Develop Performance Prediction Models Research Articles

Modeling retroreflectivity degradation of pavement markings across the US with advanced machine learning algorithms

Retroreflectivity is the primary metric that controls the visibility of pavement markings during nighttime and in adverse weather conditions. Maintaining the minimum level of retroreflectivity as specified by Federal Highway Administration (FHWA) is crucial to ensure safety for motorists. The key objective of this study was to develop robust retroreflectivity prediction models that can be used by transportation agencies to reliably predict the retroreflectivity of their pavement markings utilizing the initially measured retroreflectivity and other key project conditions. A total of 49,632 transverse skip retroreflectivity measurements of seven types of marking materials were retrieved from the eight most recent test decks covered under the National Transportation Product Evaluation Program (NTPEP). Decision Tree (DT) and Artificial Neural Network (ANN) algorithms were considered for developing performance prediction models to estimate retroreflectivity at different prediction horizons for up to three years. The models were trained with randomly selected 80% data points and tested with the remaining 20% data points. Sequential ANN models exhibited better performance with the testing data than the sequential DT models. The training and testing R2 ranges of the sequential ANN models were from 0.76 to 0.96 and 0.55 to 0.94, respectively, which were significantly higher than the R2 range (0.14 to 0.75) from the regression models proposed in past studies. Initial or predicted retroreflectivity, snowfall, and traffic were found to be the most important inputs to model predictions.

Consideration of Information Loss in the Transportation Asset Management Process

Transportation asset management (TAM) relies on data to inform decisions on the maintenance or improvement of a given asset. Two of the essential data elements in the TAM process include the inventory, and the condition of an asset over time. Often, the condition is described using several factors, such as the amounts and types of cracking, and those factors are then combined to form a condition index. The index is frequently the basis for developing performance prediction models. However, when the condition of an asset is defined using a measure comprised of several metrics, the risk exists that considerable information could be lost or distorted from index-related calculations. Although the importance of the loss of information depends on the decision framework in which the asset is being managed, it does distort the relationship between the asset deterioration and the factors that drive that deterioration. This paper describes information loss using data and indexes from the U.S. National Park Service pavement network. The effect of the loss of information on the ability to develop performance prediction models is demonstrated. The aim was to present an approach for calculating information loss along with strategies for minimizing that loss. The results showed how loss of information can affect the ability to develop performance prediction models. The primary conclusion of this study is that information loss and -distortion should be evaluated for all condition indexes in an effort to minimize that loss and distortion.

Performance Evaluation and Prediction of Preservation Asphalt Overlay in Louisiana

The Louisiana Department of Transportation and Development (LaDOTD) has been using thin asphalt overlay (2 in.) as a preservation technique for asphalt-surface pavements. Although LaDOTD has specified the triggers for its application, it was not always applied on the recommended timing. It is unclear how the application timing affects the overlay performance. Additionally, no studies have been conducted to determine optimal application timing for the overlay and predict its performance using the available tools such as AASHTOWare Pavement Mechanistic-empirical design (Pavement ME) software. The objectives of the study are (1) to evaluate field performance of the preservation asphalt overlay and develop performance prediction models, (2) to determine optimal application timing for the overlay, and (3) to investigate if Pavement ME models are able to predict performance of the overlay by modifying model coefficients. To achieve the objectives, 33 preservation asphalt overlay projects were selected, and their performance was evaluated and predicted, then the optimal application timing was investigated based on the benefit and cost analysis. The results indicate that 58% of the preservation asphalt overlays had 10–20 years’ service life, and 42% had 7–10 years’ service life in relation to Pavement Condition Index threshold of 70. Additionally, good precondition or flexible base (asphalt concrete or unbounded base) can result in a longer service life in the overlay. It is also found that an average value of optimal timing of the next overlay application is the 9th year after construction. Finally, the rehabilitation models of asphalt overlay in Pavement ME software could be used for preservation asphalt overlay by modifying model coefficients.

Development of framework for performance prediction of flexible road pavement in Nigeria using Fuzzy logic theory

ABSTRACT Though several attempts have been made in the past to develop performance prediction models for flexible road pavement in Nigeria, insufficient dataset remained the major problem of using such models. This study used Fuzzy logic theory which is capable of handling the challenge of imprecise dataset to develop a framework for performance prediction of flexible road pavement in Nigeria. A Fuzzy Inference System was developed using MATLAB software. The attributes used included; Initial Pavement Surface Condition, Age of pavement, Resilient Modulus of subgrade soil, Average Truck load per day, Average Annual Air Temperature and Rainfall to predict the Future Pavement Surface Condition (FPSC). The model was calibrated using observed logical behaviour of pavement to fit engineering experience and judgement. A goodness-of-fit test showed high level of accuracy at 81.63% which was validated at 74.13% using extrapolated dataset of the same source. The study proposed 5120 mutually exclusive decision rules for performance prediction of flexible road pavement based on permutation theory. Though there was no current and well-spread dataset that described the present pavement condition to calibrate the decision rules, a framework for performance prediction of flexible road pavement using Fuzzy logic theory was developed for pavement engineers in Nigeria.

