【学术报告】Data Mining for Prognostics(数据挖掘与预警技术)

发稿时间:2008-09-22浏览次数:3756

学 术 报 告

 

     :  Data Mining for Prognostics
(数据挖掘与预警技术)

报告人 :Chunsheng Yang

National Research Council (NRC), Canada

   间:2008925日 (周四)上午 1000

    点:蒙民伟楼404会议室

 

Abstract:

Data-driven prognostics is an emerging application of data mining to real-world problems such as system health management. It has been attracting much attention from researchers in the area of sensor, reliability, data mining and so on. The main task of prognostics is to predict the likelihood of a failure and estimate the remaining lifetime (or time to failure). Data-driven prognostics is to develop predictive models from large-scale historic operational and maintenance data using the techniques from data mining and machine learning. For this purpose, we have developed a KDD methodology to build the prognostic models which are able to predict the failure and estimate time to failure. In this talk, we will introduce the KDD methodology in details by addressing several challenging issues: model building, model evaluation and time to failure prediction.  We will also present some results obtained from a real-world application--prognostics of train wheel by demonstrating the deployed prognostic systems.

 

Dr. Chunsheng Yang is a Senior Research Officer with the Knowledge Discovery Group at the Institute for Information Technology of the National Research Council Canada (NRC-IIT). He received a Ph.D. from National Hiroshima University, Japan, 1995. He worked with Fujitsu Inc., Japan as a Senior Engineer from 1995 to 1998. His research interests include data mining for prognostics, case-based reasoning for diagnostics and multi-agent-based systems for group decision-making. He serves as a guest editor for the International Journal of Applied Intelligence, and is a Program Co-Chair for the 17th International Conference on Industry and Engineering Applications of Artificial Intelligence and Expert Systems (IEA/AIE), an Industrial Track Program Committee Member of ACM SIGKDD 07 and ACM SIGKDD 05.