Monitoring system using kernel regression modeling with pattern sequences

• US 8,660,980 B2
• Filed: 07/19/2011
• Issued: 02/25/2014
• Est. Priority Date: 07/19/2011
• Status: Active Grant

• Current Assignee: SmartSignal Corporation (Emerson Electric Company)
• Original Assignee: SmartSignal Corporation (Emerson Electric Company)
• Inventors: Herzog, James P.
• Primary Examiner(s): Gaffin, Jeffrey A
• Assistant Examiner(s): Misir, Dave
• Application Number:
  US13/186,128
• Publication Number:
  US 20130024166A1
• Time in Patent Office: 952 Days
• Field of Search: None
• US Class Current: 706/52
• CPC Class Codes: G05B 23/0254 based on a quantitative model, e.g. mathematical relationships between inputs and outputs; functions: observer, Kalman filter, residual calculation, Neural Networks
• CPC Class Codes: G06F 18/2414 Smoothing the distance, e.g. radial basis function networks [RBFN]
• CPC Class Codes: G06F 2218/12 Classification; Matching

Abstract

A monitoring system for determining the condition of an object includes an empirical model with reference data that indicates the normal operational state of the object, and input pattern arrays. Each input pattern array has a plurality of time-ordered input vectors, while each input vector has input values representing a plurality of parameters indicating the current condition of the object. The empirical model generates estimate values based on a calculation that uses an input pattern array and the reference data to determine a similarity measure between the input values and reference data. A differencing module compares the estimates values to the corresponding input values so that resulting values from the comparison can be used to determine the condition of the object.

Claims

1. A monitoring system for determining the condition of an object, comprising:

an empirical model operated by at least one processor and having reference data that indicates the normal operational state of the object, and input pattern arrays, each input pattern array having a plurality of time-ordered input vectors, each input vector having input values representing a plurality of parameters indicating the current condition of the object, the reference data being grouped in equal-sized and multi-dimensional training arrays, each of the training arrays being equal in size to a corresponding input multi-dimensional pattern array and each training array including a plurality of time-ordered reference vectors sequenced in time,

the empirical model being configured to generate estimate values based on a calculation that uses an input pattern array and the reference data to determine a similarity measure between the input values and the reference data; and

a differencing module comparing the estimate values to the corresponding input values so that resulting values from the comparison can be used to determine the condition of the object.

2. The system of claim 1, wherein at least one of the input vectors is in multiple input pattern arrays.

3. The system of claim 1, wherein each input pattern array represents a time period including a plurality of time points along a sequence of the input vectors representing the operational state of the object.

4. The system of claim 3, wherein each input pattern array represents a time period that is different than, but overlaps, the time period of at least one of the other input pattern arrays.

5. The system of claim 3, wherein each input pattern array has a current primary vector from the sequence of input vectors and representing the most recent time point within the input pattern array.

6. The system of claim 5 wherein the input vectors are selected for each input pattern array by moving a look-back window along the sequence, selecting the current primary vector, and selecting one or more older vectors that follow the primary vector along the sequence and fit within the window.

7. The system according to claim 6 wherein the older vectors are consecutive along the sequence.

8. The system according to claim 6 wherein the older vectors are not consecutive along the sequence.

9. The system according to claim 1 wherein time intervals between the time points represented by the input vectors within an input pattern array are uniform.

10. The system according to claim 1 wherein time intervals between the time points represent by the input vectors within an input pattern array are unequal.

11. The system according to claim 1 wherein time intervals between time points represented by the input vectors within an input pattern array are shorter between input vectors that are more recent than the time intervals between later input vectors within the same input pattern array.

12. The system according to claim 1 wherein the total duration represented by the input pattern arrays varies.

13. The system according to claim 1 wherein the calculation is based on an equation that comprises:

14. The system according to claim 1 wherein the calculation is based on an equation that comprises:

15. The system of claim 1 wherein the similarity measure is a value on a bounded range that indicates an input value and reference value are identical at one end of the range and have no similarity at another end of the range.

16. The system of claim 1, wherein the reference data comprises at least one learned sequential pattern matrix, each learned sequential pattern matrix having a plurality of reference vectors, each reference vector having reference values representing a plurality of parameters.

17. The system of claim 16 wherein the reference data forms a three-dimensional collection of the learned sequential pattern matrices.

18. The system of claim 17 wherein the three-dimensional collection is defined by three dimensions representing the parameters along a first dimension, the learned sequential pattern matrices along a second dimension, and time represented by the reference vectors in a third dimension.

19. The system of claim 17 wherein each input pattern array is compared to each learned sequential pattern matrix in the three-dimensional collection to determine a similarity measure for each learned sequential pattern matrix.

20. The system of claim 17 herein each learned sequential pattern matrix represents a time period including time points represented by the reference vectors, wherein the time period is different than, but overlaps, the time period of at least some of the other learned sequential pattern matrices in the three-dimensional collection.

21. The system of claim 17 wherein the learned sequential pattern matrices have the same number of vectors as the input pattern arrays.

