Invited Speakers

Alexander Lin

National University of Singapore, Singapore

Alexander Lin is a Senior Lecturer in Department of the Build Environment at the National University of Singapore (NUS). He holds a Bachelor's degree in Civil Engineering from National Taiwan University and a PhD from the Structural Engineering Mechanics and Materials program at UC Berkeley's Department of Civil and Environmental Engineering. Initially focusing on advanced concrete materials and their applications in structures, his research has evolved to incorporate the innovative use of technology and 3D printing, customizing shapes and details of building components for improved performances. His current work centers on optimizing building configurations, employing advanced algorithms and AI to enhance not only structural performance but also operational performances including acoustic and thermal insulations. He is a co-lead for the Thrust of AM (Additive Manufacturing) Enabled Design and Environment in Centre for Additive Manufacturing in NUS.

 

Speech title "Integrating Graph Neural Networks and Statistics-based Genetic Algorithm for Acoustic Optimization of Hollow Partition Wall with Functionally Graded Configuration"

Abstract-This research proposes a novel method for optimizing acoustic insulation in partition walls made with hollow concrete blocks. Traditional optimization techniques, which rely on computationally intensive finite element methods (FEM) and simple Genetic Algorithms (GA), are often inefficient. To address this issue, our research proposes a novel approach which combines Graph Neural Networks (GNN) and a probability-informed GA. GNN for acoustic performance evaluation and the probability-informed GA for optimization are integrated to design advanced partition walls using Functionally Graded Materials (FGMs). Various GNN models were developed and compared, with Graph Transformer Network (GTN) achieving the highest predictive accuracy and was selected as the assessment tool for the optimization process. The probability-informed GA, leveraging Beta distributions to guide mutation points, demonstrated superior performance over conventional GA, particularly at higher mutation rates. Our proposed combined optimization method achieved significant transmission loss improvement of 36.92 dB at 500 Hz, a frequency range typically associated with traffic noise from heavy vehicles. Our approach not only enhances computational efficiency, but also provides a robust solution for developing advanced acoustic insulation materials through implementation of FGMs for built environment applications, with potential applications to broader building performance improvements, including lightweight structures and thermal insulation.