Pavement Management Research Articles

Machine learning to enhance the management of highway pavements and bridges

The adoption of machine learning in transportation asset management is hindered by the perception of it being a black box, the natural resistance to change and the challenges of integration with existing management systems. This paper aims to enhance the understanding of machine learning and provide guidance for the development and implementation of machine learning to support decision making in the management of highway pavements and bridges. The paper identifies successful research efforts using machine learning, identifies opportunities and challenges in adopting machine learning and derives recommendations on when and how to apply different machine learning algorithms to support asset-management decisions. Four main challenges were identified: the trade-off between accuracy and interpretability, the shortage of machine learning engineers, data quality and the limitations of machine learning algorithms. Although the complexities associated with training machine learning algorithms challenge short-term implementation, machine learning offers a wide range of opportunities when compared with traditional approaches. The development of hybrid systems combining machine learning algorithms with expert opinions and traditional approaches seems a reasonable step forward to support agencies’ asset-management decisions.

Application of Lightweight Deflectometer to Assess Structural Competency of Nontraditional Airfield Pavements During Contingency Aircraft Operations

This paper discusses the use of the heavy and light weight deflectometer devices, hereafter referred to as HWD and LWD, respectively, to monitor the structural condition of airfield pavements during aircraft operations at contingency environments. Four airfield pavements like those encountered at contingency environments, referred to as nontraditional airfield pavements, were selected to conduct accelerated pavement traffic testing with simulated C-17 and C-130 aircraft. Pavement rut development was monitored, and both HWD and LWD tests were also conducted at selected traffic intervals. Both HWD and LWD were capable of tracking the pavement deterioration under accelerated traffic testing, which indicates that the LWD could be considered a potential rapid and practical pavement evaluation tool for nontraditional pavements during contingency aircraft operations. Based on current rutting failure threshold, a termination criterion in relation to a reduction in the stiffness parameter from the LWD test could be preliminarily used to aid in expediently determining acceptable aircraft operations on nontraditional airfield pavements. The influence of key operational parameters such as base plate dimension, the depth of influence from the impact force, and pavement rut formation must be further investigated to improve the effectiveness and applicability of the LWD test for nontraditional airfield pavements. The results from this study could be of interest to other low-volume road applications such as in agriculture, oilfield, large construction sites, and mining where pavement evaluation and management practices could be improved with the use of practical tools such as the LWD during the operations of heavy equipment or machinery.

Pavement Management in Highway Engineering

Pavement Management in Highway Engineering

Optimal pavement maintenance and management strategies for flood affected low volume rural roads

Due to the increase in urbanization and industrial development in India, there has been substantial traffic on the existing roads. This impact had increased public demand for good quality roads with better maintenance and high serviceability. Further, catastrophic events like flooding can cause damage to pavements, especially in those with granular base layers and thin asphalt mix surface layers. The degree of damage relies upon the vulnerability of the pavement to the entrance of surface water, the measure of water immersion, and the drainage quality of the pavement. The extensive literature review found limited studies on the post-flood maintenance strategy for low-volume roads by considering flood resilience. In the present scenario of changing climatic conditions, it is necessary to conduct a study on the post-flood maintenance strategy of low-volume rural roads in India. This study focuses on an approach to developing post-flood maintenance and rehabilitation strategies for flood-affected low-volume roads. Ten road stretches adversely affected by the flood were selected to fulfil the aim of the study. The post-flood maintenance strategy was determined based on the Pavement Condition Index (PCI) and deflection-based pavement performance with flood resiliency of pavement materials and budget constraints. The extensive data analysis revealed that the pavement with a cement-treated base or subbase has a longer service life and less Net Present Value (NPV) than all other alternatives. NPV is 30 to 45 for rehabilitation and 52 to 55 for reconstruction, lower than other alternatives considered in the study.

Sustainable Infrastructure Is a Two-Way Street: Balancing Environmental and Condition Performance Goals

Transport agencies are under increasing pressure to mitigate the global warming impact of our pavement systems. This objective, however, must be carefully balanced with other performance metrics of interest (e.g., pavement condition) for federal and state agencies. This paper details a network-level tool aimed at supporting transport agencies in achieving two competing objectives: (1) maximizing the number of pavement segments in a good state-of-repair; and (2) minimizing the global warming impact of the network. The stochastic optimization model follows a two-stage bottom-up approach, where optimal policies are learned for individual facilities, and those decision-rules are subsequently used to guide the allocation of resources across the network. The model is applied to a realistic roadway network composed of 159 miles of pavement segments based on data made available via the Highway Performance Monitoring System. The case study results highlight that, over a 20-year analysis period, maximizing the condition of pavement assets across the network increases its expected global warming impact by 1% to 8%. The results also highlight that, invariant to the selected objective and/or available budget, increasing the allocation of funds toward rehabilitation activities rather than reconstruction treatments improves the overall condition of the network by as much as 25% and reduces its global warming impact by up to 7%. The results of the case study provide decision-makers with important insights around the impact of pavement management performance goals and budgetary policies on the global warming impact of pavement systems.

