Skip to Main Content
 

Global Search Box

 
 
 
 

Files

File List


12840.pdf (11.94 MB)

ETD Abstract Container

Abstract Header

A Probabilistic Approach for Prognostics of Complex Rotary Machinery Systems

Zhao, Wenyu
http://rave.ohiolink.edu/etdc/view?acc_num=ucin1423581651

Abstract Details

2015, PhD, University of Cincinnati, Engineering and Applied Science: Mechanical Engineering.
The growing demand for asset reliability and operation optimization has motivated the development of Prognostics and Health Management (PHM) methodologies and techniques, which facilitates data analytics and enables diagnosis and prognosis of machinery asset. Data-driven PHM approaches mine available data to derive condition indicators (CI) and subsequently health indicators (HI), detect faults as data anomalies, identify faults based on data classification, and predict faults based on trends and trajectories in data. Machine learning and artificial intelligence aid such data science by providing techniques that define clustering models, generate classifiers and forecast future states, as well as define thresholds and boundaries that aid decision-making outcomes. Model-based PHM approaches establish physical models that are derived from first-principal knowledge about the failure mechanism of interest, and simulate fault signatures with high fidelity computation based on the model. There is a growing trend to merge the complementary prognostics methodologies, motivated by the rapid development of big data infrastructure. The challenges associated with learning from large amount of data with high dimension involve comprehending the interdependencies between data variables and the complex physical systems where data is collected from, as well as managing the uncertainty that is underlying with the data due to modeling uncertainty and measurement noise. In this dissertation, a Bayesian theory based modeling and reasoning approach is developed, for learning variable dependency in the form of conditional probability. Two case studies are analyzed and discussed: one addresses the issue of learning the dependency between system operating regime and its performance outcome in a data-driven environment; the other focuses on coupling physical model and sensory measurement with an adaptive filtering approach, which is capable of representing a non-linear system over discrete time slices, sequentially filtering state estimation and predicting remaining useful life of the system. Both case studies are validated with data sets provided by industry partners, and are benchmarked with previously developed and recognized techniques. Finally, conclusions of this research are discussed and prospect research direction is proposed as future work for interested researchers.
Jay Lee, Ph.D. (Committee Chair)
Raj Bhatnagar, Ph.D. (Committee Member)
Mark Schulz, Ph.D. (Committee Member)
David Thompson, Ph.D. (Committee Member)
141 p.
Mechanics
Prognostics; PHM; Bayesian Network; Wind Turbine; Jet Engine; Prediction

Recommended Citations

Citations

  • Zhao, W. (2015). A Probabilistic Approach for Prognostics of Complex Rotary Machinery Systems [Doctoral dissertation, University of Cincinnati]. OhioLINK Electronic Theses and Dissertations Center. http://rave.ohiolink.edu/etdc/view?acc_num=ucin1423581651

    APA Style (7th edition)

  • Zhao, Wenyu. A Probabilistic Approach for Prognostics of Complex Rotary Machinery Systems. 2015. University of Cincinnati, Doctoral dissertation. OhioLINK Electronic Theses and Dissertations Center, http://rave.ohiolink.edu/etdc/view?acc_num=ucin1423581651.

    MLA Style (8th edition)

  • Zhao, Wenyu. "A Probabilistic Approach for Prognostics of Complex Rotary Machinery Systems." Doctoral dissertation, University of Cincinnati, 2015. http://rave.ohiolink.edu/etdc/view?acc_num=ucin1423581651

    Chicago Manual of Style (17th edition)

ucin1423581651
1,361
© 2015, all rights reserved.
This open access ETD is published by University of Cincinnati and OhioLINK.