Ye Ma

Machine learning techniques applied to expedite the performance-driven design of hygro-sensitive kinetic façade

My research aims to accelerate the generative design of hygro-sensitive kinetic façade according to the machine learning techniques and the empirical data. Although the previous studies identified the influential parameters on the hygroscopy of wood, the weight of the parameters has not been quantified. Accordingly, the methods producing the prototypes required the extensive research on the composite, which repeatedly went through the similar experiment procedure. This proposal mainly targets to figure out the weights of the influential parameters and predict the new composite combination by building up the framework of machine learning models. The framework integrates the learning, generation, and evaluation process in the form of visual output. The five different learning models are applied for the enhancement of data, evaluation of models, and prediction of structural performance of new composite: support vector machine (SVM), Gaussian Naive Bayesian (GNB), generative adversarial network (GAN), convolutional neural network (CNN), and deep neural network (DNN). The contribution of this research would offer the clear relationship among input parameters and a real-time simulation tool as use of environment control, which extends the application scenarios of the hygro-sensitive kinetic façade.

I am currently working on the Ph.D. degree in Constructed Environment under supervision of Dr. Ehsan Baharlou at University of Virginia. I have gained the M.Arch degree at University of Southern California and B.Arch degree at Xi’An University of Architecture and Technology. The undergraduate experience on craftmanship and heritage conservation has led me to rethink the current direction of fabrication and materials. I gradually realized the possibility on the exploration of the mass customization with aid of the latest technology like digital fabrication and responsive materials after the hands-on experience at USC. Then the practice on the research and projects, concerning design, shape shifting structure, computer vision, and so forth, offered me the opportunity to be exposed to interdisciplinary synergy and encouraged me to bolster my skillset. Combining my interest in advanced methodology under the diversified background, I decided to join the vibrant community of UVA for pursuing the Ph.D. degree.