Internal damage evaluation and remaining fatigue life prediction of asphalt pavements in highway widening projects

1 Introduction

Expressway widening has become the predominant strategy for capacity expansion in mature highway networks, particularly in China where approximately 10,000 km of early-built four-lane expressways require widening or reconstruction (Zhu et al., 2024; Wu et al., 2023). Longitudinal widening—constructing new lanes adjacent to existing carriageways via stepped joints—minimizes land acquisition and traffic disruption, yet creates a persistent structural vulnerability at the old-new overlap zone (Ahmed et al., 2023; Aripov et al., 2023). Historical maintenance records indicate that this region is the preferential locus for premature distress, dominated by reflection cracking and longitudinal joint deterioration (Famewo and Mehdi, 2025; Li et al., 2023; Shaikh et al., 2022).

Current pavement performance evaluation under JTG 5210-2018 relies predominantly on surface-oriented indicators, including the Pavement Quality Index (PQI), Pavement Structure Strength Index (PSSI), and Pavement Condition Index (PCI) (JTG 5210-2018, 2018). These indices, however, focus on functional and surface performance; even PSSI—intended to characterizing structural strength—relies on surface deflection measurements, providing only an empirical index that fails to resolve internal damage states. This arises from the fact that the deflection measurements are conducted at fixed intervals, which fails to quantitatively evaluate the structural load-bearing capacity reduction resulting from the longitudinal cracking at the joint zones between the existing and the newly widening pavements. This creates a critical diagnostic blind spot, particularly in joint zones where internal cracking initiates at the old-new interface and propagates upward, often reaching advanced states before surficial manifestation. The Hangzhou-Jinhua section of the G60 Shanghai-Kunming Expressway exemplifies this discrepancy: sections with high PCI ratings may still harbor substantial internal crack damage concentrated above widening joints, as revealed by field surveys.

Three-dimensional ground-penetrating radar (3D-GPR) has emerged as a powerful tool for internal damage screening, capable of resolving crack networks at decimeter-scale resolution (Tang et al., 2022; Cao et al., 2025; Bruno et al., 2022). However, GPR outputs remain largely qualitative in practice—anomaly maps indicate damage presence but do not provide mechanistic descriptors (e.g., crack volume fraction, damage density) required for structural capacity assessment (Elseicy et al., 2022). This quantification gap has motivated the use of the falling weight deflectometer (FWD) for internal condition evaluation. Yet current FWD practice predominantly relies on backcalculation frameworks—layered elastic theory (LET)-based programs such as EVERCALC and KGPBACK—to infer equivalent layer moduli as proxies for material performance (Mishra et al., 2022; Adigopula, 2022). The inverse problem inherent in such backcalculation is ill-posed, yielding non-unique solutions sensitive to measurement noise, layer thickness assumptions, and initial seed values (Wang and Zhao, 2026). More critically, modulus is an indirect indicator that convolves temperature effects, viscoelastic relaxation, and true structural damage into a single scalar, thereby precluding damage-specific characterization. Although some studies have linked FWD deflection basins to cracking states through dissipated energy or basin shape indices (Yao et al., 2025; Fu et al., 2022), these approaches remain phenomenological: they detect structural degradation signatures but cannot localize damage to specific layers, nor can they track damage evolution over time.

The prediction of remaining fatigue life in asphalt pavements has traditionally relied on two approaches. Empirical models, notably Miner’s linear cumulative damage rule (Miner, 1945; Ayalew et al., 2026) and the AASHTO mechanistic-empirical design framework, estimate life from external loading spectra and laboratory-derived material properties, offering computational simplicity for network-level management. Mechanistic-empirical extensions incorporate structural response analysis through finite element simulation or pseudo J-integral formulations (Sudarsanan and Kim, 2022), linking stress and strain states to material fatigue resistance with greater physical fidelity. More recently, data-driven methods leveraging artificial intelligence have emerged, correlating multi-source detection data with historical performance to bypass explicit mechanical modeling (Hanyu Zhang et al., 2024; Zhang et al., 2026). Despite these advances, all existing approaches share a fundamental limitation: the damage state within the widening joint zone is neither measured nor incorporated as an initial condition.