22. The system of claim 17 wherein the learned sequential pattern matrices represent the same time points as the input pattern arrays.

23. The system of claim 17 wherein the similarity between one of the learned sequential pattern matrices and one of the input pattern arrays is based on the similarity between corresponding reference and input values that represent the same vector position relative to the position of the other vectors represented within their respective array.

24. The system of claim 17 wherein the similarity between one of the learned sequential pattern matrices and one of the input pattern arrays comprises calculating an average scalar similarity value for each parameter represented by the arrays.

25. The system of claim 17 wherein at least some of the input pattern arrays represent a time period that is different than the time period represented by at least one of the learned sequential pattern matrices.

26. The system according to claim 17 wherein the time period represented by the learned sequential pattern matrices varies.

27. The system according to claim 1 wherein the empirical model is configured to generate the estimate values in the form of an estimate matrix having a plurality of estimate vectors.

28. The system of claim 27 wherein the estimate vectors in each estimate matrix corresponds to the time points represented by the input vectors in the input pattern arrays.

29. The system of claim 27 wherein less than all of the estimate values within the estimate matrix are used to determine the condition of the object.

30. The system of claim 27 wherein the differencing module only uses the estimate values that represent the most recent time point within each estimate matrix to determine the condition of the object.

31. The system of claim 27 comprising providing a single estimate vector for each time point represented across multiple estimate matrices to determine the condition of the object.

32. The system of claim 31 wherein the single estimate vector is provided by using at least one of:

selecting the estimate vector with one or more estimate values that has a maximum, minimum, or median value relative to other estimate vectors at the time point across multiple estimate matrices,

calculating an average, a weighted average, or weighted norm of the estimate values at the time point across multiple estimate matrices, and

selecting an estimate vector with the greatest similarity to a corresponding input vector relative to other estimate vectors at the time point across multiple estimate matrices.

33. The system of claim 27 wherein the empirical model provides an estimate vector to determine the condition of the object by using at least one of:

an estimate vector within each estimate matrix that has a maximum, minimum, or median estimate value,

a weighted average or weighted norm of the estimate values in the estimate matrix to form a single estimate for each matrix, and

the estimate vector within the estimate matrix with the greatest similarity to a respective corresponding input vector.

34. The system of claim 1 wherein the empirical model is configured to use the similarity measures as weights in a calculation with a three-dimensional collection of learned sequential pattern matrices that represents parameters that are not represented in the input pattern arrays.

35. The system of claim 34 wherein the reference data used to generate the similarity measures do not represent all of the parameters represented by the three-dimensional collection.

36. The system of claim 34 wherein the three-dimensional collection represents only parameters that are not represented by the input pattern arrays.

37. The system of claim 35 wherein the three-dimensional collection represents both parameters that are not represented by the input pattern arrays and parameters that are represented by the input pattern arrays.

38. The system of claim 1 comprising a localization module to predetermine which reference data is to be used to generate the estimate values for each input pattern array.

39. The system of claim 1 comprising an adaption module configured to add the input pattern arrays to the reference data when the estimate values indicate the corresponding input pattern array represents normal operation.

40. A monitoring system for determining the condition of an object, comprising:

an empirical model operated by at least one processor and comprising reference data that indicates the normal operational state of the object and in the form of a plurality of learned, equal-size, time sequential and multi-dimensional pattern matrices, each learned, equal-size, time sequential and multi-dimensional pattern matrix having a plurality of reference vectors, each reference vector having reference values representing a plurality of parameters, each of the plurality of learned, equal-size, time sequential and multi-dimensional pattern matrices being equal in size to a corresponding input multi-dimensional pattern array, and

input data representing a plurality of parameters indicating the current condition of the object, the input data being a plurality of input multi-dimensional pattern arrays,

wherein the empirical model is configured to generate estimate values based on a calculation that uses the input data and the learned, equal-size, time sequential and multi-dimensional pattern matrices to determine a similarity measure between the input data and reference values in the plurality of learned, equal-size, time sequential, and multi-dimensional pattern matrices; and

a differencing module comparing the estimate values to corresponding input values so that resulting values from the comparison can be used to determine the condition of the object.

41. A monitoring system for determining the condition of an object, comprising:

an empirical model operated by at least one processor and comprising reference data that indicates the normal operational state of the object, the reference data being grouped in equal-sized and multi-dimensional training arrays, each of the training arrays being equal in size to a corresponding input multi-dimensional pattern array and each training array including a plurality of time-ordered reference vectors sequenced in time, and

input data representing a plurality of parameters indicating the current condition of the object, the input data being a plurality of input multi-dimensional pattern arrays,

wherein the empirical model is configured to generate estimate values based on a calculation that uses both the input data and the reference data to determine similarity measures between the input data and the reference data, wherein the estimate values are generated in the form of an estimate matrix having a plurality of time-ordered estimate vectors, each estimate vector having estimate values representing multiple parameters; and

a differencing module comparing at least one estimate vector for each time period represented by the estimate matrix to the input data so that the resulting values from the comparison can be used to determine the condition of the object.