Accurate prediction of hygrothermal behavior in the concrete is vital requirements to make more realistic service-life extension decisions. In this work, a neural network based hygrothermal prediction model to estimate a temporal hygrothermal condition in surface-protected concrete façade members is developed and presented. The model learns the case-specific features of hygrothermal behavior using the two years temperature and relative humidity data obtained from the installed probes. The performance evaluation confirms that the model describes the hygrothermal behavior inside the concrete façade with a high accuracy. This in turn enables to assess the corrosion rate as well as deterioration risk levels caused by frost and chemical attacks while identifying the appropriate surface protection system.