Empirical Models Investigation of Pavement Management for Advancing the Road's Planning Using Predictive Maintenance

Empirical Models Investigation of Pavement Management for Advancing the Road's Planning Using Predictive Maintenance

Effect of Data Imbalance on the Performance of Pavement Deterioration Models

Pavement performance models that can model consistently well across different pavement types, structures, traffic, and environmental conditions as well as throughout pavements’ lives are essential for the efficient management of transportation infrastructures. However, network pavement performance data are inherently imbalanced, and most pavement performance models are built on algorithms that assume balanced data distribution and equal error cost. As a result, the performance of most pavement performance models may vary for different pavement groups. Considering this, the objective of this study was to evaluate the data imbalance of the network pavement performance data and its effect on the performance of pavement deterioration models. Specifically, the analysis result for the longitudinal cracking model is presented in this paper. The data imbalance of the network pavement performance data mainly arises from two sources: the intrinsic pavement characteristics and pavement life expectancies. The intrinsic data imbalance arises either from the design or the management process, whereas the data imbalance introduced by the pavement life expectancies is primarily the result of the pavement management process. The distribution of the data used in this study exhibits substantial skewness owing to the data imbalance caused by the intrinsic pavement characteristics and life expectancies. The existence of the data imbalance leads to the biased performance of the deterioration models toward pavement classes with more data.

A Top-Down Approach Based on the Circularity Potential to Increase the Use of Reclaimed Asphalt

Resource depletion and climate change, amongst others, are increasingly worrying environmental challenges for which the road engineering sector is a major contributor. Globally, viable solutions that comply with the principles of circular economy (CE) are being investigated that can replace conventional asphalt mixtures in a post-fossil fuel society. The use of reclaimed asphalt (RA) is a widely used and well-established method to reduce the environmental and economic impacts of asphalt mixtures while increasing their circularity. However, RA’s market supply and demand have not yet been systematically analyzed and established. Moreover, the actual circularity potential and the opportunity of re-circulating RA in a closed-loop model have not yet been methodically defined. To address this, a three-layered framework to quantify and assess the circularity potential (Ω) of RA has been developed. To give stakeholders and legislative bodies a simple method to assess the opportunities available to them to become “more circular”, a novel equation has been formulated. This takes the form of a three-level indicator that considers: technical aspects, the effect of the RA market, and the legislative restrictions. A case study in Germany was structured and undertaken to develop and verify the proposed approach. The results indicate that the available RA is insufficient to cover the needs of asphalt mixture production; even though RA production is significantly lower than the actual need of asphalt mixtures, it is not utilized in its entirety. An impactful step forward is the alteration of the regulations to support the higher utilization of RA in asphalt mixtures, and subsequently, the increased circular opportunity and potential of RA. Thus, Circularity potential (Ω) is a composite indicator that can support stakeholders, designers, and asset managers during the process of decision-making, to follow more circular operational, design, and asphalt pavement management patterns.

Development and Temperature Correction of Piezoelectric Ceramic Sensor for Traffic Weighing-In-Motion.

Weighing-In-Motion (WIM) technology is one of the main tools for pavement management. It can accurately describe the traffic situation on the road and minimize overload problems. WIM sensors are the core elements of the WIM system. The excellent basic performance of WIMs sensor and its ability to maintain a stable output under different temperature environments are critical to the entire process of WIM. In this study, a WIM sensor was developed, which adopted a PZT-5H piezoelectric ceramic and integrated a temperature probe into the sensor. The designed WIM sensor has the advantages of having a small size, simple structure, high sensitivity, and low cost. A sine loading test was designed to test the basic performance of the piezoelectric sensor by using amplitude scanning and frequency scanning. The test results indicated that the piezoelectric sensor exhibits a clear linear relationship between input load and output voltage under constant environmental temperature. The linear correlation coefficient R2 of the fitting line is up to 0.999, and the sensitivity is 4.04858 mV/N at a loading frequency of 2 Hz at room temperature. The sensor has good frequency-independent characteristics. However, the temperature has a significant impact on it. Therefore, the output performance of the piezoelectric ceramic sensor is stabilized under different temperature conditions by using a multivariate nonlinear fitting algorithm for temperature compensation. The fitting result R2 is 0.9686, the root mean square error (RMSE) is 0.2497, and temperature correction was achieved. This study has significant implications for the application of piezoelectric ceramic sensors in road WIM systems.

