Journal of Transportation Engineering, Part B: Pavements

With the continuous advancement in data-acquisition devices, computer vision techniques, and machine-learning (ML) algorithms over the past decades, artificial intelligence (AI) technology has increasingly been applied to research and practice in pavement engineering and related fields. This paper is aimed at systematically synthesizing the state-of-the-art in applying AI algorithms and techniques to various areas of pavement management. To achieve this goal, the authors reviewed major work published in archival journals from 2015 to 2020. Key findings from the review are synthesized and presented based on three broad categories of pavement management activities: distress evaluation, performance modeling, and maintenance and rehabilitation (M&R) programming. Most of the reviewed studies have achieved positive and/or promising results, proving the effectiveness of leveraging AI algorithms for pavement management. Distress detection and classification are found to be the areas that attracted the most attention in terms of applying AI techniques and algorithms. In contrast, applying AI techniques to M&R programming represents a major research gap. Based on the review, it can be concluded that AI algorithms have made noticeable achievements in most activity areas of pavement management, although some major research gaps remain to be filled.

For the safe operation of vehicles, pavement should provide adequate skid resistance, which can be achieved by using high polishing–resistant aggregate in wearing courses. However, supplying high-quality aggregate is not always feasible due to high transportation costs. For this reason, a method called gritting was adapted to meet the Highway Technical Specification (HTS) of Turkey in 2013. According to the method, for certain parts of the country, the wearing course can be constructed with local aggregates that have minimum polished stone value (PSV) of 40 ($\mathrm{PSV}\ge 40$), but, in this case, the surface must be covered with a high polishing–resistant aggregate ($\mathrm{PSV}\ge 50$), after the rollers’ first pass. The objective of this study was to improve the present gritting method by investigating the effect of grit parameters on pavement performance under real traffic conditions. In this regard, during its construction, the wearing course of O-51 Highway was gritted with different aggregate types (slags and natural), sizes (1–3; 1–5 mm), spreading amount (1.5; 2; $2.5\mathrm{kg}/{\mathrm{m}}^2$), and spreading time (before and after the first pass of a roller) on eight test sections. Then, the macrotexture and skid resistance performance of these sections were evaluated under real traffic and environmental conditions for longer than 4 years. Changes in surface texture and skid resistance with respect to traffic were determined for each section. The results showed that higher skid resistance values were obtained at the sections gritted with metallurgical slags. Additionally, the sections gritted with 1–5 mm aggregates had better skid resistance than those gritted with 1–3 mm, while the change in mean texture depths were not very significant.

The cracking resistance of a high modulus asphalt mixture (HMAM) containing reclaimed asphalt pavement (RAP) was studied in this paper. Asphalt binder #30 and a high modulus additive were used to prepare HMAM containing RAP. Mixtures were designed by the Marshall mix design method. To improve the mixing of RAP and virgin material, the regeneration effects of the improved and traditional regeneration methods were compared. Different HMAM was obtained by changing the RAP content, the dosage of rejuvenator, the type of asphalt and the addition method of the high modulus additive. A dynamic modulus test was applied to the test mixtures, and splitting tensile strength and relaxation spectrum parameters were used to evaluate the cracking resistance of HMAM. The Glover–Rowe (G-R) parameter and R-value, which are related to aging, were also used to evaluate the cracking resistance of HMAM. The results indicate that the mixing effect of HMAM obtained by the improved regeneration method is better than that of HMAM obtained by the traditional regeneration method. Relaxation spectrum parameters can evaluate the cracking resistance of HMAM well. The R-value has a good correlation with relaxation spectrum parameters, and it shows a good linear correlation with the G-R parameter. This study provides a reference for research on the cracking resistance of HMAM and contributes to the application of RAP.

The objective of this study was to assess the performance of full-scale cold in-place recycled (CIR) asphalt sections using accelerated pavement testing (APT). A balanced mix design approach was followed to optimize the binder contents of CIR mixtures, and these mixtures were subsequently used to construct three full-scale sections (7.6 by 3.7 m) at Rowan University’s Accelerated Pavement Testing Facility. Foamed asphalt was added in varied contents: 2%, 3%, and 4% by total mix weight. All CIR mixtures were prepared at a constant water content of 3%. Each full-scale section was instrumented with asphalt strain gauges, pressure cells, and thermocouples to evaluate the structural responses within each section. A heavy vehicle simulator (HVS) was utilized to apply accelerated loading on each full-scale pavement section. A truck tire was utilized to apply a 40-kN load while an aircraft tire was utilized to apply a 100-kN load. As accelerated loading was applied, a number of field tests and visual inspections were performed to determine (1) permanent deformation using a surface profiler, (2) the structural integrity using a heavy weight deflectometer before and after APT, and (3) cracking potential by assessing stress and strain responses. The results showed that the CIR section with 2% binder content presented the best rutting performance under truck loading and the highest rutting susceptibility under aircraft loading. Conversely, the CIR section with 3% binder content presented the highest cracking resistance under accelerated truck and aircraft loading.

