Markov process for deterioration modeling and asset management of community buildings

The current management process for community buildings in Australia is mainly reactive. Data collected using regular inspections are used for maintenance decision making in the period between consecutive inspections, disregarding the future degradation of the assets and the resultant levels of service. Forecasting future conditions using historic data is difficult because of the uncertainty and stochastic nature of deterioration. A major gap in knowledge is the lack of methods for predicting this highly uncertain degradation process for components of community buildings to support a strategic decision-making process. This paper presents a Markov process-based method for deterioration prediction of building components using condition data collected by the City of Kingston in Australia. Markov transition matrices for building components have been derived using a modified method combining the genetic algorithm with Monte Carlo sampling called direct absolute value difference, which offers superior accuracy. The derived matrices are validated using a new data set collected in 2011. Fourteen transition matrices for building components are proposed. The paper presents a typical decision-making method based on the Markov process.