Continuous remote monitoring of a motorway section using geophones

Pavement monitoring has been a widespread area of research. Monitoring pavements using embeddable sensors and a data acquisition system is an active research topic, with important questions about the selection of appropriate sensors and data acquisition systems, and interpretation of data. This work presents results of the instrumentation of a French motorway section, using geophones and temperature probes. The sensors are connected to a remote monitoring system which is compact, has a low power consumption, and real-time data processing capabilities. This system performs continuous monitoring, under a heavy motorway traffic and transmits the data wirelessly to a remote server. This work focuses on the treatment and analysis of the sensor signals. It indicates a good reliability of the system and embedded sensors, and presents different methods of interpretation of the data, to monitor pavement deflections, but also traffic composition and traffic loads, which could be used for pavement management, for detection of pavement damage and preventive maintenance.

Tire-pavement friction modeling considering pavement texture and water film

The accurate estimation of tire-pavement friction, especially under wet conditions, is critical to ensure pavement safety. For this purpose, this paper develops a modified tire-pavement friction model which takes the effect of pavement texture and water film into consideration. The influence of pavement texture is quantified by a newly proposed parameter called texture influence coefficient, which is related to the real contact patch of tire-pavement. The water effect is calculated from two parts, namely lubrication effect and hydrodynamic effect. Based on these two steps, a modified average lumped LuGre (ALL) model is developed. The proposed model is calibrated and verified by GripTester data collected under different vehicle velocities and water film thicknesses. The root mean square error between the calculated value of the model and the measured value is 0.023. In addition, the effects of vehicle velocity, slip rate, water film thickness, and pavement type on the friction coefficient are analyzed by numerical calculation. The results show that the friction coefficient reaches the maximum when the slip rate is in the range of [0.15, 0.20]. The increases in the vehicle speed and water film thickness will lead to the decrease in the friction coefficient. Besides, in thin water film (<1 millimeter) conditions, the deterioration effect of water film thickness on the friction coefficient is more remarkable. The results prove that the modified tire-pavement friction model provides a precise and reliable way to estimate the friction coefficient of pavement, which can assist the pavement management systems in risk warning and safety guarantee.

Machine-Learning-Based Framework for Prediction of the Long-Term Field Performance of Asphalt Concrete Overlays in a Hot and Humid Climate

Pavement performance prediction models are used by state agencies to determine pavement maintenance and rehabilitation strategies. However, most performance prediction models are based on a limited number of parameters and a maximum prediction period of five years. With the ever-increasing amount of available pavement performance data, machine-learning techniques have become a promising alternative to traditional performance prediction models. The objective of this study was to develop a machine-learning-based framework for states with a hot and humid climate that can predict the long-term field performance (up to 11 years) of asphalt concrete (AC) overlays on asphalt pavements based on key project conditions. The pavement condition index (PCI) was used as the pavement performance indicator. Two machine-learning algorithms, namely, random forest (RF) and CatBoost, were examined. A total of 892 log-miles of AC overlay data were obtained from the Louisiana Department of Transportation and Development Pavement Management System database. Based on the collected data, six models were trained (for each algorithm) and validated to predict the PCI of AC overlays for up to 11 years. The results indicated that the RF algorithm yielded higher accuracy than the CatBoost algorithm. Therefore, the RF-based models were considered in the proposed decision-making framework.

Field Validation of Falling Weight Deflectometer Deflection-Based Critical Strain Prediction Models for Full-Depth Asphalt Pavements

Critical strains are one of the most reliable mechanistic parameters representing the structural capacity of flexible pavements. However, critical strains are currently not considered in pavement management systems (PMS) because of a lack of methodology for measuring critical strains in in-service pavements. Recently, critical strain prediction models have been developed by the Indiana Department of Transportation and Purdue University research team to estimate the critical strains of existing full-depth asphalt pavements using falling weight deflectometer (FWD) data. Even though the accuracy of these models was verified with finite element (FE) analyses, additional field validation is needed before incorporating the model results into PMS. This study focuses on validation of these previously developed critical strain prediction models using field-measured strains. Nine field sections, representing typical full-depth asphalt pavements in Indiana, were selected for study inclusion. FWD testing was conducted on the field sections and field cores were collected. Field strains were measured by a strain gauge or were determined by FE analysis using the material properties obtained from the field cores. Based on the comparative analysis, the critical strain prediction models were successfully validated with strains measured from the field. The coefficients of determination were up to 0.98 for the transverse tensile strain at the bottom of the asphalt layer and 0.93 for the vertical compressive strain at the top of the subgrade. Therefore, it is concluded that the critical strain prediction models can be used to estimate critical strains of in-service full-depth asphalt pavements for structural evaluation of PMS.

