Using Genetic Algorithms and Markov Chain to Develop the Optimal Maintenance Strategy for Building's Facilities

• Main Authors: Chien-Kuo Yu, 尤建國
• Other Authors: Chung-Wei Feng
• Format: Others
• Language: zh-TW
• Published: 2006
• Online Access: http://ndltd.ncl.edu.tw/handle/77990924714507016557

碩士 === 國立成功大學 === 土木工程學系碩博士班 === 94 === Building facilities start to deteriorate after they are put in operation. In addition, building facilities must be operated at certain level to provide adequate service. Because of deterioration, building facilities could not perform well as planned if they are not properly maintained. Since deterioration is inevitable and maintenance requires certain amount of money, it is necessary for the building manager to develop a maintenance strategy that can keep facilities functional without going over the budget. However, the status of the building facility usually varies with time and is hard to predict. Moreover, different building facilities have different deterioration rates which would change over time according to the time interval and the maintenance method applied. Therefore, developing a maintenance strategy is not an easy task. Previous research on developing the maintenance strategy mostly took mathematical programming approach, which showed inefficient for large-scale problems in terms of computation efforts. In addition, there is lack of quantified model for predicting the status of the building facility. Therefore, it is needed to develop a maintenance model not only can estimate the status of the facility and the status after it has been but also minimize the cost of maintaining facilities. This research presents a model that develops the optimal maintenance strategy for building facilities by minimizing the net present value of the money spent over the planning horizon. At first the status prediction model of the building facility is developed according to the theory of the Markov-chain. This status prediction model is presented in the form of a transition probability matrix, in which the current status of any facility depends on the previous state of that particular facility. In addition, the status of the facility after being maintained is determined by the maintenance method applied. Then, an optimization model that incorporates the genetic algorithms and the aforementioned status model is developed. Results show that the proposed optimal maintenance strategy along with the computer implementation can provide an easy-to-use interface and satisfactory results for building managers to develop the optimal maintenance strategy which determines how long each facility which be maintained by what method.