Interlocking Concrete Block Pavement (ICBP) is a pavement type consisting of a discrete wearing layer with a unique load transfer mechanism. The main objective of the present study was to understand the influence of geometric properties of Interlocking paver block (IPB) on the deflection behavior of ICBP. The geometric parameters considered in the study include the block laying pattern, laying angle, laying thickness, and laying shape. Plate load tests were used to evaluate the deflection of the ICBP, and numerical modeling was carried out using PLAXIS Software. The PLAXIS software was validated using experimental and case study deflection test results. The optimization and significance of the geometric parameters were analyzed using Design Expert software. The test result showed that less deflection was experienced for the ICBP when laid in a herringbone pattern, 90° angle, 80–100-mm thickness, and zigzag shape.

A series of large-scale model experiments were carried out on different geogrid- and geocell-reinforced base courses to evaluate realistic base-layer coefficients to design flexible pavements. The placement depth of reinforcement was first determined under monotonic loading on designed unreinforced pavement sections over a very weak subgrade (resilient modulus of 10 MPa) prepared in a $2.25\text{-}{\mathrm{m}}^3$ size test chamber. A structural support offered by the reinforcement alone in the base layer was quantified through the modulus improvement factor (MIF) for varying subgrade conditions. The MIF values ranged between 1.5 and 3.5 for geogrid-reinforced bases and 1.4 and 5.0 for geocell-reinforced base layers placed over different subgrade conditions. Further, a range of laboratory-produced MIF values and semiempirical mechanistic design principles were used to analyze the flexible pavements to get the base-layer coefficients for various geogrid- and geocell-reinforced pavements. In this analysis, the traffic was considered from 2 million to 150 million equivalent single-axle loads, subgrade resilient modulus ($M_{rs}$) from 10 to 85 MPa [corresponding a California bearing ratio (CBR) from 1% to 8%], and MIF from 1.2 to 3.5 for geogrids and 1.2 to 5.0 for geocells. A new set of apt base-layer coefficients for geogrid- and geocell-supported base layers was developed through a systematic analysis. The layer coefficients for geogrid-reinforced bases ranged from 0.15 to 0.35 and 0.175 to 0.425 for geocell-reinforced base layers. The proposed models were validated with an as-built pavement section from Montana state and the available design approaches. The proposed design approach has reduced the thickness of a geogrid-reinforced base layer by about 40%, and it is 50% for the geocell-reinforced base layer.

Utilization of reclaimed asphalt pavement (RAP) at lower production temperatures is a low-carbon and sustainable technology in the asphalt industry. This study provides a forensic evaluation of a low-volume road in Florida that used 100% RAP cold mix with a rejuvenator. Two field distress surveys were completed after 7 and 22 months of service, and the effects of staged construction and edge compaction on the road conditions were evaluated. Alongside the field survey, the asphalt mixture’s properties were analyzed in terms of field cores’ volumetric properties, Cantabro loss, Hamburg wheel tracking, indirect tensile resilient modulus, and fracture tests. The ride quality was good, but the main distresses were weathering and raveling. It was determined that the field cores had high air voids, resulting in high Cantabro loss, which was consistent with field raveling observations. The Hamburg wheel-tracking test results showed that rutting resistance and moisture resistance were good for a 100% RAP cold mix with rejuvenator. To improve field performance, it is recommended to mix and compact a 100% RAP cold mix with rejuvenator on the same day, as well as improve compaction on the edge to achieve higher in-place density.

The use of controlled low-strength material (CLSM) to backfill the cut-and-cover pipelines trench under busy streets is getting popular nowadays. Due to the slow hydration rate of cement in CLSM, extra waiting hours are usually needed before it can develop sufficient strength to sustain traffic. To shorten the waiting time, this paper presents a method that can provide a good immediate bearing capacity for the CLSM backfilled trench right after completing narrow trench work. This modified method adds extra gravel to the top of the fresh CLSM, which is already in the trench. The gravel is then pushed down and mixed with CLSM and forms a strong upper layer for the CLSM trench. It can provide a good immediate bearing capacity to endure traffic weight. To verify the suitability of this modified CLSM trench filling method, a numerical analysis method [discrete-element method (DEM)] was adopted in this study, along with the laboratory model and field tests. The DEM results showed that high-friction material (gravel) on the upper portion of the trench could form a strong layer that sustains a large share of the surcharge load. So, it could result in a significant pressure reduction for the lower portion of the trench and around the buried pipes, even though the CLSM had not hardened yet. Such a result was confirmed by the laboratory model test and the full-scale field test, in which it showed an increase in the initial load-bearing capacity and a much smaller settlement on the surface of the CLSM backfilled trench.

