详情

原文传递 PREDICTING ROUGHNESS PROGRESSION IN FLEXIBLE PAVEMENTS USING ARTIFICIAL NEURAL NETWORKS.

题名：	PREDICTING ROUGHNESS PROGRESSION IN FLEXIBLE PAVEMENTS USING ARTIFICIAL NEURAL NETWORKS.
作者：	Attoh-Okine-NO
关键词：	CONFERENCES-; PERFORMANCE-PREDICTION-MODELS; PAVEMENT-PERFORMANCE; DETERIORATION-; PAVEMENT-CONDITION; PAVEMENT-MANAGEMENT; DECISION-MAKING; ROUGHNESS-; FLEXIBLE-PAVEMENTS; FIELD-DATA; PAVEMENT-MANAGEMENT-SYSTEMS; ARTIFICIAL-NEURAL-NETWORKS; KNOWLEDGE-PROCESSING; PATTERN-RECOGNITION
摘要：	To develop a balanced expenditure program for a highway network, the rate of deterioration of the pavement and the nature of changes in the condition need to be predicted so that timing, type, and cost of maintenance can be estimated. A pavement deterioration model, or pavement performance, is therefore a key component of the analysis supporting pavement management decision making. Models for predicting roughness progression have been developed on the basis of traffic and time-related models, interactive time, traffic, or distress models. These models differ in form, in level of initial roughness, and in the influence of roughness on the subsequent progression rate. A characteristic feature of the models is that they are formulated and estimated statistically from field data. To date, modeling pavement performance has been extremely complicated; no pavement management system (PMS) can consider more than a few of the parameters involved, and then only in highly simplified manner. The capabilities of artificial neural networks (ANNs) are evaluated in predicting roughness progression in flexible pavement from structural deformation, which is the function of modified structural number, incremental traffic loadings, extent of cracking and thickness of cracked layer, incremental variation of rut depth; surface defects, which are the function of changes in cracking, patching and potholing; and environmental and non-traffic-related mechanisms, which are the function of pavement environment, time, and roughness. ANNs have attracted considerable interest in recent years because of growing recognition of the potential of these networks to perform cognitive tasks. The tasks include prediction, knowledge processing, and pattern recognition. ANNS offer a number of advantages over more traditional statistical prediction methods: they are capable of generalization, and because of their massive parallelism and strong interconnectivity, they are capable of offering real-time solutions to complex problems. The back-propagation algorithm, which uses supervised learning, is used to train the networks.
总页数：	Conference Title: Third International Conference on Managing Pavements. Location: San Antonio, Texas. Sponsored by: Transportation Research Board
报告类型：	科技报告