Asset Management System for Educational Facilities Considering the Heterogeneity in Deterioration Process

In asset management of infrastructures, predicting deterioration of structures is one of an essential technique to make a decision on the optimal maintenance policy. However, as for large-scaled infrastructures that consist of a huge number of structural components, in order to estimate their deterioration process with high accuracy, the heterogeneity of individual components has to be considered because each component possesses different material characteristics and designs and is in unique service under various environmental conditions. This paper focuses on especially educational facilities from among infrastructures, and constructs its asset management system considering the heterogeneity in deterioration process of individual components. This system mainly consists of 3 functions as: 1. database, 2. deterioration prediction and 3. life-cycle cost evaluation. The database stores the basic structural and component’s information and visual inspection data. Based on these information and data, the deterioration prediction is statistically carried out. Specifically, the deterioration process can be basically expressed by random proportional hazard model, and the heterogeneity can be modeled as probability fluctuation in the hazard rate. Furthermore, the time-dependent hazard rate is formulated by the Weibull hazard model. The heterogeneity of the hazard rates across the individual characteristics of components is explained by the random proportional Weibull hazard model in which the hazard rates are subject to Gamma distribution. In the 3rd function, through the comparisons of the life-cycle costs between the multiple repair/renew strategies, the optimum one is decided. Here, as the deterioration process of individual components can be formulated by the Markov transition probabilities defined by the estimated hazard rates, the proposed life-cycle cost evaluation method organically links to deterioration prediction results via Markov decision process. In addition, an empirical study employing visual inspection data for an actual university facility is carried out to verify the validity and applicability of the system.