Empirical models implicitly assume an undamaged or uniformly degraded structure. Mechanistic-empirical methods rely on laboratory-calibrated damage evolution laws from pristine specimens, with no field-validated starting point. Data-driven models infer condition from surface indices or FWD deflections without resolving internal crack architecture. Consequently, predicted life and actual structural condition remain decoupled—a pavement may exhibit substantial internal damage while the model forecasts years of remaining service, or trigger premature rehabilitation based on surficial distress alone.

To bridge these gaps, this study establishes an integrated approach for assessing widening pavements, hierarchically combining 3D-GPR field screening, targeted coring, laboratory characterization, and viscoelastic continuum damage-based damage density modeling. The specific objectives are: (i) to quantitatively characterize both surface and internal damage distributions using 3D-GPR and targeted coring, with explicit spatial alignment and logical verification between methods; (ii) to experimentally determine the degradation of material properties (dynamic modulus, fatigue life) across the old-new interface, providing validation data for damage state characterization; (iii) to formulate a viscoelastic damage constitutive model using damage density (cracked-area ratio) as the quantitative descriptor of internal structural degradation; and (iv) to establish a remaining fatigue life prediction methodology that couples field-calibrated initial damage states with laboratory-determined damage evolution parameters, supporting the long-life pavement principle of hierarchical layer durability and enabling segment-based maintenance decision-making.

2 Field investigation and core sampling

2.1 Study site and pavement structure

The study focuses on the Hangzhou–Jinhua widening project of the G60 Shanghai–Kunming expressway. The original four-lane dual carriageway opened in September 2003 and was widened to eight lanes between January 2015 and September 2018. The widening section was constructed by splicing new Lanes three and 4 with the existing old pavement, followed by a uniform overlay. The original hard shoulder was repurposed as the sub-base course for Lane 3.

The pavement structures of the old and new sections are summarized in Table 1. The longitudinal joint was formed with a 50 cm stepped overlap, resulting in a composite structure with a surface course thickness differential between old and new sections, as shown in Figure 1.

TABLE 1. Pavement structure configurations.

Since 2022, damage intensity in Lanes 3 and 4 has consistently exceeded 20 m²/lane-km, significantly surpassing the average of 10 m²/lane-km. Longitudinal cracks are preferentially concentrated at the old–new pavement splice locations (Figure 2), accompanied by spalling and progressive structural loosening, indicating that the joint zone constitutes a persistent structural vulnerability.

The field investigation was conducted in two stages. Stage I entailed network-level condition screening, combining automated surface distress surveys with three-dimensional ground-penetrating radar (3D-GPR) scanning to identify anomalous segments and establish the spatial distribution of both manifest and latent damage. Stage II comprised project-level forensic investigation at selected locations, involving targeted coring and laboratory characterization to quantify material degradation and predict remaining fatigue life. This staged approach enables maintenance prioritization based on residual structural capacity rather than surface distress indices alone.

2.2 Pavement surface distress survey

A comprehensive pavement condition survey was conducted over the K289-K314 subsection (single downstream direction, Lanes three and 4) using a CiCS multifunctional pavement condition assessment vehicle, as shown in Figure 3. The integrated system enables simultaneous acquisition of five functional indices at survey speeds of 30–100 km/h, with sub-millimeter detection precision and decimeter-scale longitudinal sampling resolution. The survey outputs include pavement distress rate (DR), rutting depth (RDI), ride quality index (RQI, derived from IRI) and surface texture depth. Among these, PCI (derived from DR%), RDI, and RQI serve as the primary diagnostic indicators for surficial functional degradation and precursor identification of structural distress. The multi-parameter survey strategy captures both the manifest surface condition and the latent structural anomalies, providing the empirical basis for correlation with 3D-GPR internal damage screening (Figure 4) and subsequent layer-specific mechanistic investigation.

