Jidong Yang

Modeling Crack Deterioration of Flexible Pavements: Comparison of Recurrent Markov Chains and Artificial Neural Networks

Pavement cracking and rutting are two of the most critical distress types manifested on flexible pavements, and they often govern the overall pavement condition. Hence, many models have been developed for forecasting the deterioration of the crack condition accurately, with the traditionally preferred technique being the use of a regression relationship developed from laboratory or field statistical data, or both. However, it becomes tedious for regression techniques to predict crack performance accurately and robustly in the presence of the multitude of tributary factors, material nonlinearity, and uncertainty involved in the cracking process. With the advancement of modeling techniques, two innovative breeds of models, neural networks and recurrent Markov chains, have drawn increasing attention from researchers for their use in modeling complex phenomena such as pavement cracking. This paper compares the ability of neural networks and recurrent Markov chains to model crack performance, using the Florida...

Forecasting Overall Pavement Condition with Neural Networks: Application on Florida Highway Network

Timely identification of undesirable crack, ride, and rut conditions is a critical issue in pavement management systems at the network level. The overall pavement surface condition is determined by these individual pavement surface conditions. A research project was carried out to implement an overall methodology for pavement condition prediction that uses artificial neural networks (ANNs). In the research, three ANN models were developed to predict the three key indices—crack rating, ride rating, and rut rating—used by the Florida Department of Transportation (FDOT) for pavement evaluation. The ANN models for each index were trained and tested by using the FDOT pavement condition database. In addition to the three key indices, FDOT uses a composite index called pavement condition rating (PCR), which is the minimum of the three key indices, to summarize overall pavement surface condition for pavement management. PCR is forecast with a combination of the three ANN models. Results of the research suggest that the ANN models are more accurate than the traditional regression models. These ANN models can be expected to have a significant effect on FDOTs pavement management system.

Factors Affecting Students’ Walking/Biking Rates: Initial Findings from a Safe Route to School Survey in Florida

A Safe Route To School (SRTS) survey on students’ travel modes for parents and students was conducted at 18 schools (16 elementary and 2 middle schools) in Pinellas County, Florida. A new diagnosis approach was used to pin down the factors significantly affecting students’ travel modes, especially safety and security factors. The analysis was conducted in multiple perspectives, including (1) overview perspective that gives the general statistical information of each potential factor, (2) forward direction perspective that explores the cause–effect (how walking/biking rates change as different factors levels change), and (3) backward direction perspective used to identify the similar properties of the student group with the same travel mode (walking/biking). Multiperspective diagnosis analysis identified different factors that significantly affect the walking/biking rate for different groups of students. Generally, students living in different distance intervals are subject to different barriers. Security and safety remain the primary factors of concern for parents to allow their children to walk or bike to school, especially for those living at short walkable distances. The other significant subjective variables include grade levels, school attitudes, enjoyment, healthy, allowable grade level, and students attitude.

Integrating Left-Turn Lane Geometric Design with Signal Timing

The available design guidelines indicate a lack of proper methodologies for determining the length of left-turn lanes at signalized intersections. This is largely attributable to the alternating nature of stopped and unstopped approaches as a result of cyclic signal phases. The existing design guidelines recommend the design length of left-turn lanes as a sum of two components: queue storage length and deceleration length. This is appropriate for unstopped approaches at unsignalized intersections, but it is inapplicable to signalized intersections, in which the approaches switch between stopped and unstopped by phase. This paper introduces a new methodology that coordinates the requirement of each component with signal timing for the proper design of left-turn lanes. With the underlying assumption of Poisson arrivals, the model developed is limited to uncoordinated intersections or approaches. Compared to the existing guidelines, the proposed methodology provides a rational approach for determining left-turn lane lengths and generally results in shorter left-turn lanes that necessarily meet both safety and operational needs.

Investigating the performance of as-built and overlaid pavements: a competing risks approach

Overlay has been a commonly used rehabilitation measure for flexible pavements. Properly designed and constructed overlays restore pavement service conditions and economically extended pavement service lives. In the light of different failure modes that could be experienced by a flexible pavement, such as cracking, rutting and roughness, and complex interactions among them, this paper considers these failure modes as competing risks and evaluates performance differences between overlaid pavements and as-built pavements in terms of hazard function, cause-specific cumulative incidence function and cause-specific conditional cumulative incidence function. A case study was undertaken using the pavement condition survey data in Florida. It was found that the overlaid pavements generally have a higher failure risk compared with the originally built pavements across all failure modes and exhibit three distinct phases of failure hazard change: increasing, stabilising and decreasing. For as-built pavements, the failure hazard increases dramatically after exceeding the design lives of pavements.

