Statistical properties of signal entropy for use in detecting changes in time series data

Detecting changes in an underlying time series model for a system is an important task in many different fields, including econometrics, geophysics and process control. Specifically, in process control, detecting model changes is often the first step for fault detection, plant‐model mismatch assessment and data quality assessment for system identification. Signal entropy, which basically measures the amount of disorder in a given signal, can, not only segment a time series, but can also determine which regions have similar underlying models. Thus, the changes between the input and output signals can be used to determine when model is no longer an accurate representation of the system by comparing the current differential entropy against the historical differential entropy. This paper presents the statistical properties of signal entropy for discrete time systems. An example of the general results is provided by determining the entropy characteristics for first‐order systems driven by white noise. As well, a change detection index is proposed to assess changes in the time series model, which is tested on an experimental system. Copyright © 2013 John Wiley & Sons, Ltd.

[1]  Mudassir Rashid,et al.  Multiway independent component analysis mixture model and mutual information based fault detection and diagnosis approach of multiphase batch processes , 2013 .

[2]  Fredrik Gustafsson,et al.  Adaptive filtering and change detection , 2000 .

[3]  Michèle Basseville,et al.  On-board Component Fault Detection and Isolation Using the Statistical Local Approach , 1998, Autom..

[4]  Michèle Basseville,et al.  The asymptotic local approach to change detection and model validation , 1987 .

[5]  Biao Huang,et al.  Closed-loop identification with routine operating data: Effect of time delay and sampling time ☆ , 2011 .

[6]  Eamonn J. Keogh,et al.  Segmenting Time Series: A Survey and Novel Approach , 2002 .

[7]  S. Joe Qin,et al.  Discriminating between disturbance and process model mismatch in model predictive control , 2009 .

[8]  David Cuesta-Frau,et al.  Automatic segmentation of long-term ECG signals corrupted with broadband noise based on sample entropy , 2010, Comput. Methods Programs Biomed..

[9]  Michèle Basseville,et al.  Detecting changes in signals and systems - A survey , 1988, Autom..

[10]  S. Joe Qin,et al.  Statistical MIMO controller performance monitoring. Part II: Performance diagnosis , 2008 .

[11]  A. Denis,et al.  Using the Entropy of Curves to Segment a Time or Spatial Series , 2002 .

[12]  Biao Huang,et al.  On-line closed-loop model validation and detection of abrupt parameter changes , 2001 .

[13]  Alan Liu,et al.  Pattern discovery of fuzzy time series for financial prediction , 2006, IEEE Transactions on Knowledge and Data Engineering.

[14]  Michèle Basseville,et al.  Fault Detection and Isolation in Nonlinear Dynamic Systems: A Combined Input-Output and Local Approach , 1998, Autom..

[15]  Sang Joon Kim,et al.  A Mathematical Theory of Communication , 2006 .

[16]  Sirish L. Shah,et al.  Quantifying the impact of model-plant mismatch on controller performance , 2010 .

[17]  Padhraic Smyth,et al.  Segmental Semi-Markov Models for Endpoint Detection in Plasma Etching , 2000 .

[18]  John H. L. Hansen,et al.  Discrete-Time Processing of Speech Signals , 1993 .