2.3 Three-dimensional GPR internal damage screening

Three-dimensional GPR scanning was conducted over the K289-K314 subsection (Lanes three and 4) using a step-frequency antenna array system (3D-Radar, Norway) operating at 200–3000 MHz with 20 channels, achieving 10–25 mm vertical resolution and 5 cm horizontal sampling at surveyspeeds up to 55 km/h. The 180 cm detection swath and 120 cm penetration depth enable fullvolumetric coverage of the asphalt layer stack and upper CTM base. Raw data were processedthrough time-zero correction, migration imaging, and 3D voxel reconstruction to resolve internalstructural anomalies.

2.4 FWD testing across the widening joint

Falling weight deflectometer (FWD) testing was conducted using an SHN-FWD-MV device (Figure 5) equipped with nine geophones (S1-S9) positioned at radial distances of 0, 200, 300, 450, 600, 900, 1200, 1500, and 2400 mm from the load center. Three load levels (50, 70, and 90 kN) were sequentially applied at each measurement point to capture both linear and nonlinear structural responses under varying stress magnitudes.

Measurement points were arranged on transverse cross-sections across the widening joint zone. Five positions were selected: Point 5 (P5) at the Lane 3 center, P20 30 cm toward Lane three from the joint, P33 30 cm toward Lane four from the joint, P46 at the Lane four left wheel path, and P58 at the hard shoulder. The hard shoulder served as low-stress references. All points were located on the same transverse cross-section to ensure comparability of deflection basin parameters. These positions were paired with the core extraction array (Section 2.5) to correlate FWD deflection basins with laboratory material degradation.

2.5 Cross-sectional coring array

Core sampling was conducted at a typical cross-section selected based on the integrated surface-GPR screening results, prioritizing segments with both pronounced surface distress and GPR anomalies. The transverse layout spans five extraction points (Figure 6): Lane 3 center (14#), 30 cm from the crack toward Lane 3 (15#), 30 cm from the crack toward Lane 4 (16#), Lane four left wheel path (17#), and the hard shoulder (18#). This ensures a minimum of three specimens within the stepped joint influence zone for direct comparison of joint-proximate degradation against intact and reference materials.

Φ 150 mm cores were extracted to approximately 40 cm height, encompassing the full asphalt layer stack. Prior to laboratory testing, specimens were visually documented and categorized by structural condition for dynamic modulus and fatigue characterization (Sections 3.1–3.2), with damaged specimens reserved for damage mechanism validation.

3 Laboratory characterization

3.1 Dynamic modulus test and viscoelastic master curves

Specimens were prepared following a layer-specific cutting protocol with thickness governed by nominal maximum aggregate size. Field cores extracted from the five transverse positions (Section 2.5) were first trimmed to isolate individual structural layers, then machined into cylindrical specimens: Φ100 mm × 40 mm for the SMA-13 layer, and Φ150 mm × 60 mm for both AC-20 and AC-25 layers, with dimensions matched to respective layer thicknesses and aggregate gradations, as shown in Figure 7.

Dynamic modulus tests were performed on the cylindrical specimens using a DTS universal testing system in accordance with JTG3410-2025 (JTG3410-2025, 2025). Tests were performed at four temperatures (−10, 5, 20, 30 °C) and five loading frequencies (10, 5, 1, 0.5, and 0.1 Hz) under indirect tensile loading. The dynamic modulus for each temperature-frequency combination was computed as the average of the final five loading cycles. To ensure comparability of material responses across thermal and loading conditions, the test sequence followed a standardized protocol: temperatures progressed from low to high, and at each temperature, frequencies descended from high to low. Each temperature condition required a minimum 8-h thermal conditioning period, with 100 cycles per frequency and a 2-min rest interval between successive frequencies.