In this study, 18 different unbound aggregates were collected from various locations of Idaho and several characterization tests were conducted in the laboratory. The resilient modulus of these collected aggregates was measured with the help of repeated load triaxial (RLT) testing. The results of the laboratory characterization tests were used to develop a regression model for the Mechanistic Empirical Pavement Design Guide (MEPDG) model parameters (e.g., $k_1$, $k_2$, $k_3$). The $k_1$, $k_2$, and $k_3$ models were correlated to compaction and gradation characteristics of the test aggregates. The proposed models can be used to predict the MEPDG model regression coefficients for unbound layers in AASHTOWare Pavement ME Design software. The resilient modulus measured in the laboratory and the resilient modulus predicted with these $k_i$ models are found to be in good agreement.

The objective of this study was to develop a deep neural network (DNN)-based approach to predict the overlay thickness of asphalt pavements using deflection bowl parameters measured with a falling weight deflectometer and the estimated traffic. The scope of the effort was two-fold: (1) Develop a DNN to determine the overlay thickness using deflection and traffic parameters; and (2) train and test the model’s performance. Over 1,300 datapoints from datasets collected from different geographical locations, such as the USA, Canada, and India, were used to train, validate, and test the performance of the model, so that the insufficiency of the historical data could be overcome. The developed network architecture was efficient in predicting the overlay thickness with a reasonably high coefficient of determination (${\mathrm{R}}^2>80\%$). The Morris method of sensitivity analysis was performed to understand the importance of each input parameter in predicting the asphalt overlay thickness. The absolute mean and standard deviation of elementary effects of individual parameters were in close approximation, indicating that each input variable contributed to the overlay thickness prediction. It is noteworthy that the developed model eliminates resource intensive methods of quantifying the pavement thickness, such as cutting and coring of the pavement and rigorous back-calculation processes, thus helping in the prediction of overlay thickness at the project level. Overall, the developed DNN model can help roadway agencies in making rapid and appropriate decisions pertinent to pavement maintenance, rendering it as one of the quality control toolkits easily adoptable during pavement design and operation phases.

Automated or semi-automated pavement condition data collection is replacing manual data collection in many state and local highway agencies due to its advantages of reducing labor, time, and cost. However, the practical experience of highway agencies indicates that there are still data quality issues with the pavement condition data collected using existing image and sensor-based data collection technologies. This study aims to investigate the implementation experiences and issues of automated or semi-automated pavement condition surveys. An online questionnaire survey was conducted, along with scheduled virtual/phone interviews to gather information from government, industry, and academia about the state of the practice and state of the art. Open questions about the data quality and quality control & quality assurance (QC/QA) were used to receive first-hand inputs from highway agencies and pavement experts. The study has compiled the following observations: (1) Highway agencies urgently need a uniform data collection protocol for automated data collection; (2) the current QA requires too much human intervention; (3) cost ($100–$200 per mile) is a significant burden for state and local agencies; (4) the main issues regarding data quality are data inconsistencies and discrepancies; (5)  agencies expect a greater accuracy once the image processing algorithms are improved using artificial intelligence technologies; and (6) existing automated data collection methods are not available for project-level data collection.

Abstract

Cracks commonly develop in expressways owing to repeated vehicle loads and environmental impact. Therefore, timely and accurate detection of minor cracks can help prevent severe and expensive rehabilitation efforts as well as assist in inspections to avert significant transportation system failures. Consequently, crack detection and measurement in expressways has been extensively researched. Traditional two-dimensional approaches present challenges, as they typically require distance measurement data and large workload. A pan–tilt–zoom (PTZ) camera–based crack detection and measurement method based on deep learning to obtain the size measurement of detected cracks is proposed in this paper. A YOLOv5-based crack recognition model is first trained to detect and classify crack images captured by PTZ cameras. Our previously proposed crack size estimation method is then applied on these crack images to estimate length and width. The proposed approach successfully detects and measures cracks on highway pavements; the YOLOv5-based crack recognition results yield 87.7% mean average precision and more than 90% average precision for crack real size measurement on 12 cracks. Case studies conducted in the G4/Jingshi Highway demonstrate the applicability of the method and enable tuning of the measurement algorithm parameters, confirming the viability of our proposal.

