and the corresponding average values of the maximum defections measured by the geophones are given in Table 8. Te strain gauges did not yield readings because they were damaged during installation. Te defections measured in July and April, respectively, are the same (�0.062 mm) in a distance of 2.1 m from the center of the falling weight. Tis indicates that the subgrade and the bottommost layers of the pavement Table 7: Experimental results from sledgehammer experiments on feld-testing site #1: mean values ± standard deviation (from n slh tests) of the time of fight through the cement-stabilized layer (Δt), its wave speed (v L ), and its modulus of elasticity (E), see Table 1 for layer thickness, h � 17.6 cm and mass density ρ � 2568 kg/m 3 , as well as Table 4  structure had the same stifness. With increasing proximity to the center of the falling weight, however, the diferences between defections measured in July and April increase both in absolute and in relative terms. At the center of the falling weight, the absolute diference is equal to 0.202 mm, and this equal to 79% of the defection measured in April. Tis indicates that the stifness of the topmost layers of the pavement structure were signifcantly diferent in July and April. Te temperatures measured by means of a digital infrared thermometer at the surface of the expressway, and by means of the Pt100 sensors inside the pavement structure, are listed in Table 9. Both times, the temperature in the lean concrete layer was virtually constant. Te temperature difference across this layer, T 4 -T 3 , was as small as −0.1°C in July and +0.6°C in April. Tis underlines that (i) curling of the lean concrete slab is very unlikely, and (ii) the seasonal diferences of the measured surface defections must have a diferent origin. Averaging the temperatures measured at the top and the bottom of each one of the three asphalt courses, yields 42.8°C (surface course), 34.4°C (binder course), and 31.7°C (base course) in July and 16.6°C (surface course), 13.5°C (binder course), and 12.9°C (base course) in April. Te seasonal temperature diferences have resulted in signifcant stifness changes of all three asphalt layers. Tis provides the motivation to discuss layer stifnesses quantifed from sledgehammer tests.
Te number of sledgehammer tests performed, corresponding results, and the derived values of moduli of elasticity are listed in Table 10. Te values of the thickness, the mass density, and Poisson's ratio of the lean concrete and unbound layers were taken from Tables 2 and 4. As for     Table 9: Experimental results from measurement point MP2 of feld-testing site #2: temperature measured at the surface of the slab (T surf ), the interface between surface and binder courses (T 6 ), between binder and base courses (T 5 ), binder course and lean concrete (T 4 ), lean concrete and unbound layer (T 3 ), as well as between unbound layer und subgrade (T 2 ), see Figure 3.   asphalt, the measured time of fight refers to a sandwich structure consisting of the binder and base courses, see Fig.  3. Its thickness is equal to 17.0 cm, its average mass density to 2,459 kg/m 3 , and its Poisson's ratio to 0.3, see Tables 2 and 4.
Te seasonal diference in stifness of the sandwich asphalt layer is signifcantly larger than that of the lean and unbound layers. For all three layers, the determined stifness was smaller in July than in April. It was by 54% smaller for the sandwich asphalt layer, by 6.6% smaller for the lean concrete layer, and by 2.5% smaller for the unbound layer.
Both in July and in April, the accelerometer readings from sledgehammer tests could be evaluated in a straightforward fashion, in order to compute times of fight, see Figure 14(c), while the determination of the time of fight was impossible based on the accelerometer readings from FWD tests, see Figure 14(d). Tis can be explained as follows. Te maximum force of the falling weight is much larger than that of the sledgehammer tests. In order to ensure that the pavement structure is not damaged during dynamic testing, the impact of the falling weight must be damped much more than that of the sledgehammer. Te wave front of a sledgehammer test is, therefore, much sharper than that of an FWD test, compare Figures 14(e) and 14(f).
When comparing the accelerations recorded during the sledgehammer and FWD tests, see Figures 14(c) and 14(d), the following additional points appear to be interesting: (1) Te peak accelerations produced by the sledgehammer test near the surface are much larger than those obtained during FWD tests, but decrease much more rapidly with increasing depth. In the exemplary sledgehammer test of Figure 14(c), peak accelerations recorded by the sensor closest to the surface and by the deepest sensor amounted to some ±400 m/s 2 and some ±10 m/s 2 , respectively. In the exemplary FWD test of Figure 14(d), in turn, the same sensors recorded peak accelerations of some ±20 m/s 2 and ±8 m/s 2 .
(2) Te waves caused by the sledgehammer had a basefrequency of 1 kHz to 2 kHz. Tis is by two orders of magnitude larger than the base-frequency of the waves caused by the falling weight. It ranges from 0.01 kHz to 0.03 kHz. Note the diferent scales of the abscissas of Figures 14(c) and 14(d).
It is concluded that sledgehammer tests are signifcantly better suited for determination of layer stifness, based on measured times of fight and the theory of elastic wave propagation through isotropic media. Since sledgehammer tests are not standardized, the testing conditions and the reproducibility of the results are discussed next. Te mass of the hammer and the dimensions of both the steel plate and the rubber pad are given in Figure 14(a). Te hammer is swung from a vertical distance of approximately 50-70 cm above the road surface without exerting signifcant extra force so that the hammer falls primary under its own weight. Te metal-on-metal impact creates a dynamic excitation with a sharp wave front. Te stif rubber pad ensures full-face contact to the cleaned pavement surface and prevents the latter from being damaged. Te height of fall of the hammer is optimized in an iterative fashion by the operator, until acceleration signals are obtained which can be reliably measured by the buried sensors. Given that the tests can be repeated within seconds, this initial optimization takes a few minutes only. In general applications, the optimal height of fall will depend on the damping properties of both the rubber pad and the pavement layers, as well as the depth at which the sensors are installed. Te essential feature is the creation of a sharp wave front because this is needed for reliable quantifcation of the time of fight through the layer of interest. Regarding the tests which were carried out at the described testing site, signals with suitably sharp wave fronts could be reproduced very simply in a highly satisfactory fashion. Te limiting factor for the evaluation quality is, therefore, rather related to the sensors. An even higher data acquisition rate would have further increased the quality with which the time of fight, the wave speed, and the stifness of the cement-stabilized layer could be quantifed.