3.2 Indirect tensile (IDT) fatigue test and damage evolution parameters

Indirect tensile (IDT) fatigue tests were performed on specimens prepared from the same field core positions and layer-specific configurations as the dynamic modulus tests, as shown in Figure 8. A stress-controlled mode was adopted at 15 °C ± 0.5 °C and 10 Hz, with a haversine waveform and no rest period. Prior to formal loading, a pre-conditioning contact stress of 35 kPa was applied through curved loading strips to ensure uniform load distribution and prevent local crushing. The loading stress level was then determined from the stress–horizontal radial resilient strain curve, targeting a resilient strain of approximately 200 microstrain (0.02%) to constrain fatigue life within the range of 10⁴ to 10⁶ cycles.

During testing, load and horizontal deformation were continuously recorded to calculate stress, strain, and stiffness modulus in real time. Horizontal permanent strain was measured using a linear variable differential transformer (LVDT) and fatigue life was recorded. Fatigue failure was defined as the point at which the horizontal stiffness modulus degraded to 50% of its initial value, and the corresponding number of cycles was recorded as the fatigue life.

4 Viscoelastic damage theory and remaining life prediction framework

4.1 Viscoelastic constitutive model for asphalt mixtures

Asphalt mixtures are thermorheologically simple materials exhibiting pronounced time- and temperature-dependent mechanical responses (Fan et al., 2022; Sun et al., 2026). Under small-strain conditions, their behavior is described by linear viscoelasticity theory (Gopalsamy, 2023).

To capture the relaxation behavior and frequency-dependent stiffness, the generalized Maxwell model is adopted, comprising a solitary equilibrium spring in parallel with M Mx elements; each element consists of a spring and dashpot in series. The total stress tensor σ i j is given by the Prony series as:where K ∞ and G ∞ denote the long-term equilibrium bulk and shear moduli (MPa); K m and G m are the relaxation bulk and shear moduli (MPa) of the m -th Maxwell element; ε k k and e i j are the volumetric and deviatoric strains; ε k k m · v i and e i j m · v i are the corresponding viscous strains of the mth element; and δ i j is the Kronecker delta.

σ i j = K ∞ ε k k δ i j + 2 G ∞ e i j + ∑ m = 1 M K m ε k k − ε k k m · v i δ i j + 2 G m e i j − e i j m · v i (4.1)

The constitutive relation in Equation 4.1 presumes small deformations and intact material. In service, however, microcrack initiation and propagation progressively degrade material stiffness and alter stress redistribution within the pavement structure. To capture this degradation, continuum damage mechanics (CDM) is employed (Zhang et al., 2025). A scalar damage density ξ is introduced, defined as the ratio of cracked area to total cross-sectional area:where A c is the cracked area (mm²) induced by voids, defects, and fractures, and A T is the total cross-sectional area (mm²).

ξ = A c / A T (4.2)

Based on the strain equivalence hypothesis in CDM, the effective stress and strain in the undamaged configuration relate to the apparent quantities through: where σ i j and ε i j denote the effective stress (MPa) and strain tensors on the undamaged configuration, respectively; σ i j A and ε i j A denote the apparent stress (MPa) and strain tensors on the damaged configuration measurable in laboratory tests.

σ i j A = σ i j 1 − ξ (4.3)

ε i j = ε i j A 1 − ξ (4.4)

Substituting Equations 4.2–4.4 into Equation 4.1 yields the viscoelastic damage constitutive model for asphalt mixtures:

σ i j = 1 − ξ 2 K ∞ ε k k A δ i j + 2 G ∞ e i j A + 1 − ξ ∑ m = 1 M K m ε k k A 1 − ξ − ε k k m · v i δ i j + 2 G m e i j A 1 − ξ − e i j m · v i (4.5)

The viscoelastic damage constitutive model (4.5) is implemented in finite element analysis via the relaxation modulus E t, which characterizes stress decay under constant strain and is expressed in Prony series form as:where E ∞ is the long-term equilibrium modulus (MPa), E m and τ m are the relaxation modulus (MPa) and relaxation time of the mth Maxwell element, respectively, and M is the total number of elements.