Nested Markov Decision Framework for Coordinating Pavement Improvement with Capacity Expansion

Pavement improvement and capacity expansion traditionally fall in two different decision-making processes. Pavement improvement decisions are typically made at the maintenance level and focus on maintaining, rehabilitating, and reconstructing the existing pavements with respect to the physical conditions, such as poor riding quality, severe cracking, or rutting. In contrast, capacity expansion decisions are normally made at the planning level in regard to the operational conditions, such as levels of service, travel speeds, or delays. The recently adopted asset management approach calls for integrated decision-making that balances both types of decisions. In this context, this paper introduces a nested Markov decision process (NMDP) framework that can be used to obtain the optimal policy for joint pavement improvement and capacity expansion decisions. The applicability of the proposed NMDP framework is demonstrated through a numerical example showing how a special capacity expansion decision, road widening, can be integrated with conventional pavement improvement decisions for upgrading roadway facilities.

Application of Improved Crack Prediction Methodology in Florida's Highway Network

With the growing need to maintain roadway systems despite increasing competition for resources while ensuring safety and comfort for travelers, sound network-level decision making becomes more vital than ever. A stochastic process known as the Markov chain has been used extensively to capture the uncertainty associated with pavement performance over time and to support this critical decision-making process. By application of the Markov chain, this paper investigates the crack histories of flexible pavements to gain insight into the impacts of two primary factors that contribute to the rapid deterioration of surface cracks in flexible pavements: excessive traffic loading and delayed maintenance and rehabilitation. The empirical results of the investigation, obtained by using the data from the Florida Department of Transportations pavement condition survey database, are presented. The results show that the impacts of the two factors mentioned above are statistically different from one another in terms of the rate of deterioration of Floridas pavements because of cracks. These findings will assist highway authorities in making more timely and efficient network-level decisions.

Estimate of Resilient Modulus of Graded Aggregate Base in Flexible Pavement

The development of model to estimate the Graded Aggregate Base (GAB) resilient modulus is described in this paper. Eleven (11) different sources of GAB available locally in Georgia were subjected to the resilient modulus test in accordance with AASHTO T 307-99 with two replicates. The stress state and physical properties of resilient behavior of GAB were then successfully correlated using an Artificial Neural Network (ANN) model. The study indicates that the stress state and physical properties of GAB affect the resilient behavior of GAB, which in turn has a substantial effect on the pavement response and design. In this study, the ANN was trained using Bayesian regulation back-propagation algorithm to overcome problems associated with over-fitting. The model results indicated that the ANN reasonably estimated the GAB resilient modulus based on the stress state and physical properties of the aggregate.

Factorial effects of mix design variables on the coefficient of thermal expansion of concrete mixtures

The coefficient of thermal expansion (CTE) is one of the most critical parameters for concrete pavement design. Concrete paving mixtures with high CTEs are generally subjected to temperature-related stresses and deterioration. In this study, a 2k design of experiment was applied to evaluate the effects of concrete mix design variables on the CTE of resultant concrete mixtures. Laboratory experiments were conducted by controlling six mix design variables, each at two levels, including aggregate type (granite or dolomite), fine sand type (manufactured or natural), fly ash type (Class C or Class F), cement content (314 or 273 kg/m 3), coarse aggregate content (1246 or 682 kg/m 3), and air content (3% or 6%). The respective and collective effects of these factors on the CTE were analysed. It showed that the type of coarse aggregate, fine sand, fly ash, and their respective contents are significant in explaining the variance of the lab-measured CTEs. The minimum CTE is a result of the concrete mixture consisting of low content of granite and cement in combination with high content of manufactured sand and Class F fly ash. To generalise the results, a parsimonious model was formulated and is capable of explaining over 92% of the total variance of the CTE and predicting it in a reasonably accurate manner.

Forecasting pavement remaining service life with limited causal data

Assessment of pavement remaining service life assists decision making on pavement maintenance and rehabilitation (M&R) such that proper M&R actions can be selected and scheduled to optimize the use of resources over the life cycles of pavements. In this paper, a pavement remaining service life model was developed dealing with the limited causal data present in the pavement management system databases. The model achieves this by including the current pavement condition rating in the model specification and considering the boundary conditions of the pavement deterioration process. Empirical results of model estimation and verification are presented in the context of the Florida pavement condition data sets.