Network-level maintenance decisions for flexible pavement using a soft computing-based framework

Abstract

An effective pavement management system (PMS) is one that is guided by a software program that ensures that all pavement sections are maintained at adequately high serviceability levels and structural conditions with a low budget and resource usage, without causing any significant negative effect on environment, safe traffic operations and social activities. PMS comprises of section classification; performance prediction; and optimization for decision-making. For section classification, this research presents a fuzzy inference system (FIS), with appropriate membership functions for section classifications and for calculating the pavement condition index (PCI). The severity and extent of seven distress types (alligator cracking, block cracking, longitudinal and transverse cracking, patching, potholes, bleeding and raveling) were used as fuzzy inputs. The result showed a good correlation for fuzzy model. A sensitivity analysis showed a pavement crack has the greatest influence on section classification compared to the other distress types. A novel network level deterministic deterioration model was developed for flexible pavement on arterial and collector roads in four climatic zones considering the impact of maintenance, age, area and length of cracks, and traffic loading. The prediction models showed good accuracy with high determination coefficient (R2). The cross validation study showed that the models for arterial roads yield better accuracy than the models for collector roads. A sensitivity analysis showed that the area and length of cracks have the most significant impact on the model performance. A novel discrete barebones multi-objective particle swarm algorithm was applied for a discrete multi-objective problem. Conventional particle swarm optimization (PSO) techniques require a manual selection of various control parameters for the velocity term. In contrast, the bare-bones PSO has the advantage of being velocity-free, hence, does not involve any parameter selection. The discrete barebones multi-objective PSO algorithm was applied to find optimal rehabilitation scheduling considering the two objectives of the minimization of the total pavement rehabilitation cost and the minimization of the sum of all residual PCI values. The results showed that the optimal maintenance plan found by the novel algorithm is the better than found by conventional algorithm. Although the results of performance metrics showed that the both algorithms perform on a par, the novel algorithm is clearly advantageous as it does not need parameter selection.