E t = E ∞ + ∑ m = 1 M E m ⁡ exp − t τ m (4.6)

To characterize the thermorheological behavior of asphalt mixtures, dynamic modulus tests were conducted at various temperatures and frequencies. The time-temperature superposition principle (TTSP) was employed to shift the modulus data to a common reference temperature of 293.15 K (20 °C). The temperature shift factor described by a quadratic polynomial:where T is the test temperature (K), a = 6.03 × 10⁻⁵, b = −0.16, and c = 41.72.

log α T = a T 2 + b T + c (4.7)

The dynamic modulus master curve was constructed by plotting the shifted modulus data against the reduced frequency f r = f r · a T, and fitted by the sigmoidal function:where E * is the dynamic modulus (MPa), δ is the lower asymptote, α is the logarithmic span, η and γ are shape parameters controlling the transition location and steepness, respectively, and f r is the reduced frequency in Hz. The fitted coefficients are δ = 1.21, α = 3.20, η = 0.043, and γ = 0.505.

log E * = δ + α 1 + e η − γ ⁡ log f r (4.8)

4.2 Remaining fatigue life prediction

Under repeated traffic loading, microcrack initiation and propagation constitute the fundamental mechanism of fatigue damage accumulation and structural performance degradation in asphalt pavements. Existing viscoelastic fatigue analyses are susceptible to interference from material viscous dissipation, making it difficult to accurately characterize true crack evolution. This section establishes a quantitative linkage between damage density, crack evolution, and remaining fatigue life based on viscoelastic fracture mechanics and continuum damage mechanics (CDM), introducing the pseudo J-integral and pseudo-strain concepts to eliminate viscoelastic relaxation dissipation.

The crack evolution process is described by a damage-density-based pseudo J-integral Paris law:where ξ is the damage density, N is the number of load cycles, A and n are Paris-law material parameters, and Δ J R is the incremental pseudo J-integral (MPa), representing the energy release rate that drives damage evolution per loading cycle. The pseudo J-integral is thermodynamically equivalent to the energy release rate associated with crack propagation in classical fracture mechanics.

d ξ d N = A Δ J R n (4.9)

Based on the energy balance principle in CDM, the pseudo J-integral is expressed in terms of damage density and stress state. For asphalt mixtures under cycle loading, the crack evolution equation becomes:where σ A is the applied stress amplitude (MPa), E 0 is the initial undamaged modulus (MPa), and H is a geometry-dependent factor. n is the stress-dependent exponent, N is the number of loading cycles.

d ξ d N = A H 4 σ A 2 E 0 3 + ξ 1 − ξ 3 n (4.10)

The annual equivalent single axle loads (ESALs) are computed as:where N _*e* is the annual ESALs, N _*i* is the annual traffic volume of vehicle class i, and α _*i* is the ESAL conversion coefficient. The design-lane ESALs N_d are obtained by directional distribution:

N e = ∑ N i α i (4.11)

N d = 0.6 N e (4.12)

Based on traffic data from 2021 to 2025, the design lane ESALs increased from 7.2186 × 10⁶ to 7.9016 × 10⁶ per year, yielding a compound growth rate of 2.29%. This calibrated rate, rather than the empirical 10% assumption used in early design, is adopted for subsequent life prediction.

The remaining service life t is back-calculated from the residual load-bearing capacity:where N_r is the remaining ESAL capacity (10⁴ cycles) from the damage model, N ₁ = 790.16 × 10⁴ cycles/year is the annual design-lane ESALs in 2025, g = 2.29% is the annual traffic growth rate, and t is the equivalent remaining service life (years).

t = ln N R · g N 1 + 1 ln 1 + g (4.13)

For a remaining ESAL capacity of 56.5 × 10⁴ cycles, substituting N _*R* = 56.5 × 10⁴, N ₁ = 790.16 × 10⁴ cycles/year, and g = 2.29% into the rearranged Equation 4.13 yields t = ln (56.5 × 0.0229/790.16 + 1)/ln (1.0229)≈0.07 years. This corresponds to approximately 0.07 years of remaining service life.

