Integration of time series modeling and wavelet transform for monitoring nuclear plant sensors

Belle R. Upadhyaya*, Chaitanya Mehta, Duygu Bayram

*Corresponding author for this work

Research output: Contribution to journalArticlepeer-review

11 Citations (Scopus)

Abstract

Dynamic fluctuations of nuclear plant sensors contain information about their response characteristics and bandwidth features. The random fluctuations can be characterized by using auto-regression (AR) time-series models. These discrete-time models are then utilized to estimate time-domain and frequency-domain signatures. Prior to developing these models, the sensor measurements are enhanced by filtering both low-frequency and high-frequency components using wavelet transforms. The use of wavelet transform for signal conditioning results in minimum distortion of the signal bandwidth, and thus provides an effective approach for data pre-processing. This integrated approach is applied to plant data from a pressurized water reactor (PWR). Univariate AR models were established for several pressure transmitter data, and used to estimate response time parameters of sensors and their frequency spectra. The results of this integrated approach demonstrate the improvement in the sensor signature estimation compared to the direct use of plant measurements.

Original languageEnglish
Article number2341035
Pages (from-to)2628-2635
Number of pages8
JournalIEEE Transactions on Nuclear Science
Volume61
Issue number5
DOIs
Publication statusPublished - 1 Oct 2014

Bibliographical note

Publisher Copyright:
© 2014 IEEE.

Funding

FundersFunder number
U.S. Department of Energy

    Keywords

    • Auto-regression models
    • Response time
    • Sensor monitoring
    • Signal enhancement
    • Wavelet transform

    Fingerprint

    Dive into the research topics of 'Integration of time series modeling and wavelet transform for monitoring nuclear plant sensors'. Together they form a unique fingerprint.

    Cite this