利用動態數據做系統之監控偵錯與數據調諧(2/3)
Date Issued
2005
Date
2005
Author(s)
黃孝平
DOI
932214E002014
Abstract
The objective of this research is to utilize dynamic-PLSR1 (DPLS) to identifying dynamics of a process, and utilize PLS on modeling wavelet transformed input data for FDI. After modeling the dynamics of the process, filters are constructed and applied to the input data which have been pre-treated with Discrete Wavelet Transform (DWT). The DWT decomposes each input signal into three basic components -- trend, seasonal, and, stationary and random components. By this DWT pretreatment, it provides a capability to identify the source of sensor faults. The concepts and techniques are demonstrated using a simulated Wood and Berry system. It shows this presented method is effective in detecting and identifying either the faults caused by high frequency noises or by biases in relating outputs.
Subjects
subspace identification
impulse response sequence
closed loop
PLS
fault detection
wavelets
Publisher
臺北市：國立臺灣大學化學工程學系暨研究所
Type
report
File(s)![Thumbnail Image]()
Loading...
Name
932214E002014.pdf
Size
3.8 MB
Format
Adobe PDF
Checksum
(MD5):93ae031db05de8f61ac8dcdb9980cb2d