The remaining service life t is back-calculated from the residual load-bearing capacity using Equation 4.13. It is emphasized that this prediction is a model-based estimate contingent on the adopted material characterization, traffic characterization, and field-calibrated initial damage state. Actual field performance may deviate due to environmental variability, spatial heterogeneity in material properties, and uncertainties in traffic projection.

5 Results and discussion

5.1 Surface and internal damage distribution

The network-level survey of the downstream K289–K314 section (51.44 lane-km) revealed that crack-type distresses overwhelmingly dominated the surface condition, constituting 97.91% of the total damage area (1,191.91 m², 23.12 m²/lane-km). Longitudinal cracks were predominantly concentrated in Lane 3 (the widening joint zone), with peak severities at K294–K296, K299–K300, K309–K310, and K312–K313. Concurrently, the 3D-GPR survey identified 252.85 m² of internal damage (4.86 m²/lane-km), with crack-type defects again dominant at 93.3%. Figure 9 reveals a pronounced transverse gradient: Lane 3 consistently exhibits the highest damage intensities in both surface and internal records. However, the longitudinal distributions are not synchronous—internal damage peaks at K298–K299, K305–K306, K309–K310, and K312–K313 are spatially offset from surface distress maxima, indicating that severity rankings based on surface inspection alone would misidentify critical structural segments.

The K304–K306 subsection (2 km, embankment section) was selected through a tri-criterion screen: distance from ramp hubs, concurrence of internal and surface distress exceeding section averages, and exclusion of cut sections. Satisfying all three, this segment exhibits pronounced damage in both surface and internal records, providing a representative basis for mechanistic investigation, as shown in Figure 10.

A refined 3D-GPR survey of K304–K306 resolved internal damage architecture at decimeter-scale resolution. Over the 2-km segment, 34.15 m² of internal damage was identified (17.08 m²/lane-km), far exceeding the general survey mean of 4.86 m²/lane-km. Crack-type defects accounted for 93.2%.

Despite PCI values meeting 100% compliance thresholds, significant internal damage persists along the surveyed section, indicating that PCI alone is insufficient for comprehensive pavement health characterization. A pronounced spatial mismatch between internal and surface distress is evident: at K305 + 500 (Lane 3), substantial internal cracking is detected with relatively minor surficial manifestation, whereas at K304 + 800, surface damage exceeds internal damage. This inconsistency demonstrates that internal and surficial deterioration do not develop synchronously; in some locations, subsurface damage progresses ahead of surface expression, while in others, surficial distress may dominate. Maintenance strategies based solely on surface-visible criteria therefore cannot adequately support structural decision-making. Layer-wise damage assessment, incorporating material deterioration at different pavement depths, is essential for reliable characterization of structural health.

5.2 Dynamic modulus and viscoelastic characterization

Dynamic modulus master curves for four representative materials (old pavement SMA-13, new pavement SMA-13, AC-20, and AC-25) are presented in Figure 11. At the reference temperature of 20 °C and 10 Hz, the dynamic modulus results demonstrate typical temperature–frequency sensitivity: higher moduli at low temperatures and high frequencies, with significant reduction at elevated temperatures and low frequencies. This trend is corroborated by the master curves, where increasing reduced frequency corresponds to rising dynamic modulus. The Maxwell parameters extracted from the master curves exhibited low fitting errors (Table 2), confirming reliable parameter extraction. Table 2 lists the equilibrium modulus (E ∞), instantaneous modulus (E ₍₀₎), and the corresponding bulk and shear moduli (K ₍₀₎ and G ₍₀₎) converted using a Poisson’s ratio of 0.28. These parameters were subsequently incorporated into the generalized Maxwell viscoelastic constitutive model described in Section 4.1.

TABLE 2. Dynamic modulus master curve fitting errors and moduli.

