A comprehensive evaluation algorithm for project-level bridge maintenance decision-making

Abstract As an indispensable process to identify the optimal treatment measures for bridge maintenance, the importance of decision-making techniques have been recently highlighted. In this paper, a multi-parameters project level bridge maintenance decision-making algorithm, conducted by the Probabilistic Neural Networks method, radial basis function method, and the Principal Component Analysis method, is investigated. A detailed study on the parameter selection, data classification, and prediction are performed by using the former two methods for the time-series index data, in terms of technical condition, average daily traffic, and sufficiency rating. The sub-item weight among different parameter indexes is calculated by the Principal Component Analysis method and evaluated by correlation evaluation. Besides, an overall evaluation model for the project level bridge maintenance decision-making is also suggested, in which the comprehensive state index model, target-based reliability model, the degradation rate factor model, and the life-cycle cost estimate and determination of maintenance time, are conducted. The optimal process is chosen and applied in three bridges as a case study. Results demonstrate that the optimal algorithm is preferable for bridge maintenance decision-making.