Application of back-propagation artificial neural network to predict maintenance costs and budget for university buildings

The study focuses on the operation maintenance phase of buildings on the National Taiwan University campus. Using historical data on maintenance and repair over a 40-year period, life-cycle cost analyses are conducted based on the statistical quantization methods and expert opinions. The study in connection with periodic maintenance; non-periodic repair and demand change, for theses three types of maintenance management. Moreover, multiple regression analysis and back-propagation artificial neural network (BPN) are used to establish a cost model for predicting maintenance costs. The age of the building, number of storeys, and elevator facilities are used as independent variables to estimate maintenance costs. The study helps to set a legitimate standard for arranging repair maintenance costs, and proposes a plan and standard for the repair maintenance strategy of the structures.