Comparative Study of Different Condition Indices Using Michigan Department of Transportation’s Flexible Distress Data

The condition of a country’s roadway infrastructure significantly affects its economic development. Highway agencies that manage pavements often communicate their road network condition by using single or multiple composite condition indices. Each agency relies on its condition index to make maintenance decisions and perform budget allocation. Generally, these condition indices follow a similar level of the rating scale (100 to 0, where 100 denotes perfect condition). However, because of different numbers of distress inputs with different units, different severity levels, and different deduct points and mathematical formulas, condition indices are often incomparable. Consequently, research focused on comparing different distress indices using a highway agency’s pavement distress data is lacking. The objective of this study is to provide a comparative analysis of different condition indices using Michigan Department of Transportation’s (MDOT) flexible distress data, as part of an effort to develop a new index for Michigan. In this process, 2,081 flexible pavement sections were identified from MDOT’s Pavement Management System database, and five different condition indices were evaluated and compared against MDOT’s Distress Index. Comparison results were presented through scatter plots and statistical analysis results. MDOT’s maintenance records were also used to compare the ranges of magnitudes of various indices computed at the time when various types of maintenance activities took place. Finally, this study resulted in recommendations for selecting a new condition index and calibration needs of deduct points and equations.

Pemodelan Kondisi Perkerasan Jalan dengan Metode Probabilistik Rantai Markov

Road pavements are an important part of transportation infrastructure and require regular maintenance. This study aimed to model the condition of road pavement by the probabilistic method of Markov chains. The method used in this study is a literature study and data collection from a survey of road pavement conditions in the field. Data analysis was performed using the Markov chain probabilistic method, which can be used to model changes in pavement conditions over time. The results showed that the Markov chain probabilistic method can be used to model changes in pavement conditions with high accuracy. This method can also be used to make predictions of future pavement conditions and determine the right time to perform maintenance and repairs. The study recommends using the Markov chain probabilistic method as a tool to model pavement conditions and make appropriate maintenance and repair decisions. Thus, it is expected to increase the effectiveness of road pavement management and extend the service life of road pavement.

Evaluation of Pavement Performance Models Using Historical Condition Data for the U.S. National Parks Pavement Network

Pavement management, a subdivision of transportation asset management, seeks to use finite and limited resources to maintain or improve the condition of a pavement network for a defined time horizon. Pavement performance prediction is integral to pavement management, and among other uses, it contributes to budget projections and allows scenario analyses to be conducted. This paper documents the evaluation of the original pavement performance models developed for the U.S. National Park Service (NPS) pavement network using historical condition data. The Federal Lands Highway Program began to use a pavement management system for NPS pavements in 2005, and the results in this paper represent an evaluation of those models with an emphasis on developing recommendations for improvement. The main finding of the work presented in this paper is that the current NPS models result in significant overprediction of deterioration in practically all cases, although the effect is more pronounced for those measures associated with structural failure. This means that the NPS pavement network is in better condition than the models predict. The performance prediction models for those segments that received preservation treatments are also investigated in this paper, and it is found that the model bias is less for these than for the models predicting condition without a preservation treatment. However, this is attributed to the segments that received preservation treatments being in good condition. Overall, the findings point to a need to recalibrate or redevelop the models to improve their prediction capabilities.

Quantifying Uncertainty with Pavement Performance Models: Comparing Bayesian and Non-Parametric Methods

An important part of pavement management systems is accurately estimating the performance-time-degradation relationship. One common approach to establishing this relationship is to use performance family curves. These curves are developed by collecting performance data at specific points in time and collectively shifting pavements of various ages to identify the probable underlying function. This paper compares two alternative methods for characterizing such a family curve function. First, a Bayesian method (Method-A) is used, which fits both the family curve and the shift factor function in parallel by assuming a Beta distribution for pavement performance condition rating (PCR). Second, a non-parametric method (Method-B) is developed, which fits the model in two steps; (1) by fitting the family; and (2) by horizontal shift to minimize the error. PCR values from flexible pavements in North Carolina (NC-PCR) are used for this comparison. These data include a total of 30,988 pavement sections segregated according to surface type and traffic level. Data from 2013 to 2015 are used for model calibration, and data from 2016 are used for model validation. The root means square error and k-fold cross-validation test are used to conduct the comparison, and Method-A is found to be preferred. The uncertainty in both models is quantified and compared. On the basis of this uncertainty, the Bayesian method is preferred, but in cases with large data sets, a non-parametric method does result in lower uncertainty.

