[原分所演講] IAMS lecture on June 26, 2:00-4:00 pm, C.T. Chang Memorial Hall, Prof. Akeel Shah
消息來源：Academia Sinica
截止日期：2017-06-23
IAMS Lecture Announcement
中研院原分所演講公告
Title: Future challenges for modelling electrochemical devices
Speaker: Prof. Akeel Shah (School of Engineering, University of Warwick, UK)
Time: 2:00-4:00 PM, June 26 (Monday), 2017
Place: C.T. Chang Memorial Hall (4F), IAMS (本所張昭鼎紀念講堂 臺大校園內)
Contact: Dr. Kuei-Hsien Chen 陳貴賢博士
Abstract: Modelling and simulation are well-established tools in many fields, e.g., aerospace and energy technologies. Limitations of traditional computational modelling approaches have led in recent decades to the growth in alternative methods and paradigms, such as ab-initio and multi-scale modeling (to account for small scale effects averaged out in continuum models) as well as data-driven and model-order reduction (MOR) techniques (to deliver rapid results for applications such as optimization, control and uncertainty quantification). In this talk, we describe efforts by our group to develop data-driven and MOR approaches for models in science and engineering, with a focus on complex models of energy storage devices. This includes the use of machine learning and stochastic techniques (e.g., Gaussian process models, support vector machines) combined with manifold learning (e.g., diffusion maps and local tangent space alignment) to approximate maps between input spaces and very high dimensional (>104) output spaces. MOR approaches based on extended proper-orthogonal decomposition for parametric problems are also discussed. We discuss future directions, including the use of Deep Learning and Gaussian process latent variable models and outline some of the latest results.
Keywords: Batteries and fuel cells, mathematical models, supervised/unsupervised machine learning, Galerkin reduced order models
Contact information: School of Engineering, University of Warwick, Coventry CV47AL, UK, (tel: 07979465001; e-mail: Akeel.Shah@warwick.ac.uk).
Short biography: Akeel Shah graduated with a first class honours degree in Mathematical Physics from Manchester University in 1995. Between 1997 and 2000 he undertook a PhD in Applied Mathematics, also at Manchester University, before two postdoctoral positions in University of Leeds and University of British Columbia, where his interests broadened to mathematical modeling of fuel cells. He has held faculty positions at University of Southampton and University of Warwick, with interests in he development and modeling of battery and fuel cell systems. He has worked with several companies including Ballard Power Systems and Johnson Matthey Fuel Cells. More recently his interests have extended to uncertainty quantification for general scientific and engineering problems, with a particular emphasis on field inputs and outputs (high dimensional spaces) using techniques for dimensionality reduction and emulation (machine learning and model order reduction). He currently leads a group of 3 PhD students and 2 postdoctoral researchers with funding in excess of £1m.
中研院原分所演講公告
Title: Future challenges for modelling electrochemical devices
Speaker: Prof. Akeel Shah (School of Engineering, University of Warwick, UK)
Time: 2:00-4:00 PM, June 26 (Monday), 2017
Place: C.T. Chang Memorial Hall (4F), IAMS (本所張昭鼎紀念講堂 臺大校園內)
Contact: Dr. Kuei-Hsien Chen 陳貴賢博士
Abstract: Modelling and simulation are well-established tools in many fields, e.g., aerospace and energy technologies. Limitations of traditional computational modelling approaches have led in recent decades to the growth in alternative methods and paradigms, such as ab-initio and multi-scale modeling (to account for small scale effects averaged out in continuum models) as well as data-driven and model-order reduction (MOR) techniques (to deliver rapid results for applications such as optimization, control and uncertainty quantification). In this talk, we describe efforts by our group to develop data-driven and MOR approaches for models in science and engineering, with a focus on complex models of energy storage devices. This includes the use of machine learning and stochastic techniques (e.g., Gaussian process models, support vector machines) combined with manifold learning (e.g., diffusion maps and local tangent space alignment) to approximate maps between input spaces and very high dimensional (>104) output spaces. MOR approaches based on extended proper-orthogonal decomposition for parametric problems are also discussed. We discuss future directions, including the use of Deep Learning and Gaussian process latent variable models and outline some of the latest results.
Keywords: Batteries and fuel cells, mathematical models, supervised/unsupervised machine learning, Galerkin reduced order models
Contact information: School of Engineering, University of Warwick, Coventry CV47AL, UK, (tel: 07979465001; e-mail: Akeel.Shah@warwick.ac.uk).
Short biography: Akeel Shah graduated with a first class honours degree in Mathematical Physics from Manchester University in 1995. Between 1997 and 2000 he undertook a PhD in Applied Mathematics, also at Manchester University, before two postdoctoral positions in University of Leeds and University of British Columbia, where his interests broadened to mathematical modeling of fuel cells. He has held faculty positions at University of Southampton and University of Warwick, with interests in he development and modeling of battery and fuel cell systems. He has worked with several companies including Ballard Power Systems and Johnson Matthey Fuel Cells. More recently his interests have extended to uncertainty quantification for general scientific and engineering problems, with a particular emphasis on field inputs and outputs (high dimensional spaces) using techniques for dimensionality reduction and emulation (machine learning and model order reduction). He currently leads a group of 3 PhD students and 2 postdoctoral researchers with funding in excess of £1m.