Machine learning model to predict impact of climate change on facility equipment service life

Predicting the service life of facility equipment is a crucial aspect of facility design and maintenance that requires careful assessment of a wide range of building and environmental factors. Climate change may cause these environmental stressors, like temperature intensity, to deviate from historical trends, implying that past design assumptions may not reflect future service conditions. Therefore, additional techniques are needed to associate the effects of climate on equipment service life. This paper develops a machine learning method capable of modeling complex building and environmental factors, including climate change, on the service life of equipment such as air-handling components. A boosted decision tree model is developed to predict equipment service life based on the combined impact of building and environmental factors. The developed model outperforms other approaches, such as industry-standard approaches, Weibull models, and inspection-based techniques for estimating equipment service life. The model showed a 93.9% reduction in mean-squared-error when estimating component service life compared to the industry standard model. The developed machine learning model is then combined with climate projection models to measure the impact of predicted temperature intensity changes on the service life of air-handling units, showing an expected 12%-41% decrease in the service life of air-handling units in Yuma, Arizona, when subjected to 2060 climate projections.