Telegram Facebook

EACE Journal — Vol. 13, No. 2

Artificial Neural Network-Based Distress Prediction Model: A Case of ERA Asphalt Concrete Pavements

Belete Yismachew, Asregedew Woldesenbet

Abstract

Abstract In pavement management, predicting the amount of distress of pavement plays a substantial role to determine the overall level of performance of a pavement. In current practice, even though the Federal Government of Ethiopia is investing large amounts of money on road construction projects every year, poor & untimely maintenance and rehabilitation is leading to the deterioration of pavement performance. The aim of this research is to develop artificial neural network (ANNbased) models in order to predict pavement distresses as a decision making tool for planning, budgeting, prioritizing and optimizing maintenance and rehabilitation strategies throughout a pavement service time. The study collected asphalt concrete pavement distresses such as cracking, rutting, raveling, pothole and age break from ERA Road Asset Management Department covering about 179 case study projects spread across the nation’s road network. Goodness of Fit (R2) and Root Mean Square Error (RMSE) were used to evaluate the developed models. The trained models (85%) have been validated using validating dataset (15%). R2 between ANN predicted vs. observed distresses to both training and validating dataset showed a value of 0.84 to 0.97 for the developed ANN distress prediction models. The results identified significant independent variables or factors affecting asphalt concrete pavement distresses. Average Annual Daily Freight Vehicles (AADFV), Adjusted Structural Number (SNP), Drainage Condition (DC), Annual Rain Fall (ARF), Average Annual Traffic (AADT) and Service Life (SL) were significant factors affecting rutting, while Service Life (SL) Minimum temperature (MinT), Average Annual Daily Traffic (AADT), AADFV and ARF were identified as critical factors affecting raveling. In addition, SNP, AADT, Equivalent Single Axel load (ESAL) and AADFV are contributing factors for potholes, while Deflection (Def), DC, AADFV, ESAL, SNP and maximum temperature (MaxT) are the root causes for cracking. Furthermore, ESAL, SNP, ARF, Def and AADFV are critical factors affecting edge break distress. The developed models can serve as a useful tool in the decision-making process for maintaining road assets. Keywords: Asphalt Concrete Pavement, Pavement Management System, Distresses, Artificial Neural Network, Multilayer Perceptron, Prediction Model

Read Full Article (PDF)