Artificial Neural Networks for Mechanistic-Empirical Concrete Pavement Design: An Efficient Development Approach

The existence of multi-dimensional tradeoffs among performance metrics is an impediment to the development of reliable Electronic Design Automation (EDA) tools for the design of analog circuits. To overcome this limitation, soft computing techniques that have functionality similar to the human brain are being explored. This work proposes a methodology for the development of a fast and accurate Artificial Neural Network (ANN) model for analog circuits, to achieve optimization of its design, considering a 2-stage op-amp for illustration. In order to map complex and multiple input-output relationships of analog circuits, an ANN architecture of a single hidden layer with its neuron count selected to be between the number of neurons in the input and output layers is proposed. The ANN model is developed in the MATLAB environment, and trained using SPICE-generated performance dataset. For a given set of performance metrics, the ANN model generated designs are validated against SPICE simulations. It is demonstrated that trained ANN models substitute the circuit simulator, to achieve design optimization, in a single iteration, for a new set of design specifications.

As the leading global producer of palm oil, Indonesia encounters substantial environmental challenges arising from the waste generated by empty palm oil fruit bunches (EPOFB). This research aims to develop an accurate Artificial Neural Network (ANN) model to predict the tensile strength of EPOFB fiber-reinforced composites. The method involves two types of ANN, namely Radial Basis Function (RBF) and Backpropagation, with testing using variations in immersion time, volume fraction, and length of EPOFB fibers. The research results show that both ANN models can predict tensile strength with a Mean Absolute Error (MAE) below 10%. However, the Backpropagation ANN shows superior performance with a training MAE of 0.0078 and a testing MAE of 0.45, compared to the RBF ANN, which has a training MAE of 0.371 and a testing MAE of 0.53. In conclusion, ANN Backpropagation is superior in prediction accuracy and characterization efficiency of EFB fiber-reinforced composites, offering an economical solution and supporting sustainable palm oil waste management.

Mechanistic–empirical pavement design has received significant attention from the pavement community as the method for designing asphalt pavements in the future. Currently available software for mechanistic–empirical pavement design includes the AASHTOWare Pavement ME Design (Pavement ME) program. The Pavement ME program allows users to predict pavement distresses by applying layered elastic theory for the mechanical responses and using empirical models for the distress predictions. The layered viscoelastic pavement design for critical distresses (LVECD) program, which employs three-dimensional viscoelastic finite element analysis with moving loads, can also be used to predict the fatigue and rutting performance of pavements. The LVECD program employs the simplified viscoelastic continuum damage (S-VECD) model as the material model for the fatigue performance predictions of asphalt mixtures under complex loading and environmental conditions. This paper examines and compares the performance of 33 pavement sections from five research projects located in the United States, Canada, and South Korea by using both the Pavement ME and LVECD computer programs. To verify the results obtained from these two programs, the simulations were compared with the field performance data. In terms of ranking, the LVECD simulations provided better agreement with the field performance data than did the Pavement ME simulations. One of the main reasons for the better predictions obtained by the LVECD program is that its fatigue performance predictions depend on the mixture properties of all the layers, whereas the Pavement ME program considers the fatigue properties of only the bottom layer mixture.

Medical prognosis has become an emerging area in health care. Several reliable prognostic models based on survival analysis procedures have been useful to a variety of domains, with different degree of success. An enhanced model with further advancement of computerized medical decision support systems help clinicians and doctors in building strategies for understanding surgical outcomes. A tenfold cross validation (CV) was applied in the medical input dataset which was obtained from United Network Organ Sharing database. In order to perform the dimensionality reduction of a huge database, principal component analysis (PCA) with ranking was done. The relation between attributes was recognized and proved using various association rule mining techniques, such as apriori, tertius, and treap algorithms. For confirming the results, we compared the rules generated by the association rules mining algorithm before and after PCA was also performed. We proposed an effective and accurate artificial neural network (ANN) model for the prediction of long-term survival of liver patients who undergo liver transplantation (LT). A tenfold CV was applied in the medical input dataset which was obtained from United Network Organ Sharing database. We made survival analysis of 13 years in the prediction of liver patients after LT. We trained the liver follow up data of 13 years separately using multilayer perceptron ANN model with proper selection of data attributes. Our model outperformed the existing models in terms of specificity, sensitivity, and accuracy. From the comparison of existing approaches, our model promised high accuracy in survival analysis after LT. We have also evaluated the survival probabilities of 13 years of follow up liver data and proved that our prediction model is suitable for long-term prognosis of survival of patients after LT.

Since the release of the first research version of Mechanistic–Empirical Pavement Design Guide software, the State of Louisiana has conducted a series of studies in preparation for adopting the new AASHTO design guide. The objective of this study was to develop an implementation guideline for the full implementation of the AASHTOWare Pavement ME Design program (Pavement ME) in Louisiana. For this purpose, a Louisiana default input strategy that reflected state and local conditions and practices was first developed. Design reliability and performance criteria were established on the basis of a consensus survey and pavement management system (PMS) distress triggers. A total of 162 projects were selected from the existing Louisiana highway network, including flexible pavements with five types of base, rigid pavements with three types of base, and rehabilitation with asphalt overlays. Performance data were retrieved from the PMS database and validated through additional field performance surveys. Through comparison of the performance predicted by the Pavement ME with the measured performance recorded in the PMS database, distress models were evaluated and calibrated according to local Louisiana conditions. Subsequently, a Louisiana Pavement ME Design Guideline that contains all necessary design input information and calibration coefficients was developed. Finally, the practical and economic benefits of the full implementation of the Pavement ME for Louisiana pavement design were demonstrated by a direct comparison with the current 1993 AASHTO design method through a group of thickness design examples and life-cycle cost analyses.

Axle load distribution factors (ALDFs) are used as one of the primary traffic data inputs for mechanistic-empirical (ME) pavement design methods for predicting the impact of varying traffic loads on pavement performance with a higher degree of accuracy than empirical methods that are solely based on equivalent single axle load (ESAL) concept. Ideally, to ensure optimal pavement structural design, site-specific traffic load spectra data—generated from weigh-in-motion (WIM) systems—should be used during the pavement design process. However, because of the limited number of available permanent WIM stations (in Texas, for example), it is not feasible to generate a statewide ALDFs database for each highway or project from permanent WIM data. In this study, two possible alternative methods, namely, the direct measurement using a portable WIM system and the cluster analysis technique, were explored for generating site-specific ME-compatible traffic data for a highway test section, namely, state highway (SH) 7 in Bryan District (Texas). The traffic data were then used for estimating pavement performance using a ME pavement design software, namely, the Texas Mechanistic-Empirical Thickness Design System (TxME). The TxME-predicted pavement performance (e.g., rutting) using the portable WIM-generated traffic input parameters closely matched with the actual field performance. Overall, the study findings indicated that the portable WIM (with proper installation and calibration) constitutes an effective means for rapidly collecting reliable site-specific ME-compatible traffic data.

The reflection of existing cracks from the underlying pavement is one of the primary modes of failure for rehabilitated asphalt concrete (AC) pavements with AC overlays. The AC overlay design philosophy is to delay the onset of reflection cracking or to eliminate it entirely. In recent years, rehabilitation designs are being optimized not just for cost effectiveness, but also to produce more reliable, sustainable, and environmentally friendly designs that require more detailed and sophisticated methodologies. The new finite element and fracture mechanics based reflection cracking models were developed under National Cooperative Highway Research Program (NCHRP) Project 1-41 to improve design methodology. The new models were generally compatible with the AASHTOWare Pavement ME Design (AASHTOWare) framework. The difference included the computation of critical pavement responses (i.e., stress intensity factors) and others, and thus modified. The modified models were calibrated and validated using Long-Term Pavement Performance projects. The calibration of reflection cracking models produced satisfactory goodness of fit with no significant bias. Thus, the modified reflection cracking models were successfully adapted and deployed in the AASHTOWare software to create a new improved AC overlay of existing AC pavement design methodology. This new methodology represents a paradigm shift from the traditional pavement design procedures. It significantly increases the capability of the pavement design software to simulate all the loadings experienced by AC overlay, estimate pavement responses, and predict reflection cracking based on pavement responses. This paper discusses the implementation of NCHRP 1-41 reflection cracking models for AC overlay of existing AC pavements into AASHTOWare software.

This report focuses on the calibration of a design framework for continuously reinforced concrete pavement (CRCP) developed in 2009 by using the most recent performance data acquired from existing CRCP sections in Illinois. Field performance data were used to update the fatigue damage to punchout model coefficients in the design framework. A sensitivity analysis was performed on the updated CRCP design program to determine its sensitivity to traffic levels, shoulder type, and support conditions as well as its magnitude relative to jointed plain concrete pavement design curves. Additionally, the new CRCP design charts were compared to AASHTOWare Pavement ME Design predicted CRCP slab thicknesses for the same inputs. Lastly, AASHTOWare Pavement ME Design was run to predict the performance of CRCP overlays in Illinois and compare the performance data of seven unbonded concrete overlays constructed in Illinois. AASHTOWare Pavement ME Design was then used to generate CRCP overlay thickness design tables for different traffic levels, shoulder types, and support conditions.