Predicting the Retroreflectivity Degradation of Waterborne Paint Pavement Markings using Advanced Machine Learning Techniques

Waterborne paint is the most common marking material used throughout the United States. Because of budget constraints, most transportation agencies repaint their markings based on a fixed schedule, which is questionable in relation to efficiency and economy. To overcome this problem, state agencies could evaluate the marking performance by utilizing measured retroreflectivity of waterborne paints applied in the National Transportation Product Evaluation Program (NTPEP) or by using retroreflectivity degradation models developed in previous studies. Generally, both options lack accuracy because of the high dimensionality and multi-collinearity of retroreflectivity data. Therefore, the objective of this study was to employ an advanced machine learning algorithm to develop performance prediction models for waterborne paints considering the variables that are believed to affect their performance. To achieve this objective, a total of 17,952 skip and wheel retroreflectivity measurements were collected from 10 test decks included in the NTPEP. Based on these data, two CatBoost models were developed with an acceptable level of accuracy which can predict the skip and wheel retroreflectivity of waterborne paints for up to 3 years using only the initial measured retroreflectivity and the anticipated project conditions over the intended prediction horizon, such as line color, traffic, air temperature, and so forth. These models could be used by transportation agencies throughout the United States to 1) compare between different products and select the best product for a specific project, and 2) determine the expected service life of a specific product based on a specified threshold retroreflectivity to plan for future restriping activities.

Developing Performance Prediction Models for Pavement Management Systems in Local Governments in Absence of Age Data

Pavement Management Systems (PMS) use a strategic and data-driven approach to optimize budget allocation to various maintenance and rehabilitation (M&R) projects. Performance prediction models are used in PMS to determine the optimum timing for M&R interventions on every pavement section. Many local roadway agencies use empirical regression models which are based on past condition and age data. Often, these agencies are faced with limited resources for data collection and a high staff turnover rate, which all result in inadequate or unreliable construction history and pavement age data. This paper recommends a simple practical approach for local governments to develop performance prediction models in the absence of reliable pavement age data. Also, best practices for data pre-processing and validation of the model prediction capability are synthesized. Instead of using regression models based on condition and age, the pavement deterioration rate at each condition level is estimated. Similar to the Markovian transition probability concept, it is assumed that deterioration rates for every family of pavements are independent of time and only dependent on the current condition level. For every pair of subsequent condition measurements on a section, the difference in condition score is normalized by the difference in measurement time. These deterioration rates are then classified into bins based on the initial condition level for every pair of measurements. The average deterioration rate for all data records in each bin is then used to build a deterioration curve. This approach is demonstrated in this paper using real but anonymous agency data.

Models for estimation of TBM performance in granitic and mica gneiss hard rocks in a hydropower tunnel

The risk of excavation operations due to high capital costs can be reduced by correct estimation of machine performance. Many models have been proposed to study this issue, but considering the nature of the problem, it is rather difficult to estimate tunnel boring machine performance by simple linear prediction models. The purpose of the present study is to construct linear and non-linear multivariate prediction models to estimate TBM performance as a function of rock mass properties in granitic and mica-gniess rocks. For this purpose, rock properties and machine data were obtained from a historical TBM tunneling project in Norway and then the database was established to develop performance prediction models utilizing the linear and the non-linear multiple regression methods. This study proposes more accurate and practical statistical models compared to the previous ones based on multivariate regression analyses to estimate the performance of hard rock TBMs.

Predicting performance of impact hammers from rock quality designation and compressive strength properties in various rock masses

Predicting the performance of the impact hammers is one of the major subjects in determining the economics of the underground excavation projects in which they are utilized. Therefore, researchers have been attracted to developing performance prediction models for these machines. Physical and mechanical properties of rocks have been used to estimate the performance of impact hammers over the last few decades. In this study, the instantaneous breaking rate (IBR, m3/h) of an impact hammer used in construction of Levent-Hisarüstü metro tunnel (Istanbul) is recorded in detail. Sixty rock samples are obtained from tunnel route during the excavation of which the machine is employed. Physical and mechanical property tests are performed on the obtained samples. A data set including uniaxial compressive strength (UCS), rock quality designation index (RQD), Brazilian tensile strength (BTS), density (ρ), Schmidt hammer hardness (SHH), Shore scleroscope hardness (SSH), Cerchar abrasivity index (CAI), and IBR is formed. Regression analysis techniques are applied to the created data set in order to develop a performance prediction model. The investigation results in a model that can predict IBR based on UCS, RQD, and the output power of the impact hammer. The proposed model passes both F-test and t-test at 0.95 confidence level. The soundness of the model is successfully tested against two formerly developed models. Covering a wide range of application and requiring only two of the most common and versatile rock properties as input parameters are the other advantages of the suggested model.