In terms of material ranking, AC-25 exhibits the highest E ₍₀₎, followed by AC-20, with new SMA-13 showing the lowest; old SMA-13 falls between AC-20 and new SMA-13. The old SMA-13 also differs from new SMA-13 in its contribution from the long relaxation-time branch, indicating that long-term aging alters the viscoelastic spectrum distribution.

The Prony-series parameters, comprising 11 Maxwell branches with relaxation times τ _*i* and corresponding stiffness moduli E _*i* for each asphalt mixture, are listed in Table 3.

TABLE 3. Prony-series parameters for the generalized Maxwell model.

5.3 FWD deflection characteristics across widening joint

FWD deflections at five transverse positions under sequential load levels (50, 70, and 90 kN) are presented in Figure 12. While deflections increased monotonically with load at all positions, the most pronounced structural anomaly was concentrated at the widening joint. Under 90 kN, peak deflections at P33 (147.3 μm) and P20 (124.4 μm) reached 1.96× and 1.33× the Lane 4 benchmark (P46, 75.0 μm), and 1.58× and 1.33× the Lane three benchmark (P5, 93.5 μm), respectively. The deflection attenuated sharply away from the joint, recovering by 49.1% from P33 to P46 and by 24.8% from P20 to P5; given that P20 and P33 lie only 30 cm from the joint interface and P46 and P5 have returned to sound-pavement benchmarks, the lateral influence zone is constrained to less than 1.0 m.

The simultaneous elevation of deflections on both sides of the joint relative to their respective lane benchmarks indicates structural damage at the interface, likely involving through-thickness cracking, base–subbase debonding, or localized voiding. These elevated deflections reflect reduced structural stiffness and material degradation, yet surface deflections alone cannot quantify the degree of base modulus reduction, remaining structural capacity, or fatigue life consumption; these require further mechanistic-empirical analysis and laboratory validation against extracted cores.

5.4 Fatigue and damage evolution parameters

IDT fatigue tests were performed on the extracted cores. Table 4 summarizes the test results, and Table 5 presents grouped statistics by material type. The hard shoulder (18#) and Lane 4 (17#) serve as low-stress and intact new pavement references, respectively.

TABLE 4. IDT fatigue test results and fitted parameters.

TABLE 5. Grouped fatigue life statistics by material type.

SMA-13 surface mixtures exhibited the largest variability (CV = 58.3%), with fatigue life ranging from 15,425 to 130,401 cycles. This substantial dispersion reflects the combined influence of long-term aging, localized joint damage, and construction-induced heterogeneity. AC-20 intermediate course specimens demonstrated superior fatigue performance relative to AC-25 binder course, with mean fatigue life of 35,398 versus 9,634 cycles. Specimen 15#4 (AC-20, joint zone) achieved the highest fatigue life among all tested specimens (80,321 cycles), whereas AC-25 specimens at 17# (Lane 4) and 14#5 (joint zone) failed prematurely at 433 and 3,400 cycles, respectively, indicating pre-existing internal damage or material embrittlement.

The damage evolution parameters A and n in the Paris law model were calibrated from the S–N curves. For the joint-zone mixture, A = 2.3 × 10⁻⁶ and n = 2.14 exceed typical values for intact asphalt (A ≈ 1.1 × 10⁻⁶, n ≈ 1.85), indicating accelerated damage initiation and heightened stress sensitivity near the joint. For the hard shoulder reference, A = 1.0 × 10⁻⁶ and n = 1.82. The progressive increase from reference to joint-zone material reflects cumulative aging and construction-induced defects, and these parameters were directly embedded in the damage constitutive model.

