United States Department of Transportation
i
Prediction of Road Condition and Smoothness for Flexible and Rigid Pavements in Louisiana Using Neural Networks
-
2024-03-01
-
[PDF-3.67 MB]
Details:
-
Creators:
-
Corporate Creators:
-
Corporate Contributors:
-
Subject/TRT Terms:
-
Publication/ Report Number:
-
Resource Type:
-
Geographical Coverage:
-
Edition:Final Report August 2020 – June 2023
-
Corporate Publisher:
-
Abstract:The Louisiana Department of Transportation and Development (DOTD) currently utilizes pavement performance prediction models in treatment selection and budget planning. These models are solely based on the non-linear curve-fitting regression of existing pavement condition data available in its pavement management system (PMS) database. The objective of the research was to develop short-term and long-term pavement performance prediction models to estimate future pavement condition and smoothness for flexible and rigid pavements using artificial neural network (ANN) modeling. To achieve the objective, two different pavement condition datasets were prepared– one each for short-term and long-term pavement performance prediction. The datasets were assembled based on DOTD’s PMS and other pavement project management data sources. A feedforward neural network technique was used in the training, validation, and testing of the ANN modeling. Specifically, this study developed three groups of ANN pavement performance prediction models: 17 individual neural network models for short-term federal-designated cracking percent prediction, 8 incremental ANN models for long-term asphalt overlay pavement performance prediction, and 5 ANN-based regression models for asphalt pavement family curve generation. The developed short-term models will be used to support DOTD’s prediction of 2- and 4-year performance target values for federal pavement condition assessment. On the other hand, the incremental long-term performance models can be utilized to forecast pavement condition even with limited historical performance records, which are insufficient for developing site-specific curves. The developed ANN-based family curves, which incorporate additional factors such as climate and traffic, may replace the current family curves used in DOTD’s PMS with improved accuracy and flexibility.
-
Format:
-
Funding:
-
Collection(s):
-
Main Document Checksum:
-
Download URL:
-
File Type:
