Fault detection system utilizing dynamic principal components analysis

Patent metadata

Field	Value
Publication number	US-2018136019-A1
Application number	US-201715811477-A
Country	US
Kind code	A1
Filing date	Nov 13, 2017
Priority date	Nov 11, 2016
Publication date	May 17, 2018
Grant date	—

Abstract

Methods and systems for detecting a fault in a data set from an industrial process are disclosed. One method includes forming a first data matrix at a data processing framework from time-series training data, and performing a principal component pursuit on the first data matrix to form an uncorrupted, unscaled matrix and a sparse matrix in the memory, and scaling the uncorrupted, unscaled matrix to form an uncorrupted scaled matrix. The method also includes performing a dynamic principal component analysis (DPCA) on the uncorrupted scaled matrix to form a DPCA model, and determining a squared prediction error from the DPCA model. Based on the squared prediction error, faults are detected in a different data set from operation of the industrial process. At least one of (1) correcting the one or more faults in the different data set or (2) performing a repair operation on a sensor is performed.

First claim

1 . A computer-implemented method for detecting a fault in a data set from an industrial process, the method comprising:
forming, in memory of a computing system, a first data matrix at a data processing framework from training data, the training data from operation of an industrial process having at least two sensors, wherein the training data comprises time-series data; performing a principal component pursuit on the first data matrix to form an uncorrupted, unscaled matrix and a sparse matrix in the memory; scaling the uncorrupted, unscaled matrix to form an uncorrupted scaled matrix; performing a dynamic principal component analysis on the uncorrupted scaled matrix to form a dynamic principal component analysis model; determining a squared prediction error from the dynamic principal component analysis model; based on the squared prediction error, detecting one or more faults in a different data set from operation of the industrial process having the at least two sensors; and performing at least one of (1) correcting the one or more faults in the different data set or (2) performing a repair operation on a sensor from among the at least two sensors.

2 . The computer-implemented method of claim 1 , wherein the industrial process comprises a process occurring at a hydrocarbon facility.

3 . The computer-implemented method of claim 1 , wherein the first data matrix comprises an augmented matrix including time-lagged variables from an original matrix including the training data.

4 . The computer-implemented method of claim 3 , wherein the augmented matrix is only generated before the step of performing the principal component pursuit.

5 . The computer-implemented method of claim 1 , further comprising augmenting the uncorrupted matrix to form a two-dimensional matrix having a predetermined number of rows and a predetermined number of columns, the predetermined number of rows corresponding to a number of time-sequence samples included in the training data and the predetermined number of columns corresponding to a number of time-lagged variables being considered.

6 . The computer-implemented method of claim 1 , wherein the principal component pursuit includes a parameter defining a threshold for noise from the first data matrix.

7 . The computer-implemented method of claim 6 , further comprising generating a noise matrix based on the threshold for noise from the first data matrix.

8 . The computer-implemented method of claim 7 , further comprising adding the noise matrix to the uncorrupted, unscaled matrix and using the noisy, uncorrupted, unscaled matrix for performing the dynamic principal component analysis.

9 . The computer-implemented method of claim 6 , wherein the parameter is greater than zero.

10 . The computer-implemented method of claim 1 , further comprising generating an alert indicating the existence of the one or more faults.

11 . The computer-implemented method of claim 1 , further comprising tuning a parameter associated with the sparse matrix to balance a rate of fault detection against a rate of false alarming.

12 . The computer-implemented method of claim 1 , wherein the training data includes a plurality of errors represented by a sparse data matrix.

13 . A fault detection system useable to detect a fault in a data set from an industrial process, the fault detection system comprising:
a computing system including a processor and a memory communicatively connected to the processor, the computing system configured to execute, based on instructions stored in the memory, a method, the method comprising: forming, in the memory, a first data matrix at a data processing framework from training data, the training data from operation of an industrial process having at least two sensors, wherein the training data comprises time-series data; performing a principal component pursuit on the first data matrix to form an uncorrupted, unscaled matrix and a sparse matrix in the memory; scaling the uncorrupted, unscaled matrix to form an uncorrupted scaled matrix; performing a dynamic principal component analysis on the uncorrupted scaled matrix to form a dynamic principal component analysis model; determining a squared prediction error from the dynamic principal component analysis model; based on the squared prediction error, detecting one or more faults in a different data set from operation of the industrial process having the at least two sensors; and performing at least one of (1) correcting the one or more faults in the different data set or (2) initiating a repair operation on a sensor from among the at least two sensors.

14 . The system of claim 13 , wherein the first data matrix comprises an augmented matrix including time-lagged variables from an original matrix including the training data.

15 . The system of claim 14 , wherein the augmented matrix is only generated before the step of performing the principal component pursuit.

16 . The computer-implemented method of claim 13 , wherein the principal component pursuit includes a parameter defining a threshold for noise from the first data matrix.

17 . The system of claim 16 , further comprising generating a noise matrix based on the threshold for noise from the first data matrix.

18 . The system of claim 17 , further comprising adding the noise matrix to the uncorrupted, unscaled matrix and using the noisy, uncorrupted, unscaled matrix for performing the dynamic principal component analysis.

19 . The system of claim 16 , wherein the parameter is greater than zero.

20 . The system of claim 13 , further comprising a display configured to display an alert to the user based on detection of the one or more faults.

21 . The system of claim 13 , wherein the computing system comprising one or more computing devices.

22 . The system of claim 13 , further comprising generating an alert indicating the existence of the one or more faults to a user of the system.

23 . The system of claim 13 , wherein the training data includes a plurality of errors represented by a sparse data matrix.

24 . A fault detection system useable to detect a fault in a data set generated from at least two sensors monitoring a process within a hydrocarbon facility, the fault detection system comprising:
a computing system including a processor and a memory communicatively connected to the processor, the computing system configured to execute, based on instructions stored in the memory, a method, the method comprising: forming, in the memory, a first data matrix at a data processing framework from training data, the training data from at least two sensors associated with a process within a hydrocarbon facility, the training data comprising time-series data including any errors represented by a sparse data matrix; performing a principal component pursuit on the first data matrix to form an uncorrupted, unscaled matrix and a sparse matrix in the memory, wherein performing the principal component pursuit includes tuning a parameter associated with the sparse matrix to balance a rate of fault detection against a rate of false alarming; scaling the uncorrupted, unscaled matrix to form an uncorrupted scaled matrix; performing a dynamic principal component analysis on the uncorrupted matrix to form a dynamic principal component analysis model; determining a squared prediction error from the dynamic principal component analysis model; based on the squared prediction error, det

Assignees

• Chevron Usa Inc
• Univ Southern California

Inventors

• Deshpande Alisha
• Qin Si-Zhao J
• Brenskelle Lisa Ann

Classifications

• G06F2218/08

  Feature extraction · CPC title

• G06F18/2135

  based on approximation criteria, e.g. principal component analysis · CPC title

• G06F17/16

  Matrix or vector computation {, e.g. matrix-matrix or matrix-vector multiplication, matrix factorization (matrix transposition G06F7/78)} · CPC title

• G01D18/002 Primary

  Automatic recalibration (G01D18/008 takes precedence) · CPC title

• G06F17/18

  for evaluating statistical data {, e.g. average values, frequency distributions, probability functions, regression analysis (forecasting specially adapted for a specific administrative, business or logistic context G06Q10/04)} · CPC title