Conclusions and Future Outlook
One rigid and two fexible pavement structures were equipped, during their construction, with temperature sensors, accelerometers, and strain gauges. Te following conclusions are drawn from the experience gained with the installation of the sensors: (1) Te used Pt100 temperature sensors and IEPE accelerometers are suitable for installation in all types of pavement layers. Tey particularly withstood high temperatures and compaction forces during the construction of asphalt layers. (2) Sensor overload during hot-state roller compaction of asphalt layers was the main problem encountered with the strain gauges. In order to avoid such problems, it is recommended to install a steel dummy as a placeholder into hot asphalt layers, immediately after their construction and right before their compaction, to replace the dummy by the actual sensor right before the installation of the next layer, and to fll small gaps between the sensor and the asphalt by cement paste.
Te following conclusions are drawn from results of newly proposed sledgehammer tests and FWD experiments performed at the three feld-testing sites: (1) Strokes with a sledgehammer onto a metal plate, transmitted to the pavement structure via a rubber pad, are well suited for quantifcation of the time of fight of elastic waves through asphalt, cementstabilized, and unbound aggregate layers, as long as the individual layers directly underneath the hit surface position are in full-face contact. In such cases, stifness quantifcation of individual layers is possible using the theory of propagation of elastic longitudinal waves through isotropic media. (2) Regarding rigid pavements, sledgehammer tests are capable of detecting curling-induced partial loss of contact of concrete slabs from lower layers by which Structural Control and Health Monitoring they are supported. Loss of layer-to-layer contact underneath the falling weight signifcantly increases measured defections during FWD testing. (3) Regarding fexible pavements, seasonal variations of FWD results can be primarily traced back to temperature-induced stifness variations of asphalt layers. Te other unbound and bound layers were found to exhibit signifcantly smaller stifness variations.
Tese conclusions provide motivation for the following future studies: (1) Data from feld testing together with results from laboratory characterization of the stifness of bound layers of pavement structures will provide a valuable database for the assessment of software which is designed to back-calculate layer stifness from defections measured during FWD tests. (2) Performing sledgehammer tests and FWD experiments repeatedly at the frst feld-testing site, during the morning of a day with signifcant solar heating of the surface of the concrete slab, will allow for studying the evolution of slab curling and its infuence on surface defections measured during FWD testing. (3) Large-scale application of the proposed approaches to real-time pavement monitoring will provide valuable information for the assessment of pavement performance and the optimization of maintenance strategies. Currently, the biggest challenges for such applications are the development of robust low-cost wireless sensors with autonomous energy supply. Te proposed approaches for quantifcation of the stifness of individual layers will be particularly valuable when used in combination with this sensor technology in the future.

Data Availability
Te experimental data used to support the fndings of this study are available from the corresponding author upon reasonable request.

Conflicts of Interest
Te authors declare that they have no conficts of interest.