The major objective of a pavement maintenance decision support system (PMDSS) is to assist decision makers in selecting an appropriate maintenance and repair (M&R) action for a defected pavement. This is typically performed through collecting condition data, analyzing and reducing condition data (e.g., development of condition indices), and selecting appropriate M&R actions. This paper reveals the results of implementing artificial neural networks (ANN) to recommend appropriate M&R actions. For an ANN to diagnose an M&R action accurately, it must be trained with correctly diagnosed M&R actions (training sets). Each training set consists of a pavement condition represented by deduct values for each distress present in the pavement and the corresponding recommended M&R action. Pavement condition data used in this study were obtained from comprehensive visual inspection data conducted on the Riyadh road network in Saudi Arabia. The associated M&R actions were obtained based on consulting human expertise and M&R actions recommended by PMDSS software. Results of this study reveal that ANN is appropriate for implementation in identifying appropriate M&R actions.
Read full abstract