Phenotype Variability Mimicking as a Process for the Test and Optimization of Dynamic Facade Systems

Abstract

A genetic algorithm and an artificial neural network are deployed for the design of a dynamic multi-layered façade system that adapts in real-time to different weather and occupants’ needs scenarios. The outputs are a set of different performances of the façade insulation cushions, optimized by the previous run of the genetic algorithm. A façade system of ETFE cushions is considered for them to learn from environmental data models. Each façade cushion is set up as an artificial neuron that is linked to the behavior and temperature of the others. The proposed outputs are a set of different performances of the façade system that are optimized through running the genetic algorithm. Façade neurons are configured as genes of the system that is abstractly represented on a digital model. The computational model manages cushion patterns’ performances through several phenotypical adaptations, suggesting that the proposed facade system maximizes its thermal efficiency in different scenarios.