Neural networks applied to service life prediction of exterior painted surfaces

Service life prediction is assuming a primary role as it allows a more rational use of scarce resources; its methods are useful for defining preventive maintenance plans, thereby increasing performance and reducing costs. A new mathematical model is presented that uses artificial neural networks to evaluate the service life of painted surfaces. The data on facade degradation were collected from field observations on 160 buildings (220 painted surfaces) in Lisbon, Portugal, examining several degradation agents. In service conditions, the mean estimated service life of exterior painted surfaces is found to be 9.49 years, with a standard deviation of 0.633 years. Detailed factors are identified and incorporated into the model, which account for variations in degradation. Some statistical parameters are used to evaluate the validity and efficiency of the model. The values obtained are consistent with the existing perception relative to the durability of painted coatings. These values can be used to evaluate the economic and environmental performance of painted surfaces throughout their life cycle. The use of this model can optimize inspection and maintenance plans as well as the implementation of inform decisions at the design and construction stages.