1.汇报安排

题目：参加IEEE I²MTC 2016国际会议总结报告会

时间：2016年6月2日19:30-19:50

地点：科技园西五楼北楼会议室A228

报告人：博1216班——张晗

学号：4112001068

指导教师：陈雪峰 教授

2 .参加国际会议信息

·会议名称：IEEE I²MTC 2016(2016 IEEE International Instrumentation and Measurement Technology Conference)

·会议时间：23 – 26 MAY 2016

·会议地点：台北, 台湾

·会议简介：国际仪表与测量技术会议（International Instrumentation and Measurement Technology Conference，I²MTC）是IEEE仪表与测量学会（IEEE Instrumentation and Measurement Society）主办的仪器仪表领域顶级国际学术会议，会议主题包括：仪表与测量基础、传感器与换能器、物理量测量、测量系统、测量应用、信号与图像处理、监测与故障诊断和工业应用等。其分会SPECIAL SESSION: Advances in instruments and measurement专场讨论机械系统的健康监测与故障诊断，涉及图像处理技术、电机故障诊断分类、深度学习等。

·会议交流工作：

Presentation –Sparsity-aware tight frame learning for feature subspace recognition.

3.参会论文信息

·论文标题：Sparsity-aware tight frame learning for feature subspace recognition

·作者：张晗，陈雪峰，杜朝辉，马猛，张小丽

·摘要：It is a challenging problem to design excellent dictionaries to sparsely represent diverse fault information and simultaneously discriminate different fault sources. Therefore, this paper describes and analyzes a novel tight frame learning framework incorporating with an adaptive subspace recognition strategy for machine fault diagnosis. By introducing the tight frame constraint into the popular dictionary learning model, the proposed tight frame learning model could be formulated as a nonconvex optimization problem which can be solved by alternatively implementing hard thresholding operation and singular value decomposition. The noises are then effectively eliminated through transform sparse coding techniques. Meanwhile, structured signals are collaboratively represented through many different feature subspaces. Subsequently, the optimal feature subspace for every fault pattern is adaptively identified utilizing physics-driven indexes, and naturally different fault components are fully decoupled and reliably recovered. Extensive numerical experiments are firstly implemented to investigate the sparsifying capability of the learned tight frame as well as its comprehensive de-noising performance. Most importantly, the feasibility and superiority of the proposed framework is verified through performing multiple fault diagnosis of motor bearings. Compared with the state-of-the-art fault detection techniques, some important advantages have been observed: firstly, the tight frame filter banks are directly learned from the noisy observation itself and thus can be adaptively matched to various types of sparse modes, i.e., it is sparsity-aware. Secondly, each tight frame filter locally tailors to one type of signal pattern and all filters globally expand the whole space of structure information, which sufficiently eliminates the ill-posedness of traditional learning schemes for single feature subspace design. Thirdly, fast convolution operation has been exploited to enhance its computational efficiency and translation-invariance property. Fourthly, its joint subspace learning property leads to great superiority in high-fidelity waveform recovery and multiple feature separation.

欢迎有兴趣的同学届时听讲座。