Abstract:

The rapid advancements of sensor technology, communication networks, and computing power have resulted in temporally and spatially dense data-rich environments, which provide unprecedented opportunities for improving operations in complex industrial systems. Meanwhile, it also raises new research challenges on data analysis and decision making, such as heterogeneous data formats, high-dimensional and big data structures, inherent complexity of the target systems, and potential lack of complete a priori knowledge, etc. To address those challenges, Dr. Kaibo Liu's research focuses on System Informatics and Data Analytics in Industrial Engineering, with emphasis on the data fusion for process modeling, monitoring, diagnosis and prognostics.

In this talk, two topics will be discussed in detail to present the need of developing multidisciplinary data analysis methods for effective prognostic analysis and system monitoring. The first topic is to describe a generic real-time data-level fusion methodology, which is capable of integrating multiple sensor signals to enhance degradation modeling and prognostic analysis. This methodology is tested and validated through a degradation dataset of an aircraft gas turbine engine. In the second topic, a systematic, simple and computationally ef?cient sampling algorithm, which is named as "Top-? based Adaptive Sampling (TRAS)"is proposed for real-time detection of the occurrences of solar flares in a large video stream generated by NASA satellites. If time is allowed, an overview of other research topics for manufacturing and service systems will also be provided.