This paper discusses opportunities and challenges for the creation of evolving artificial neural network (ANN) and more general ? computational intelligence (CI) models inspired by principles at different levels of information processing in the brain ? neuronal-, genetic-, and quantum, and mainly ? the issues related to the integration of these principles into more powerful and accurate ANN models. A particular type of ANN, evolving connectionist systems (ECOS), is used to illustrate this approach. ECOS evolve their structure and functionality through continuous learning from data and facilitate data and knowledge integration and knowledge elucidation. ECOS gain inspiration from the evolving processes in the brain. Evolving fuzzy neural networks and evolving spiking neural networks are presented as examples. With more genetic information available now, it becomes possible to integrate the gene and the neuronal information into neuro-genetic models and to use them for a better understanding of complex brain processes. Further down in the information processing hierarchy, are the quantum processes. Quantum inspired ANN may help solve efficiently the hardest computational problems. It may be possible to integrated quantum principles into brain-gene inspired ANN models for a faster and more accurate modeling. All the topics above are illustrated with some contemporary solutions, but many more open questions and challenges are raised and directions for further research outlined.

Accelerated pavement testing (APT) applies intensive traffic loading for the purpose of observing long-term performance of pavements in a short time. The testing is beneficial for researchers as a timely and valuable data resource. As Mechanistic-Empirical (M-E) pavement design approaches emerge as a favorable alternative to traditional empirical approaches, there is a need for conducting calibration of the performance models in the M-E design framework. Data from APT testing may be considered advantageous for calibration due to research-grade accuracy conducted over shorter time intervals, but even so, M-E calibration is still time-consuming because many combinations of the respective calibration coefficients need to be evaluated against observed performance from pavement sections. The current version of the AASHTOWare™ Pavement ME Design software, though serving as a powerful pavement design and analysis tool, is not necessarily conducive to calibration studies as many interactions are needed by the user to populate the software with trial calibration coefficients and extract performance predictions after simulations have been completed. This study proposes using automation to facilitate calibration studies whereby the repetitive manual operations required by the software are recorded and built into a macro that greatly reduces the required human interactions. The software automation is demonstrated through a calibration study conducted at the National Center for Asphalt Technology (NCAT) Test Track that examined fatigue cracking calibration. The efficiency of the automated approach is compared against conducting calibration requiring human interaction with the software.

A new concrete pavement design method is currently introduced in Sweden. It consists of two parts: (i) computations of subgrade and base distresses and (ii) computations of concrete pavement distresses. To eliminate the difference between concrete and asphalt pavement design, the first part is carried out with equal requirements, in the same manner, and with the very same computer program. The computer run also provides a coefficient of subgrade reaction that is used as input in the second part. For this part, the American FE program ISLAB2000 is used to compute stresses – both traffic and thermal stresses, simultaneously – in various locations of the concrete pavement slab. Contrary to previous Swedish tradition, both bottom-up and top-down cracking are considered in the analysis of the computations. For the pavement designer, the division into two parts will not be visible. She will use the same user-friendly computer program (that previously has only been used for flexible pavement design) for both pavement alternatives. It is the authors’ hope that the selection of pavement solution for a certain case will be dependent solely on technical and economical factors and not on tradition and intensive advertising. The new design method includes a novelty concerning safety philosophy. The safety margin should be obvious and transparent and determined by the client in co-operation with the designer by selecting an appropriate safety class that is dependent on the economical consequences of a pavement failure but independent of material in the wear layer.

