TITLE: Waveform Signal Analysis for System Performance Improvement

SPEAKER: Kamran Paynabar

ABSTRACT:

The rapid development of distributed sensing and computer technology has facilitated wide collection of various waveform signals during system operations. This has resulted in temporally and spatially data-rich environment that provides unprecedented opportunities for improving complex system operations in various applications including manufacturing, healthcare, emerging energy systems, etc. Meanwhile, it also raises new research challenges on data analysis and decision making due to the complex data structures of waveform signals, such as heterogeneous data dependency, high-dimensional and non-stationary characteristics. In this talk, two topics will be discussed in detail to present the essential need of multidisciplinary efforts in the development of new engineering-driven data analysis methods for effectively analyzing waveform signals. In the first topic, a new method of “hierarchical non-negative Garrote for group variable selection” will be presented, which is developed for informative sensor and feature selection among massive multistream sensing signals. The second topic is to discuss a new method for “characterization of nonlinear profiles variations using mixed-effect models and wavelets”, which is developed to effectively model both within-profile variations and between-profile variations for enhancing variation root cause diagnosis. Afterwards, a brief overview of my other research topics will also be presented.