Asphalt Pavement Performance Evaluation And Prediction Models

Pavement Management System has achieved great application and development since our country introduced British asphalt pavement management in 1984. This system provide not only scientific approach and analyzing tool for the kinds of department of pavement management institution, but also help to make scientific management and decision to ensure the good service level of the pavement. Pavement performance evaluation and prediction is the two most important parts of pavement management system, they are the base to understand pavement condition and change regulation, make right strategy and corresponding maintenance, reconstruction decision. Therefore, establishing and realizing the pavement performance evaluation and prediction model is the key point of pavement management.Pavement performance evaluation and prediction are the mostly research contents in this paper. Based on analyzing the present situation of the pavement performance evaluation, advantages and shortages of every existing model are presented. The first part of the paper introduces common pavement performance evaluation index and corresponding codes and standards. According to pavement performance characteristics and evaluation index feature, a new method Genetic Algorithms-based BP Neural Network model is introduced to establish pavement performance evaluation model. This structure has great character; it can not only make good use of many advantages of Neural Network, such as highly nonlinear, good fault tolerant capability, strong self study, self adapting, self organizing, but also overcome the deficiency of BP Neural Network model which is easy to be trapped in the local extremum and has low rapidity of convergence. The global search ability of GA covers the shortage of BP network and save the training time. The two part of the paper presents the pavement performance prediction model. It also adopts Genetic Algorithms-based BP Neural Network to establish prediction model. In this model, smoothness standard deviation, pavement intensity index and pavement cornering ratio are network inputs, output is pavement condition index. Actual sample is used to test the evaluation and prediction model. After the network training, simulating and testing, it is proved that this model has high evaluating and predicting precision, good extrapolating character, straightforward construction. Therefore, it is an effective method to evaluate and predict for pavement performance system.MATLAB7.0 is the mainly programming tool for this paper. Artificial Neural Network is established by the Neural Network Toolbox of MATLAB. Genetic Algorithm is realized by the Genetic Algorithm Toolbox developed by University of Sheffield, UK. This structure mainly use Genetic Algorithm to optimize the weight and threshold of BP Neural Network, then training the network with optimized weight and threshold so that the network can trip out of the local extremum, and achieve precision of convergence with minimal iterate.