Academic Report Notice of Haibin Chang: Deep Learning Experience in Seepage Inverse Problems

发布者:王健发布时间:2023-11-14浏览次数:19

Speaker: Mr. Haibin Chang , Associate Researcher

Affiliation: School of Energy and Mining, China University of Mining and Technology (Beijing)

Title: Deep Learning Experience in Seepage Inversion Problems

Time: November 15, 2023 (Wednesday) 16:00

Location: 290 Arts and Sciences Building

Abstract:

There are some traditional methods in the research of all scientific fields, but most of them tend to have some technical bottlenecks and limitations. Deep learning method is a hot research topic in recent years, which has advantages in image data, time series data processing, complex function relationship approximation, intelligent decision-making and so on. How to choose and improve deep learning methods reasonably and apply them in various research fields to solve the technical bottlenecks of traditional methods is a question worth thinking about. Taking the seepage inversion problem as an example, the presenter talks about the application experience of deep learning methods, and discusses the complementarity between deep learning methods and traditional methods in a dialectical way. The report starts from the foundation of modeling the seepage inversion problem, describes the main technical bottlenecks of the problem, and then explains how to use deep learning methods to solve the problems of traditional methods. The report will also discuss how traditional methods can be applied to deep learning frameworks. Finally, it will also briefly introduce the application of deep learning methods in seepage development optimization.

 

Personal Introduction:

Haibin Chang is an associate researcher at the School of Energy and Mining, China University of Mining and Technology (Beijing). He has long been engaged in research on seepage uncertainty analysis, data assimilation, development optimization and big data methods. He has published more than 20 SCI papers in SPE J, WRR, JCP, JGR, CMAME and other authoritative journals.

 

[Editor: Xiangyi Ma]