Practical Applications

Cracks in highway pavement are among the most critical problems in expressway maintenance. Crack detection and measurement not only have severe consequences in relation to expressway safety and decide maintenance timing, but can also prevent the rapid development of crack disease. Owing to ITS development of an expressway network in China, PTZ camera–based crack detection and measurement using YOLOv5 networks is proposed. We train a crack recognition model using YOLOv5 to classify cracks before length and width estimation. Case studies conducted in the G4/Jingshi Highway in the Hebei province of China demonstrate that the proposed method was able to detect and quantify cracks in a cost-effective manner, providing objective and timely guidance for preventive maintenance. This method shows its promise in real-world civil engineering applications for network-level highway maintenance engineering.

Crumb-rubberized asphalt (CRA) binder has been widely used for several decades because it provides a nonhazardous disposal method for scrap tires and it improves some engineering properties of asphalt pavement. The energy consumption, atmospheric emissions, and storage stability of CRA binder are the issues limiting wider use of it. This study provides a novel method to introduce CRA binder into cutback cold-mix asphalt (CMA) remedying the drawbacks of CRA binder and achieving sustainable asphalt pavement in the entire construction process. The composition design of CMA mixture based on CRA binder was proposed based on the Brookfield viscosity, Marshall stability, and uniaxial compressive tests. The CMA mixture based on CRA binder further demonstrates excellent rutting resistance, crack resistance, and storage stability in laboratory experiments, despite being produced and stored at ambient temperature. An evaluation of sustainability shows that use of the CMA mixture based on CRA binder would provide a 58.94% decrease in energy consumption and a 81.42% decrease in greenhouse gas emissions, proving its great potential in sustainable construction of asphalt pavements.

The modulus back-calculation of asphalt pavement layers using falling weight deflectometer (FWD) data has become one of the most important methods of evaluating pavement bearing capacity. Several back-calculation methods have been proposed to estimate material properties. A new modulus back-calculation method called the deflection basin regulation algorithm (DBRA) has emerged recently. This algorithm requires an inertial point and two characteristic points in a deflection basin for back-calculating the moduli of three-layer pavement. However, the existing characteristic point positions are determined based on the theoretically calculated deflections and have not been verified by measured deflections. In this research, the field-measured deflection basins of different pavement structures are used to determine the optimal characteristic point positions to improve the new back-calculation method. First, the optimal point positions are determined based on the effects of the positions on the back-calculation variability. Then the back-calculated results based on the optimal characteristic points are compared to the results obtained using the MODULUS program. Finally, the improved modulus back-calculation method is also verified by the deflections from two pavement structures at different temperatures and loading levels. It was found that the optimal characteristic point positions were located at distances of 0 and 60 cm from the load center. Both comparative analysis and independent verification prove that the determined optimal characteristic point positions work well for back-calculating the layer moduli of other pavement structures. The findings of this research may facilitate the application of the DBRA for modulus back-calculation.

The influence of the transverse unevenness indexes of the road profile on the vehicle vibration was rarely studied in the past. This study proposes new road indexes of the transverse road profile and their correlation with the whole-body vibration (WBV) measured on the passenger’s and the driver’s seats in a passenger car. The proposed new indexes are the range, interquartile range, and standard deviation of the transverse profile vertical displacements. The benchmark was the maximum rut depth specified in the European standard. The WBV was evaluated on the passenger’s and driver’s seat surface in three perpendicular axes of various vehicle types traveling on test sections of different categories and quality over 4,100 km. The correlation of the proposed indexes with the total passenger’s seat acceleration was similar to that for the maximum rut depth. Using the combination of longitudinal and the proposed transverse unevenness indexes improved the correlation with the passenger’s seat vibration in the lateral direction by 23% and the model RMS error (RMSE) decreased by $\sim 15\%$ compared with that of the maximum rut depth.

