System and method for controlling oscillatory instabilities in a device

Patent metadata

Field	Value
Publication number	US-10095247-B2
Application number	US-201415033897-A
Country	US
Kind code	B2
Filing date	Dec 3, 2014
Priority date	Dec 4, 2013
Publication date	Oct 9, 2018
Grant date	Oct 9, 2018

Abstract

A method for detecting onset of oscillatory instability in a device is described. The method includes obtaining a symbolic time series of a time series signal corresponding to a dynamic state variable of the device. The method further includes detecting the onset of oscillatory instability in the device based on the symbolic time series.

First claim

We claim:

1. A method for detecting an onset of oscillatory instability in a device, the method comprising:
performing analysis of a time series signal to detect onset of oscillatory instability, wherein said analysis is performed by considering that at said onset of oscillatory instability said time series signal exhibits specific patterns captured in a corresponding time window of said time series signal;
generating a symbolic time series from said time series signal corresponding to a dynamic state variable of said device;
identifying presence of oscillatory instability in said time series signal corresponding to said dynamic state variable by recognizing a change in said patterns;
constructing a Probabilistic Finite State Automata (PFSA) from said symbolic time series, wherein said PFSA is associated with a current state of said dynamic state variable;
computing an anomaly measure based on said PFSA and a PFSA corresponding to a reference state, wherein said anomaly measure indicates proximity of said PFSA to that of a reference PFSA, wherein said current state corresponds to a state for which said anomaly measure is being measured and a reference state is selected based on a dynamic state corresponding to said onset of oscillatory instability;
detecting said onset of oscillatory instability in response to determining that said anomaly measure is below an anomaly threshold; and
varying at least one parameter of said device in accordance to said detected onset of oscillatory instability to control an oscillatory instability.

2. The method of claim 1 , wherein detecting said onset of oscillatory instability is based on identifying whether at least one anomaly measure corresponding to the current state crosses said anomaly threshold.

3. The method of claim 1 , wherein said reference PFSA is constructed based on said reference state of said dynamic state variable representing one of a state corresponding to said oscillatory instability in said device and a state prior to said state corresponding to said oscillatory instability.

4. The method of claim 1 , wherein said anomaly measure is used to vary said at least one parameter.

5. The method of claim 1 , wherein obtaining said symbolic time series of said time series signal corresponding to said dynamic state variable of said device comprises:
obtaining said time series signal of said dynamic state variable, wherein said dynamic state variable is measured by at least one sensor in said device; and
converting said time series signal to said symbolic time series corresponding to said dynamic state variable.

6. The method of claim 1 , wherein varying said at least one parameter of said device in accordance to said detected onset of oscillatory instability to control said oscillatory instability comprises:
generating at least one control signal in accordance to said detected onset of oscillatory instability; and
varying said at least one parameter of said device based on said at least one control signal, wherein variation in said at least one parameter is dynamically performed on detecting said onset of oscillatory instability to control said oscillatory instability in said device.

7. A system for detecting onset of oscillatory instability in a device, the system comprising:
an instability detection unit configured to:
perform analysis of a time series signal to detect onset of oscillatory instability, wherein said analysis is performed by considering that at said onset of oscillatory instability said time series signal exhibits specific patterns captured in a corresponding time window of said time series signal;
generate a symbolic time series from said time series signal corresponding to a dynamic state variable of said device;
identify presence of oscillatory instability in said time series signal corresponding to said dynamic state variable by recognizing a change in said patterns;
construct a Probabilistic Finite State Automata (PFSA) from said symbolic time series, wherein said PFSA is associated with a current state of said dynamic state variable;
compute an anomaly measure based on said PFSA and a PFSA corresponding to a reference state, wherein said anomaly measure indicates proximity of said PFSA to that of a reference PFSA, wherein said current state corresponds to a state for which said anomaly measure is being measured and a reference state is selected based on a dynamic state corresponding to said onset of oscillatory instability;
detect said onset of oscillatory instability in response to determining that said anomaly measure is below an anomaly threshold; and
vary at least one parameter of said device in accordance to said detected onset of oscillatory instability to control an oscillatory instability.

8. The system of claim 7 , wherein said instability detection unit is configured to detect said onset of oscillatory instability is based on identifying whether at least one oscillation corresponding to said oscillatory instability crosses said anomaly threshold.

9. The system of claim 7 , wherein said instability detection unit is configured to construct said reference PFSA based on said reference state of said dynamic state variable representing one of a state corresponding to said oscillatory instability in said device and a state prior to said state corresponding to said oscillatory instability.

10. The system of claim 7 , wherein said anomaly measure is used to vary said at least one parameter.

11. The system of claim 7 , wherein said instability detection unit is configured to obtain said symbolic time series of said time series signal corresponding to said dynamic state variable of said device comprises:
obtain said time series signal of said dynamic state variable, wherein said dynamic state variable is measured by said at least one sensor on said device; and
convert said time series signal to said symbolic time series corresponding to said dynamic state variable.

12. The system of claim 7 , wherein vary said at least one parameter of said device in accordance to said detected onset of oscillatory instability to control said oscillatory instability comprises:
generate at least one control signal in accordance to said determined onset of oscillatory instability; and
vary said at least one parameter of said device based on said at least one control signal, wherein variation in said at least one parameter is dynamically performed prior to said detected onset of oscillatory instability to control said oscillatory instability in said device.

13. The system of claim 7 , wherein an Analog to Digital Converter (ADC) is configured to convert said time series signal measured by said sensor from an analog domain to digital domain and provide said converted time series signal to said instability detection unit.

Assignees

• Indian Institute Of Tech Madras

Inventors

• Unni Vishnu Rajasekharan
• Vinod Vineeth Nair
• Nair Sujith Raman Pillai Indusekharan
• Mukhopadhyay Achintya

Classifications

• G06N7/01

  Probabilistic graphical models, e.g. probabilistic networks · CPC title

• G05B23/0289

  Reconfiguration to prevent failure, e.g. usually as a reaction to incipient failure detection · CPC title

• F02D41/1498

  measuring engine roughness · CPC title

• G05B23/0227 Primary

  Qualitative history assessment, whereby the type of data acted upon, e.g. waveforms, images or patterns, is not relevant, e.g. rule based assessment; if-then decisions · CPC title

• G06N7/005

  Physics · mapped topic