Arrangement for predicting an abnormality of a system and for carrying out an action which counteracts the abnormality

• US 7,006,866 B1
• Filed: 11/05/1998
• Issued: 02/28/2006
• Est. Priority Date: 11/07/1997
• Status: Expired due to Fees

• Current Assignee: Siemens AG
• Original Assignee: Siemens AG
• Inventors: Deco, Gustavo, Dubé, Louis-J.
• Primary Examiner(s): Pezzuto, Robert E.
• Assistant Examiner(s): Oropeza, Frances P.
• Application Number:
  US09/530,983
• Time in Patent Office: 2,672 Days
• Field of Search: 128/925, 600/14, 600/374, 600/377, 600/378, 600/382, 600/383, 600/509, 600/513, 600/515, 600/518, 600/519, 600/544, 607/4, 607/5, 607/7, 607/9, 607/14, 607/17, 607/62, 607/119, 700/38, 700/40, 700/45, 706/20, 706/21, 706/23
• US Class Current: 607/5
• CPC Class Codes: A61N 1/36025 for treating a mental or cerebral condition
• CPC Class Codes: G05B 23/024 Quantitative history assessment, e.g. mathematical relationships between available data; Functions therefor; Principal component analysis [PCA]; Partial least square [PLS]; Statistical classifiers, e.g. Bayesian networks, linear regression or correlation analysis; Neural networks
• CPC Class Codes: G05B 23/0289 Reconfiguration to prevent failure, e.g. usually as a reaction to incipient failure detection

Abstract

An arrangement and method are presented that enable a prediction of an abnormality and implement a suitable action opposing the abnormality. An information flow underlying a dynamic system is interpreted and a prediction quantity that comprises the abnormality as characterizing quantity of the dynamic system is determined from it. A neural network is trained with measured data of the system. After the training, the abnormality can be indicated on the basis of the prediction quantity before it occurs and the occurrence can be opposed with suitable measures.

Claims

1. An arrangement for predicting an abnormality of a dynamic system and for implementing an action opposing the abnormality using a continuous information flow that describes a development of a predictability of several future system states, comprising:

a) a measured data pick-up that registers comparison measured data of said system and test measured data of said system;

b) a processor unit, having a neural network that models said system, said processor unit

(1) training said neural network using said comparison measured data;

(2) determining a comparison information flow that describes a comparison dynamic of said system using said trained neural network;

(3) determining a test information flow that describes a test dynamic of said system using said test measured data;

(4) using said comparison information flow and said test information flow, predicting said abnormality as established when said comparison information flow differs significantly from said test information flow, and predicting said abnormality as not established when said comparison information flow does not significantly differ from said test information flow;

(5) when said abnormality of the system has been predicted as established, then implementing said action; and

c) an actuator that implements said action,

wherein the information flow describes a development of a predictability of plural future system states.

2. An arrangement according to claim 1, wherein said processor unit endlessly loops from said determining a comparison information flow to said implementing said action.

3. Arrangement according to claim 1, wherein said abnormality is predicted as established when said test information flow is significantly smaller than said comparison information flow.

4. An arrangement according to claim 3, wherein said action comprises exciting said system with a chaotic signal.

5. An arrangement according to claim 4, wherein said action comprises supplying noise to said system.

6. An arrangement according to claim 5, wherein said noise is supplied by a corresponding electrical field.

7. An arrangement according to claim 6, wherein said electrical field is supplied by at least one electrode.

8. An arrangement according to claim 5, wherein said noise is supplied by a corresponding magnetic field.

9. An arrangement according to claim 8, wherein said magnetic field is supplied by at least one electrode.

10. An arrangement according to claim 1, wherein said abnormality is predicted as established when said test information flow is significantly greater than said comparison information flow.

11. An arrangement according to claim 10, wherein said action comprises exciting said system with a regular signal.

12. An arrangement according to claim 11, wherein said regular signal is supplied by an electrical field.

13. An arrangement according to claim 11, wherein said electrical field is supplied by at least one electrode.

14. An arrangement according to claim 11, wherein said regular signal is supplied by a magnetic field.

15. An arrangement according to claim 14, wherein said magnetic field is supplied to said system by at least one electrode.

16. A method for predicting an abnormality of a dynamic system and for implementing an action opposing the abnormality using a continuous information flow that describes a development of a predictability of several future system states, comprising:

a) measuring comparison measured data of said system and test measured data of said system;

b) determining a neural network that models said system using said comparison measured data;

c) determining a comparison information flow that describes a comparison dynamic of said system using said neural network;

d) determining a test information flow that describes a test dynamic of said system using said test measured data;

e) comparing said comparison information flow to said test information flow using said comparison information flow and said test information flow;

f) determining said abnormality to be predicted as established when said comparison information flow differs significantly from said test information flow;

g) determining said abnormality to be predicted as not established when said comparison information flows does not significantly differ from said test information flow; and

h) implementing said action when said abnormality of said system has been predicted as established,

wherein the information flow describes a development of a predictability of plural future system states.

17. A method for predicting an abnormality of a dynamic system using a continuous information flow that describes a development of a predictability of several future system states, comprising the steps of:

a) measuring comparison measured data of said system and test measured data of said system;

b) determining a comparison information flow that describes a comparison dynamic of said system using said comparison measured data;

c) determining a test information flow that describes a test dynamic of said system using said test measured data;

d) comparing said comparison information flow to said test information flow using said comparison information flow and said test information flow;

e) determining said abnormality to be predicted as established when said comparison information flow differs significantly from said test information flow; and

f) determining said abnormality to be predicted as not established when said comparison information flow does not significantly differ from said test information flow,

wherein the information flow describes a development of a predictability of plural future system states.

18. A method for predicting an abnormality of a dynamic system and for implementing a procedure in response to the abnormality, comprising:

training a neural network to learn the dynamics of a system;

evaluating a continuous information flow received from the system;

predicting an abnormality when the information flow differs significantly from normal state information as determined by the neural network; and

implementing a procedure, if an abnormality is predicted, to prevent or treat the abnormality,

wherein the information flow describes a development of a predictability of plural future system states.