The AASHTOWare Pavement ME Design and the Guide for Mechanistic–Empirical Design of New and Rehabilitated Pavement Structures provide a methodology for analyzing response and predicting rutting performance of unbound granular pavement layers. However, this methodology has been reported to have low sensitivity to the properties of the base and subbase and the subgrade. The primary objective of this paper is to provide a framework for improving the prediction ability of the Pavement ME Design rutting model for unbound aggregate pavement layers. This goal was achieved by using accelerated testing data from full-scale pavement test sections constructed with different aggregate materials. These test sections were loaded to failure by using the accelerated transportation loading assembly at the University of Illinois. Nonlinear finite element solutions of the pavement test sections were studied so that stress-dependent modulus characterization models could be used for the anisotropic unbound aggregate base and isotropic subgrade. This approach enabled prediction of realistic stress states and accurate pavement responses. Realistic stress states and the use of a shear stress ratio concept were incorporated into the proposed framework needed to improve the rutting predictions of the Pavement ME Design damage model. Finally, a definite correlation appeared to exist between shear stress ratio values and the shift factors that were proposed for the adjustment of the model’s predicted rut depths so they would match the field-measured performance trends. The framework studied in this paper is applicable to the three-parameter Tseng–Lytton rutting model, the original version of the Pavement ME Design damage model, for the modifications needed to generate improved granular layer rut predictions throughout pavement design service life.

Flexible pavement design requires considering a variety of factors including the materials used, variations in water tables, traffic levels, and the climatic conditions the road will experience over its lifetime. Most pavement designs are based on historical climate variables such as temperature and precipitation that are already changing across much of the United States, and do not reflect projected trends. As pavements are typically designed to last 20 years or more, designs that do not account for current and future trends can result in reduced performance. However, incorporating climate projections into pavement design is not a trivial exercise. Significant mismatches in both spatial and temporal scale challenge the integration of the latest global climate model simulations into pavement models. This study provides a national-level overview of what the impact of climate change to flexible pavement could look like, and where regional focus should be placed. It also demonstrates a new approach to developing high-resolution spatial and temporal projections that generates hourly information at the scale of individual weather stations, and applies this as input to the AASHTOWare Pavement ME Design™ model. The impact of three different future climates on pavement performance and time to reach failure thresholds in 24 locations across the United States are quantified. Changes to projected pavement performance differ by location, but nearly all result in decreased performance under current design standards. The largest increases in distress are observed for permanent deformation measures, especially toward the end of the century under greater increases in temperature.

Currently, Costa Rica’s pavement design is typically carried out using methodologies from the AASHTO or through AASHTOWare Pavement ME. However, sometimes the pavement structure designs are not optimal because these methodologies do not fully adapt to climatic conditions, material properties, and local calibration factors. Some pavement structures fail prematurely because of inadequate designs, often resulting from omission or miscalculation of certain factors in the initial design phase. Environmental effects, such as precipitation and groundwater levels, and traffic loads are among the most challenging to quantify yet most destructive factors affecting pavement structures. Increased moisture content in subgrades affects their mechanical properties and structural contribution to the pavement. Given Costa Rica’s climatic conditions, which favor critical moisture in subgrades from a structural standpoint, modeling the moisture impact on material properties allows one to quantify its influence on pavement performance. Therefore, incorporating unsaturated soil mechanics into pavement design and analysis enables the creation of structures that account for varying subgrades moisture conditions throughout the year. This approach acknowledges that soils are not constantly at 100% saturation. To obtain this moisture variation, soil-water characteristic curves were obtained for each soil and, with this information, the resilient modulus (RM) test was performed using different water contents and suctions. This report proposes a calibrated model for local conditions that predicts RM of soils under varying moisture conditions. The inclusion of this model in Costa Rica’s mechanistic-empirical design guide is suggested.

The new Mechanistic-Empirical Pavement Design Guide (MEPDG) provides methodologies for mechanistic-empirical pavement design as opposed to the empirical methodology used in the 1993 American Association of State Highway and Transportation Officials (AASHTO) pavement design guide. The objective of this study was to compare the pavement designs obtained using the 1993 AASHTO and the new MEPDG methods for typical Portland Cement Concrete (PCC) and Asphalt Concrete (AC) pavements in Kansas. Five in- service Jointed Plain Concrete Pavement (JPCP) projects were reanalyzed as equivalent JPCP and AC projects using both approaches at the same reliability level. The results show that the new MEPDG analysis yielded thinner AC sections for all projects than those obtained from the 1993 AASHTO design guide analysis. Four of the PCC sections, designed using the 1993 AASHTO design guide, were thicker than the sections obtained with MEPDG. The MEPDG analysis resulted in thicker PCC slab for the fifth project. Effect of change in performance criteria on the thickness of AC and PCC sections has also been investigated. It has been found that AC sections are more sensitive to change in performance criteria as compared to PCC sections using MEPDG versions 1.0 and 1.1. In general, difference in thickness using both versions is not significant for all practical purposes.