Model-plant mismatch detection with support vector machine for cross-directional process behavior monitoring

Patent metadata

Field	Value
Publication number	US-11449046-B2
Application number	US-201715611082-A
Country	US
Kind code	B2
Filing date	Jun 1, 2017
Priority date	Sep 16, 2016
Publication date	Sep 20, 2022
Grant date	Sep 20, 2022

Abstract

A method includes obtaining operating data associated with operation of a cross-directional industrial process controlled by at least one model-based process controller. The method also includes, during a training period, performing closed-loop model identification with a first portion of the operating data to identify multiple sets of first spatial and temporal models. The method further includes identifying clusters associated with parameter values of the first spatial and temporal models. The method also includes, during a testing period, performing closed-loop model identification with a second portion of the operating data to identify second spatial and temporal models. The method further includes determining whether at least one parameter value of at least one of the second spatial and temporal models falls outside at least one of the clusters. In addition, the method includes, in response to such a determination, detecting that a mismatch exists between actual and modeled behaviors of the industrial process.

First claim

What is claimed is:

1. An apparatus comprising:
at least one memory configured to store operating data associated with operation of a cross-directional industrial process that is controlled by at least one model-based process controller; and
at least one processing device configured to:
during a training period, perform closed-loop model identification with a first portion of the operating data to identify multiple sets of first spatial and temporal models;
identify clusters associated with parameters using a support vector machine to determine clustering of the model parameters, the support vector machine implements a kernel function that generates linear or nonlinear clustering of data by mapping original data into a space where a cluster boundary is identified and the support vector machine is used for binary classifications;
during a testing period, perform closed-loop model identification with a second portion of the operating data to identify second spatial and temporal models;
determine whether at least one parameter value of at least one of the second spatial and temporal models falls outside at least one of the clusters, comprises:
determine a functional distance of a point representing a new model parameter to the at least one of the clusters associated with the second spatial and temporal models; and
classify, based on the functional distance, whether the at least one parameter value is inside or outside the at least one of the clusters;

in response to a determination that the at least one parameter value of at least one of the second spatial and temporal models falls outside at least one of the clusters, detect that a mismatch exists between actual and modeled behaviors of the industrial process; and
generate an alarm to take a corrective action including initiation of a model identification process to identify a replacement spatial or temporal model.

2. The apparatus of claim 1 , wherein, to perform the closed-loop model identification with the first or second portion of the operating data, the at least one processing device is configured to:
iteratively:
estimate one or more temporal parameters while one or more spatial parameters are fixed;
normalize at least one of the one or more temporal parameters; and
estimate the one or more spatial parameters while the one or more temporal parameters are fixed.

3. The apparatus of claim 2 , wherein, to perform the closed-loop model identification with the first or second portion of the operating data, the at least one processing device is further configured to:
filter the first or second portion of the operating data using the temporal and spatial parameters;
estimate a temporal model using the filtered first or second portion of the operating data; and
rescale the one or more spatial parameters based on the estimated temporal model.

4. The apparatus of claim 1 , wherein, to determine whether the at least one parameter value of at least one of the second spatial and temporal models falls outside at least one of the clusters, the at least one processing device is configured to:
map parameter values of the second spatial and temporal models into the higher-dimensional feature space using the support vector machine; and
determine whether each parameter value of the second spatial and temporal models falls outside the at least one of the clusters by calculating the functional distance between a point representing that parameter value and the boundary of that cluster.

5. The apparatus of claim 4 , wherein the support vector machine comprises a one-class support vector machine.

6. The apparatus of claim 1 , wherein:
the at least one processing device is configured to perform the closed-loop model identification with the first portion of the operating data multiple times using a sliding window within the first portion of the operating data; and
the at least one processing device is configured to perform the closed-loop model identification with the second portion of the operating data multiple times using the sliding window within the second portion of the operating data.

7. The apparatus of claim 1 , wherein:
the at least one processing device is configured to perform the closed-loop model identification with the first portion of the operating data to identify first noise models;
the at least one processing device is configured to perform the closed-loop model identification with the second portion of the operating data to identify second noise models; and
the at least one processing device is further configured to determine that no mismatch exists between the actual and modeled behaviors of the industrial process in response to a determination that at least one parameter value of the second noise models falls outside a cluster associated with parameter values of the first noise models.

8. A method comprising:
obtaining operating data associated with operation of a cross-directional industrial process that is controlled by at least one model-based process controller;
during a training period, performing closed-loop model identification with a first portion of the operating data to identify multiple sets of first spatial and temporal models;
identifying clusters associated with parameters using a support vector machine to determine clustering of the model parameters, the support vector machine implements a kernel function that generates linear or nonlinear clustering of data by mapping original data into a space where a cluster boundary is identified and the support vector machine is used for binary classifications;
during a testing period, performing closed-loop model identification with a second portion of the operating data to identify second spatial and temporal models;
determining whether at least one parameter value of at least one of the second spatial and temporal models falls outside at least one of the clusters, comprises:
determine a functional distance of a point representing a new model parameter to the at least one of the clusters associated with the second spatial and temporal models; and
classify, based on the functional distance, whether the at least one parameter value is inside or outside the at least one of the clusters;

in response to a determination that the at least one parameter value of at least one of the second spatial and temporal models falls outside at least one of the clusters, detecting that a mismatch exists between actual and modeled behaviors of the industrial process; and
generate an alarm to take a corrective action including initiation of a model identification process to identify a replacement spatial or temporal model.

9. The method of claim 8 , wherein performing the closed-loop model identification with the first or second portion of the operating data comprises:
iteratively:
estimating one or more temporal parameters while one or more spatial parameters are fixed;
normalizing at least one of the one or more temporal parameters; and
estimating the one or more spatial parameters while the one or more temporal parameters are fixed.

10. The method of claim 9 , wherein performing the closed-loop model identification with the first or second portion of the operating data further comprises:
filtering the first or second portion of the operating data using the temporal and spatial parameters;
estimating a temporal model using the filtered first or second portion of the operating data; and
rescaling the one or more spatial parameters based on the estimated temporal model.

11. The method of claim 8 , wherein determining whether the at least one parameter value of at least one of the second spatial and temporal models falls outside at least one of the cluste

Assignees

• Honeywell Ltd

Inventors

• Lu Qiugang
• Gopaluni R Bhushan
• Forbes Michael G
• Loewen Philip D
• Backstrom Johan U
• Dumont Guy A

Classifications

• G05B13/041

  in which a variable is automatically adjusted to optimise the performance · CPC title

• G05B2219/31088

  Network communication between supervisor and cell, machine group · CPC title

• G05B19/4185 Primary

  characterised by the network communication · CPC title

• G05B13/048

  using a predictor · CPC title

• G06N20/10

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