5.5 Remaining fatigue life prediction

5.5.1 Traffic loading characterization

For the K304-K306 subsection, design-lane ESALs increased from 7.2186 million to 7.9016 million per year during 2021–2025 (Table 6), yielding a compound growth rate of 2.29%. This calibrated rate, rather than the empirical 10% assumption used in early design (Luo et al., 2023), is adopted for subsequent life prediction. The 2025 annual design-lane ESAL value of 7.9016 million cycles is adopted as the baseline input N ₁ in Equation 4.13 for remaining-life prediction. The 2025 axle load spectrum (Table 7) reveals that heavy trucks dominate: Class 6 accounts for 55.43%, with Class 4 and 5 at 12.31% and 12.30%, respectively; these three classes total 80.04%.This indicates that heavy-truck standard axle contribution, particularly Class 6, controls remaining life prediction rather than total mixed traffic volume. The remaining fatigue life was computed using the framework in Section 4.2 with these inputs.

TABLE 6. Traffic data from 2021 to 2025.

TABLE 7. Main components of 2025 axle load spectrum.

5.5.2 Spatial variation of remaining life

Damage density and remaining fatigue life exhibit sharp transverse gradients across the joint zone, as shown in Figure 13. Under the 2025 traffic loading scenario, damage concentration and life decay are most pronounced at joint-proximate positions, with base-course remaining life at the joint edge converting to approximately 0.07 years against substantially higher values at remote positions. This pronounced spatial disparity stems from material mismatch, interface bonding degradation, differential deformation, and repeated heavy loading.

Cross-sectional analysis further confirms that the base course governs structural performance degradation, as its damage density consistently exceeds surface-course values and its remaining life is lowest across all positions. General positions away from the joint exhibit markedly higher converted remaining years under identical traffic growth rates, indicating that life discrepancies originate from localized joint distress rather than uniform traffic loading. Consequently, remaining life assessment should prioritize joint-proximate positions, with the base course serving as the critical layer.

IDT fatigue tests provide mechanistic interpretation: joint-adjacent SMA-13 and AC-20 specimens exhibit weakened fatigue resistance, entering stable crack propagation at smaller damage accumulation. Field back-calculation further demonstrates that base-course damage and life decay are most sensitive near the interface. These mutually consistent findings indicate that performance deterioration at the old–new interface arises from a dual mechanism: reduced material crack resistance coupled with amplified structural stress concentration.

From an engineering perspective, the joint vicinity should be prioritized for maintenance and overlay design. Heavy-axle loading magnifies base damage propagation and surface fatigue cracking risks. Structural rehabilitation should emphasize interface bonding enhancement, crack-resistant layer placement, and differential deformation control to retard crack extension from the joint into traffic lanes.

6 Conclusion

• PCI-based assessments fail to capture internal damage in widening joint zones, where internal and surface distress exhibit pronounced spatial asynchrony. Maintenance strategies relying solely on surface-visible criteria risk deferred intervention until structural capacity is critically compromised.
• An integrated hierarchical approach was established for damage-specific assessment in widened pavements, comprising three progressively refined stages: 3D-GPR screening for internal damage detection, FWD testing for structural response characterization at the joint interface, and laboratory characterization for material degradation quantification. This staged framework enables progressive refinement of joint-zone damage characterization from network-level anomaly identification to project-level mechanistic validation.
• Joint-proximate materials exhibit a dual degradation mechanism: the AC-25 binder course shows embrittlement (high stiffness concurrent with low fatigue life), while SMA-13 and AC-20 specimens demonstrate reduced fatigue resistance with earlier onset of stable crack propagation. Damage density and remaining fatigue life exhibit sharp transverse gradients across the joint zone, with the base course serving as the critical layer. At 30 cm from the joint centerline, base-course remaining life is approximately 0.07 years under 2025 design-lane ESALs.
• Heavy-axle loading amplifies base-course damage propagation and surface fatigue cracking risks at the old-new interface. Heavy-load adaptability assessment is therefore essential for widening pavements.

Generative AI statement

The author(s) declared that generative AI was not used in the creation of this manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Footnotes

Edited by: Peng Lin, Directorate-General for Public Works and Water Management, Netherlands

Reviewed by: Yangming Gao, Liverpool John Moores University, United Kingdom

Hanyu Zhang, Liverpool John Moores University, United Kingdom