ASSEMBLY OF INTERCONNECTED COMPUTING ELEMENTS, METHOD FOR COMPUTER-ASSISTED DETERMINATION OF A DYNAMICS WHICH IS THE BASE OF A DYNAMIC PROCESS, AND METHOD FOR COMPUTER-ASSISTED TRAINING OF AN ASSEMBLY OF INTERCONNECTED ELEMENTS

• US 6,493,691 B1
• Filed: 04/07/2000
• Issued: 12/10/2002
• Est. Priority Date: 08/07/1998
• Status: Expired due to Fees

• Current Assignee: Siemens AG
• Original Assignee: Siemens AG
• Inventors: Zimmermann, Hans-Georg, Neuneier, Ralf
• Primary Examiner(s): Starks, Jr., Wilbert L.
• Application Number:
  US09/529,195
• Time in Patent Office: 977 Days
• Field of Search: 706/45, 706/23, 187/247, 700/51
• US Class Current: 706/45
• CPC Class Codes: G06N 3/049 Temporal neural networks, e.g. delay elements, oscillating neurons or pulsed inputs

Abstract

An input signal is transformed into a predetermined space. Transformation computer elements are connected to one another such that transformed signals can be taken at the transformation computer elements, whereby at least three transformed signals relate to respectively successive points in time. Composite computer elements are respectively connected to two transformation computer elements. Further, a first output computer element is provided at which an output signal that describes a system status at a point in time can be taken. The first output computer element is connected to the transformation computer elements. Further, a second output computer element is provided that is connected to the composite computer elements and given whose employment a predetermined condition can be taken into consideration when training the arrangement.

Claims

1. An arrangement of computer elements connected to one another, comprising:

a) input computer elements to which time row values that respectively describe a status of a system at a point in time are supplied;

b) transformation computer elements for transformation of the time row values into a predetermined space, said transformation computer elements being connected to the input computer elements;

c) connections between the transformation computer elements such that transformed signals are taken at the transformation computer elements, at least three of the transformed signals relate to respectively successive points in time;

d) composite computer elements that are respectively connected to two of said transformation computer elements;

e) a first output computer element that is connected to the transformation computer elements, an output signal being taken at the first output computer element; and

f) a second output computer element that is connected to the composite computer elements and with whose employment a predetermined condition in the prescribed space is taken into consideration in a training of the arrangement.

2. An arrangement according to claim 1, wherein the transformation computer elements are grouped into a first hidden layer and into a second hidden layer, whereby at least a part of the transformation computer elements of the first hidden layer and of the transformation computer elements of the second hidden layer are connected to one another.

3. An arrangement according to claim 2, wherein the input computer elements are allocated to an input layer that includes sub-input layers, and the transformation computer elements of the first hidden layer are grouped into hidden sub-layers such that the transformation computer elements of each of said hidden sub-layers are respectively connected to input computer elements of respective ones of said sub-input layers.

4. An arrangement according to claim 1, wherein at least a part of the computer elements are artificial neurons.

5. An arrangement according to claim 1, wherein the transformation computer elements are connected to one another such that transformed signals is taken at the transformation computer elements, at least four transformed signals relating to respectively successive points in time.

6. An arrangement according to claim 1, wherein the input computer elements are allocated to an input layer such that the input layer includes a plurality of sub-input layers, each of said sub-input layers supplied with at least one input signal that describes a system status at a respective point in time.

7. An arrangement according to claim 1, wherein at least a part of the connections between the computer elements are variably fashioned.

8. An arrangement according to claim 1, wherein the connections between the composite computer elements and the second output computer element are invariant.

9. An arrangement according to claim 1, wherein at least parts of the connections exhibit mutually same weightings.

10. An arrangement according to claim 1, wherein said arrangement is configured such that the predetermined condition is a predetermined smoothness of a dynamic process in space.

11. An arrangement according to claim 1, wherein said arrangement is configured such that the predetermined condition is a Lyapunov zero condition.

12. An arrangement according to claim 1, wherein said arrangement is utilized for determining a dynamic of a dynamic process.

13. An arrangement according to claim 12, further comprising:

a measuring arrangement for acquiring physical signals with which the dynamic process is described.

14. An arrangement according to claim 12, wherein said arrangement utilized for determining the dynamic of a dynamic process in a reactor.

15. An arrangement according to claim 12, wherein said arrangement is utilized for determining the dynamic of a dynamic process in a chemical reactor.

16. An arrangement according to claim 12, wherein said arrangement is utilized for determining the dynamic of a dynamic process for traffic modeling.

17. An method for computer-supported determination of a dynamic that forms a basis of a dynamic process, comprising the following steps:

a) describing the dynamic process by a time row having time row values in a first status space, at least one first time row value describing a status of the dynamic process at a first point in time and a second time row value describing a status of the dynamic process at a second point in time;

b) transforming the first time row value into a second status space;

c) subjecting the first time row value in the second status space to an imaging onto a second time row value in the second status space;

d) transforming the second time row value in the second status space back into the first status space;

e) performing the transformation and the imaging such that the dynamic process described by the time row values in the second status space meets a predetermined condition in the second status space; and

f) determining the dynamic of the dynamic processes from the time row values in the second status space.

18. A method according to claim 17, wherein the time row values are vectors of a prescribable dimension.

19. A method according to claim 17, further the time row values are determined from physical signals.

20. A method according to claim 19, wherein the time row values describe a dynamic process in a reactor.

21. A method according to claim 20, wherein the time row values describe a dynamic process in a chemical reactor.

22. A method according to claim 19, wherein the time row values describe a dynamic process for traffic modeling.

23. A method according to claim 17, wherein the predetermined condition is predetermined smoothness of the dynamic process in the space.

24. A method according to claim 17, wherein the predetermined condition is a Lyapunov zero condition.

25. A arrangement of computer elements connected to one another for computer-supported determination of a dynamic that forms a basis of a dynamic process, comprising:

a) input computer elements supplied with an input signal that describes a system status at a point in time;

b) transformation computer elements for transformation of the input signal into a predetermined space connected to the input computer elements;

c) connections between the transformation computer elements such that transformed signals are taken at the transformation computer elements, at least three of the transformed signals relating to respectively successive points in time;

d) composite computer elements connected to respectively two of said transformation computer elements;

e) a first output computer element connected to the transformation computer elements, said first output computer element producing an output signal that describes a system status at a point in time;

f) a second output computer element connected to the composite computer elements, a predetermined condition in the prescribed space being taken into consideration in a training of the arrangement upon employment of the second output computer element;

wherein the arrangement is supplied with the input signal, and the arrangement determines the first output signal from which the dynamic is determined.

26. A method for computer-supported training of an arrangement of computer elements connected to one another, comprising the steps of:

a) supplying input computer elements with an input signal that describes a system status at a point in time;

b) connecting transformation computer elements for transformation of the input signal into a predetermined space to the input computer element;

c) connecting the transformation computer elements to one another such that transformed signals are taken at the transformation computer elements, at least three of said transformed signals relating to respectively successive points in time;

d) connecting composite computer elements to respectively two transformation computer elements;

e) connecting a first output computer element to the transformation computer elements, an output signal that describes a system status at a point in time is taken at the first output computer element;

f) connecting a second output computer element to the composite computer elements, a predetermined condition in the prescribed space is taken into consideration in a training of the arrangement upon employment of the second output computer element;

training the arrangement taking the predetermined condition into consideration and using predetermined training data.