Mechanistic Assessment of Flexible Pavements in Wisconsin Using Traffic Speed Deflectometer

The traffic speed deflectometer (TSD) has been gaining interest for pavement management practices, as it can continuously collect deflection data at traffic speed without requiring traffic control. TSD data have been typically used to provide an estimate of the structural condition of pavements with a limited range of significant parameters considered in the analysis. To date, limited studies have been conducted in or for cold regions to consider the potential use of TSD data for pavement structural assessment. This study aimed to bridge this gap by developing a twofold approach to: (1) mechanistically characterize the structural integrity and remaining life of pavements; and (2) compare TSD and falling weight deflectometer (FWD) data collected in the field, considering the features of pavement facilities in Wisconsin. A field testing plan was developed to characterize the site-specific features of eight representative pavement sections, with a total length of about 30 mi and distinct pavement structures, truck traffic patterns, and climatic conditions. The postprocessed results underscored the significance of the TSD data and their potential for the structural evaluation of underlying layers in a multilayered pavement system when integrated with the site-specific pavement profile and truck traffic information derived from field testing. A comparison of the normalized deflections and structural parameters characterized by TSD and FWD data showed similar trends, while the average FWD-based deflections were 12%–37% higher than the TSD-based deflections.

Evaluation of Airport Pavement Base Layer Stiffness Characteristics via Embedded Field Sensors

Real-time stiffness monitoring of unbound aggregate layers using embedded field sensors is crucial for advanced airport pavement design and management. This paper describes research findings on monitoring base layer stiffness characteristics for airport pavements using an embedded bender element (BE) field sensor, inductive coil sensors, and pressure cells. A full-scale pavement test section was constructed at the Federal Aviation Administration (FAA)’s National Airport Pavement Test Facility (NAPTF). A BE field sensor, inductive coil sensor pairs, and pressure cells were installed in the unbound aggregate base to evaluate the layer stiffness, applied load stress, and deformation characteristics. A triple dual tandem gear loading module applied 58,000 lb per wheel to the tested section with a wander pattern consisting of 66 passes arranged in nine lateral wander positions. Dynamic responses of coil sensors and pressure cells were collected during the traffic test, whereas BE signals were collected after completing each wander pattern. Laboratory repeated load triaxial tests with BE instrumentation were also conducted to establish a correlation for converting BE field sensor readings to layer modulus properties. The in situ modulus characteristics of the aggregate base layer were evaluated during full-scale testing using two methods: shear wave velocities from a BE field sensor, and strain and stress measurements from coil sensors and pressure cells. The moduli estimated using both approaches fell into typical ranges of crushed aggregate bases and were highly comparable. The moduli from the dynamic sensor measurements clearly showed the effect of vehicle wander on aggregate base layer measured responses.

Development of Pavement Performance and Remaining Service Life Prediction Tools for Iowa Jointed Plain Concrete Pavement Systems

Moving Ahead for Progress in the Twenty-First Century (MAP-21) requires US state highway agencies (SHA) to utilize performance-based approaches in their pavement management decision-making processes, and use of a remaining service life (RSL) model would be one of such performance-based approaches for facilitating the pavement management decision-making process for SHAs. In this study, statistical and artificial neural network (ANN)-based pavement performance and RSL models were developed for Iowa jointed plain concrete pavement systems (JPCP) using actual pavement structural, traffic, construction history, and pavement performance records obtained from the Iowa Department of Transportation pavement-management information system database. While both models were found to be potentially useful for project and network level performance and RSL predictions, statistical and ANN-based models were respectively found to be more suitable for project and network level analysis. Using these models, efficient Microsoft Excel-based automation tools were created to predict future performance of a JPCP section and estimate RSL values based on predicted future performance and threshold limits for the performance indicators. Consequence analysis was also conducted to investigate the impact of traffic and preservation treatment (diamond grinding) on the RSL of a JPCP. The tool, also capable of estimating realistic pavement pretreatment and posttreatment performance and RSL, could be successfully used as part of performance-based pavement management strategies and helping decision-makers make better-informed pavement management decisions to properly allocate agency resource expenditures. Moreover, this study provides a better understanding of RSL and the factors that influence both the project and network level RSL.