Different approaches have been suggested for determining the optimalmix of repair projects for a pavement network. These methods range fromrandom selection to sophisticated mathematical optimization models.This paper presents an analysis of several questions regarding theeffectiveness of three possible selection methods.First, the performance of three separate single year project selectionmethods on different size networks is assessed over a broad funding spectrum.The results indicate that as funding levels increase, the benefit obtained bydifferent selection methods converge. In addition, as the size of the networkincreases, the convergence tends to occur at progressively lower funding levels.Second, the effect of the performance prediction models on these sameselection methods is assessed by altering the coefficients of the models to predict both faster and slower deterioration of the network. The "select sets" ofprojects created by priority ranking selection and Knapsack IP selection at threeseparate funding levels are compared to determine how much variation is introduced by the changes in the performance prediction. With a 30%acceleration and deceleration of the deterioration curves, there was little changein the optimal project set created by either method.Finally, a modified Monte Carlo model is used to assess the generalshape of the solution space. The results suggest that the solution space is relatively flat except in the immediate vicinity of the optimum. This, in turn,suggests that a Monte Carlo approach to this problem would require a largenumber of trials to approximate the optimum. This finding conceptually supportsfindings in this study and others, as well as the intuitive observation, that randommaintenance and repair strategies perform poorly compared to more rationalapproaches. Since only a few sets of repair projects are near the optimum, thechances of a random selection matching one of these near optimal project sets are relatively small.
展开▼
