AbstractComposite pavement structure is the main pavement structure type for airport runways worldwide. However, the damage features of composite runway pavement are different from both flexible pavement and rigid pavement. Preventive maintenance (PM) represents a sufficient way to maintain the pavement structure in a state of good serviceability, but few proper methodologies for PM decision making are available. This paper tries to establish a new set of indexes for PM decision making at the project level and solve the corresponding thresholds that were based on the distress type from field investigations and randomly generated data. The drawbacks of current indexes for pavement condition evaluation were analyzed, and 15 pavement distress types were divided into 4 categories in accordance with the damage mechanisms. The new set of PM decision-making indexes was established based on the new classification, which can strongly differentiate the pavement distress. To increase the reliability of the thresholds, an extended distress database was established from the initial distress database under constraint conditions. In the end, the thresholds of PM decision making for a composite pavement structure can be obtained based on an expert system for maintenance decision and a back-propagation (BP) neural network method. Results revealed that the indexes and thresholds could optimize PM decision making for composite pavement runways quantitatively and efficiently.