哈尔滨工业大学（深圳）学术讲座

演讲人Speaker：Prof. Peng ZHANG (张鹏)

题目Title: Dimension Reduction and Mode Recognition for Time-series Data of Complex Combustion Systems

时间Date：2024年 12月 11日     Time：10:00-11:30

地点Venue：H栋307室 

内容摘要Abstract：

Time-series data is a sequence of observable quantities in natural temporal ordering. Studying time-series data is crucial for identifying and predicting anomalies, trends, and patterns, which are vital for informed decision-making in many scientific and engineering fields. This talk will present our recent progress in deep learning (DL) analysis of time-series data collected from numerical simulation, lab experiments, and industry, focusing on dimension reduction, dynamic mode recognition, and anomaly detection. Identifying the collective behaviors of complex nonlinear systems is challenging because their time-series data is often high-dimensional. Based on the neural networks of Variational Autoencoder (VAE), we developed a comprehensive DL framework for dynamical mode recognition using the supervised and unsupervised classification approaches of Wasserstein distance (WD) and Dynamic Time Warping (DTW) in the phase space. This classification framework performs better in distinguishing dynamical patterns in complex combustion systems, such as flame oscillator system, annular turbulent flames, and supersonic combustion scramjet, indicating its potential for studying industrial turbulent combustion.

个人简介(About the speaker):

Dr. Peng Zhang joined the City University of Hong Kong as an associate professor in 2022. Before that, he was an assistant and then associate professor at the Hong Kong Polytechnic University during 2012-2022. He received a Ph.D. degree in Mechanical and Aerospace Engineering from Princeton University in 2010 and worked as a Combustion Energy Research Fellow at Princeton University during 2010-2012. He is the founding chairman of ILASS (Institute for Liquid Atomization and Spray Systems)-Hong Kong and the local committee chairman for the 16th ICCEU (International Conference on Combustion and Energy Utilization). Dr. Zhang’s current research areas are droplet and spray dynamics, theoretical and numerical combustion, theoretical chemical kinetics, and AI for science. He has published more than 110 papers in international peer-reviewed journals in these areas.