Monitoring system based on a support vector data description

Patent metadata

Field	Value
Publication number	US-9639809-B1
Application number	US-201615390236-A
Country	US
Kind code	B1
Filing date	Dec 23, 2016
Priority date	Feb 10, 2016
Publication date	May 2, 2017
Grant date	May 2, 2017

Abstract

A computing device identifies outliers. Support vectors, Lagrange constants, a center threshold value, an upper control limit value, and a lower control limit value are received that define a normal operating condition of a system. The center threshold value, the upper control limit value, and the lower control limit value are computed from the vectors and the Lagrange constants. A first plurality of observation vectors is received for a predefined window length. A window threshold value and a window center vector are computed. A window distance value is computed between the window center vector and the support vectors. Based on comparisons between the computed values and the received values, the first plurality of observation vectors is identified as an outlier relative to the normal operating condition of the system. When the first plurality of observation vectors are identified as the outlier, an alert is output.

First claim

What is claimed is:

1. A non-transitory computer-readable medium having stored thereon computer-readable instructions that when executed by a computing device cause the computing device to:
receive a set of support vectors, a Lagrange constant for each support vector of the set of support vectors, a center threshold value, an upper control limit value, and a lower control limit value, wherein the set of support vectors define a normal operating condition of a system, wherein the center threshold value, the upper control limit value, and the lower control limit value are computed from the set of support vectors and the plurality of Lagrange constants;
receive a first plurality of observation vectors, wherein a number of the first plurality of observation vectors is a predefined window length;
compute a window threshold value and a window center vector by solving an objective function defined for a support vector data description (SVDD) model using the received first plurality of observation vectors;
compute a window distance value between the window center vector and the set of support vectors using the set of support vectors and the plurality of Lagrange constants;
when the computed distance value is greater than the center threshold value, identify the first plurality of observation vectors as an outlier relative to the normal operating condition of the system;
when the computed window threshold value is greater than the upper control limit value, identify the first plurality of observation vectors as the outlier relative to the normal operating condition of the system;
when the computed window threshold value is less than the lower control limit value, identify the first plurality of observation vectors as the outlier relative to the normal operating condition of the system; and
when the first plurality of observation vectors are identified as the outlier, output an alert.

2. The non-transitory computer-readable medium of claim 1 , wherein outputting the alert comprises sending a message to a second computing device, wherein the message indicates that a system fault has occurred or that a system state has changed.

3. The non-transitory computer-readable medium of claim 2 , wherein the predefined window length is associated with a start time and an end time, wherein outputting the alert comprises presenting an indicator of the start time and the end time.

4. The non-transitory computer-readable medium of claim 1 , wherein outputting the alert comprises sending the first plurality of observation vectors to a second computing device.

5. The non-transitory computer-readable medium of claim 1 , wherein outputting the alert comprises presenting the first plurality of observation vectors on a display.

6. The non-transitory computer-readable medium of claim 1 , wherein outputting the alert comprises presenting the computed distance value or the computed window threshold value on a display.

7. The non-transitory computer-readable medium of claim 1 , wherein the distance value is computed using dist 2 (z)=K(z,z)−2Σ i=1 N α i K(x i ,z)+Σ i=1 N Σ j=1 N α i α j K(x i ,x j ), where z is the window center vector, x i and x j are a support vector of the set of support vectors, α i and α j are the Lagrange constant for the associated support vector, N is a number of support vectors of the set of support vectors, and K( ) is a kernel function.

8. The non-transitory computer-readable medium of claim 1 , wherein the distance value is computed using dist 2 (z)=(z·z)−2Σ i=1 N α i (x i ·z)+Σ i=1 N Σ j=1 N α i α j (x i ·x j ), where z is the window center vector, x i and x j are a support vector of the set of support vectors, α i and α j are the Lagrange constant for the associated support vector, and N is a number of support vectors of the set of support vectors.

9. The non-transitory computer-readable medium of claim 1 , wherein the set of support vectors are computed from a plurality of windows of data, wherein each window of data of the plurality of windows of data includes a second plurality of observation vectors.

10. The non-transitory computer-readable medium of claim 9 , wherein the upper control limit value is computed using U DCL = R 2 +CL F ×σ R 2 , where R 2 is a mean of a plurality of threshold values, CL F is a control limit factor, and σ R 2 is a standard deviation of the plurality of threshold values, wherein each threshold value of the plurality of threshold values is computed for each window of data.

11. The non-transitory computer-readable medium of claim 10 , wherein the lower control limit value is computed using L DCL = R 2 −CL F ×σ R 2 .

12. The non-transitory computer-readable medium of claim 10 , wherein each threshold value for each window of data is computed by solving the objective function defined for the SVDD model using the second plurality of observation vectors for each window of data to define a second set of support vectors for each window of data.

13. The non-transitory computer-readable medium of claim 12 , wherein each threshold value is computed using R 2 =K(x k ,x k )−2Σ i=1 N α i K(x i ,x k )+Σ i=1 N Σ j=1 N α i α j K(x i ,x j ), where x k is any support vector of the second set of support vectors that have 0<α i <C, x i and x j are a support vector of the second set of support vectors, α i and α j are the Lagrange constant of the associated support vector, N is a number of support vectors included in the second set of support vectors, and C is a penalty constant.

14. The non-transitory computer-readable medium of claim 12 , wherein each threshold value is computed using R 2 =x k ·x k ·2Σ i=1 N α i (x i ·x k )+Σ i=1 N Σ j=1 N α i α j (x i ·x j ), where x k is any support vector of the second set of support vectors that have 0<α i <C, x i and x j are a support vector of the second set of support vectors, α i and α j are the Lagrange constant of the associated support vector, N is a number of support vectors included in the second set of support vectors, and C is a penalty constant.

15. The non-transitory computer-readable medium of claim 13 , wherein the center threshold value is computed using R 2 =K(x k ,x k )−2Σ i=1 N α i K(x i ,x k )+Σ i=1 N Σ j=1 N α i α j K(x i ,x j ), where x k is any support vector of the set of support vectors that have 0<α i <C, x i and x j are a support vector of the set of support vectors, α i and α j are the Lagrange constant of the associated support vector, N is a number of support vectors included in the set of support vectors, and C is a penalty constant.

16. The non-transitory computer-readable medium of claim 14 , wherein the center threshold value is computed using R 2 =(x k ·x k )−2Σ i=1 N α i (x i ·x k )+Σ i=1 N Σ j=1 N α i α j (x i ·x j ), where x k is any support vector of the set of support vectors that have 0<α i <C, x i and x j are a support vector of the set of support vectors, α i and α j are the Lagrange constant of the associated support vector, N is a number of support vectors included in the set of support vectors, and C is a penalty constant.

17. The non-transitory computer-readable medium of claim 1 , wherein the set of support vectors are computed by sampling from a plurality of windows of data, wherein each window of data of the plurality of windows of data includes a second plurality of observation vectors.

18. The non-transitory computer-readable medium of claim 1 , wherein solving the objective function defined for the SVDD model using the received first plurality of observation vectors comprises sampli

Assignees

• Sas Inst Inc

Inventors

• Kakde Deovrat Vijay
• Peredriy Sergiy
• Chaudhuri Arin
• Mcguirk Anya M

Classifications

• G06N99/005 Primary

  Physics · mapped topic

• H04L67/02

  based on web technology, e.g. hypertext transfer protocol [HTTP] · CPC title

• G06N20/10 Primary

  using kernel methods, e.g. support vector machines [SVM] · CPC title

• G06N20/00 Primary

  Machine learning · CPC title