Prediction of roadheader performance by artificial neural network

Performance prediction of the roadheaders is one of the main subjects in determining the economics of the underground excavation projects. During the last decades, researchers have focused on developing performance prediction models for roadheaders. In the first stage of this study, the performance of a roadheader used in Kucuksu sewage tunnel (Istanbul) was recorded in detail and the instantaneous cutting rate (ICR) of the machine was determined. The uniaxial compressive strength (UCS) and rock quality designation (RQD) are used as input parameters in previously developed empirical models in order to point out the efficiency of these models, and the relationships between measured and predicted ICR for different encountered formations. In the second stage of the study, Artificial Neural Network (ANN) technique is used for predicting of the ICR of the roadheader. A data set including UCS, RQD, and measured ICR are established. It is traced that a neural network with two inputs (RQD and UCS) and one hidden layer can be sufficient for the estimation of ICR. In addition, it is determined that increase in number of neurons in hidden layer has positive optimizing on the performance of the ANN and a hidden layer larger than 10 neurons does not have a significant effect on optimizing the performance of the neural network. Furthermore, probability of memorizing is being recognized in this situation. Based on this study, it is concluded that the prediction capacity of ANN is better than the empirical models developed previously.

Performance Modeling on the Basis of Application Type in Virtualized Environments

Virtualization technology plays an essential role in resource in modern large data centers while it also causes interference among virtual machines which co-located in common physical machine contending for the shared physical resources. In this paper, we study the performance prediction models in virtualized environment. Unclassified model developed from all types of applications is quite inaccuracy to predict performance of test applications as it is too general. We respectively develop models for each type of applications classified by the resources that they use. See5/C5 technology is used to determine the type of test application before executing its corresponding performance prediction model and linear regression technique is adopt to develop performance prediction models. Finally, comparing classified models with the unclassified one, the former get 0.038 average prediction errors for test applications while unclassified model arrives 0.609 average prediction errors.

Performance Prediction Models for Cracked, Seated, and Overlaid Concrete Pavements in California

Crack, seat, and overlay (CS&O) is a rehabilitation technique used on jointed plain concrete pavements in California. Although CS&O is widely used, few studies have attempted to develop performance prediction models for pavements that have been rehabilitated with this technique. This study evaluated CS&O performance for surface roughness and transverse reflection cracking. Performance data for CS&O sections in three geographic regions in California were used to develop regression performance models. These models can be incorporated into the California Department of Transportation pavement management system and used to predict future performance, assist with life-cycle cost analyses, and optimize the allocation of resources. The results of a sensitivity analysis suggest that section age is the most significant factor affecting the deterioration of CS&O pavements. Secondary factors include the annual equivalent single-axle load and layer thickness ratio. Section age as an independent variable would include the impact of environmental loading cycles. Other factors related to subgrade conditions, temperature variations, and construction variability could not be investigated because data were not available. However, their impact is inherent to the models.

Application of two non-linear prediction tools to the estimation of tunnel boring machine performance

Predicting tunnel boring machine (TBM) performance is a crucial issue for the accomplishment of a mechanical tunnel project, excavating via full face tunneling machine. Many models and equations have previously been introduced to estimate TBM performance based on properties of both rock and machine employing various statistical analysis techniques. However, considering the nature of the problem, it is relatively difficult to estimate tunnel boring machine performance by linear prediction models. Artificial neural networks (ANNs) and non-linear multiple regression models have great potential for establishing such prediction models. The purpose of the present study is the construction of non-linear multivariable prediction models to estimate TBM performance as a function of rock properties. For this purpose, rock properties and machine data were collected from recently completed TBM tunnel project in the City of New York, USA and consequently the database was established to develop performance prediction models utilizing the ANN and the non-linear multiple regression methods. This paper presents the results of study into the application of the non-linear prediction approaches providing the acceptable precise performance estimations.

Staged survival models for overlay performance prediction

Asphalt concrete (AC) resurfacing is used by highway agencies to maintain and improve existing pavement conditions. With a large portion of their budgets being spent on road maintenance and rehabilitation, the agencies are interested in identifying the optimal timing of resurfacing and its life expectations. Using Kentucky's pavement management data, the research as presented by this paper investigated the procedures and methods of establishing thresholds and developing performance prediction models for AC overlays. Both transition probability matrix and data plots were used to investigate the performance trend of overlays. Based on the trend, two stage survival models were proposed to predict the resurfacing cycles. The first stage model can be used to predict the useful life of an overlay before a new resurfacing is required, while the second one can be used to predict the time window during which the new overlay must be applied to avoid accelerated pavement deterioration.