Asphalt binder is one of the key constitutive components of hot-mix asphalt (HMA) that considerably affects its rutting performance. In particular, the high-temperature rheological properties measured from the multiple stress creep and recovery (MSCR) test are critical for correlating to the HMA rutting resistance. In this study, the Texas flexible pavements and overlays database was used as the data source to investigate the effect of asphalt-binder high-temperature rheological properties on the HMA rutting resistance. The study methodology was based on correlating the results of the MSCR test and the Hamburg wheel-tracking test (HWTT) to HMA field rutting performance. The data matrix for the study included asphalt binder (PG 64-22) from three different sources, three widely used Texas HMA mixes (fine gradation to coarse gradation), and five in-service highway test sections constructed using the same asphalt binders and HMA mixes. In general, the MSCR nonrecoverable creep compliance parameter, $J_{\mathrm{nrdiff}}$, showed fairly strong correlations with the HMA rutting performance in the laboratory and field. The percent recovery parameter ($R$), on the other hand, exhibited the potential to ascertain and quantify the presence of modifiers in the asphalt binders. Furthermore, the test results indicated that material source/supplier has an impact on the rheological properties of the asphalt binders with the same performance grade (PG). Overall, the use of the MSCR test to quantify the asphalt-binder high-temperature rheological properties indicated the potential to compliment the laboratory HWTT test for correlating with the field HMA rutting performance in terms of the effects of asphalt binder.

Detailed climatic data are essential for analyzing and designing pavements using AASHTOWare pavement mechanistic-empirical design (PMED). Hourly climatic data are used in the enhanced integrated climatic model in the PMED to adjust material properties. The current PMED V2.6.0 has been updated using the North American Regional Reanalysis (NARR) database for rigid pavements and the Modern Era Retrospective Analysis for Research and Applications (MERRA) climatic data for flexible pavements. Because several climatic data sources are available, it is helpful to establish the most viable data in the PMED. In a recent study, the Michigan DOT (MDOT) enhanced the quality of NARR data to improve data coverage and fill data gaps. This paper compares the climatic data for the enhanced NARR stations and closest MERRA-2 stations in Michigan to investigate the differences between the two data sources. The NARR and MERRA-2 climatic data have been validated with the two National Oceanic and Atmospheric Administration (NOAA) stations in Michigan to evaluate the accuracy of the data sources. The paper also evaluated the two data sources for quantifying the effects of climatic inputs on predicted pavement performance. The relative impacts of climatic inputs based on predicted performance have also been documented. The results show that MERRA and NARR climatic data are comparable for air temperature, humidity, and wind speed. Percent sunshine shows a low correlation, and precipitation data is substantially different. MERRA and NARR data quality is similar in Michigan with the NOAA data. In flexible and rigid pavements, performance predictions are substantially different for thermal and transverse cracking, respectively. These differences can be attributed to hourly variations in sunshine data. Temperature and percent sunshine data are the most critical climate inputs for flexible and rigid pavements. Considering the advantages of MERRA data, it is a suitable climatic data source for both flexible and rigid pavements.

Traffic noise causes discomfort to road users and residents near highways. Tire–pavement noise comprises major part of traffic noise caused by high-speed vehicles. Tire–pavement noise comes from tire vibration and air pumping in between a vehicle’s tires and the pavement’s surface texture. The intensity of the noise is affected by pavement texture characteristics such as texture depth and wavelength. To investigate the effects and mechanisms of tire–pavement noise, which depend on pavement textures, regression analysis between the noise and pavement texture in an asphalt-paved section was conducted. The result showed that the tire–pavement noise generation related to pavement textures can be explained by three principal mechanisms: (1) air pumping noise increases as wavelength increases due to larger intrusion of the tire into the surface texture, (2) tire vibration noise increases as texture depth increases due to larger volume of aggregate pushed into the tire, and (3) air pumping noise decreases as texture depth increases due to lower air pressure between the tire and the surface texture. Texture depth is highly correlated to texture wavelength because both parameters are dominated by the size of coarse aggregate. However, they influence tire–pavement noise in different ways. Thus, it can be concluded that overall tire–pavement noise cannot be interpreted and analyzed only through either texture depth or wavelength. The combined effects of texture depth and wavelength make it possible to better explain the mechanism and establish a comprehensive evaluation of tire–pavement noise.

In this study, an approach in determining the asphalt pay adjustment factor was proposed by evaluating the rutting performance based on different asphalt concrete acceptable quality characteristic (AQC) parameters. These AQC parameters are asphalt binder content, percent passing a 0.075-mm sieve ($P_{200}$), percent cumulative retained on a 4.75-mm sieve ($P_4$), field air voids (AV), and thickness. New specification limits for Seoul City, Korea, were initially established using a percent within limits (PWL) analysis, and these limits were then compared with the current limits in Korea and the US. A rutting model from a previous study was then used to evaluate the relative performance (RP) of the pavement, which was utilized in determining corresponding pay factors. Two different approaches were conducted wherein the first approach included the combined PWL and RP analysis and the second used the average RP analysis only. It was found that each approach has its own advantages and disadvantages when evaluating the pavement’s pay factor; therefore, the most appropriate approach in determining the pay adjustment factor was established.