| This dissertation describes a new method of finding near optimal maintenance and rehabilitation strategies for road pavements, at both the project and network levels, using genetic algorithms and multi-objective optimization. A strategy, in this case, is defined as a set of decisions (for instance what maintenance action to take in each year of an analysis period). The methodology differs from other similar methods in a number of important respects.; Unlike many methods which optimize or prioritize based on a single objective (resulting in a single optimal strategy), this method allows generation of efficient sets of strategies which are optimal or near optimal for various combinations of user defined multiple objectives. These strategies can then be plotted, based on their objective values, to form surfaces in as many dimensions as there are objectives. Thereafter, users may visualize the tradeoffs between their objectives and choose strategies accordingly.; The generation of efficient sets at project level is also shown to allow decomposition of restricted cases of the network optimization problem into project level sub-problems. This insight allows solution of these network problems, while retaining location information, for large networks. It is believed that this is a new accomplishment. However, the common network optimization problem involving annual budget restrictions is not accommodated.; Because of the flexibility of genetic algorithms, almost any optimization criteria and prediction models can be used in the optimization. Criteria can thus include different distresses, costs, user delays or safety considerations. The only necessity is that criterion values can be predicted for any project or network strategy. It is also possible to define different models for each specific project level section.; The specific criteria and prediction models used as examples in this dissertation are linked to deterioration of the pavement and costs. General prediction models for these criteria are thus developed and discussed. A new type of deterioration/restoration modeling method termed ‘interactivity modeling’ is introduced, as well as methods to generalize this ‘point’ modeling to distress extent and severity distributions over the pavement. A chapter on cost modeling also introduces a method of incorporating economies of scale into the optimization system. |
