Multimodal data processing for building material property predictions

Building materials significantly impact a building's environmental footprint. Material inventories promote building element reuse, yet documentation of material property data for specific buildings is scarce and often confined to closed repositories. Effective material-related performance assessments demand knowledge of building characteristics, including material types, properties, and environmental indicators. Traditional (unimodal) prediction models predict material properties but overlook the complexities of real-world applications, lacking contextual insights from diverse data sources. This study introduces a multimodal data processing method that incorporates machine learning for material property prediction, integrating knowledge engineering to extract building characteristics as contextual information, and validates it on a residential brick facade.