Method for analyzing and classifying process data that operates a knowledge base in an open-book mode before defining any clusters

• US 8,036,999 B2
• Filed: 02/13/2008
• Issued: 10/11/2011
• Est. Priority Date: 02/14/2007
• Status: Expired due to Fees

• Current Assignee: Isagacity
• Original Assignee: Isagacity
• Inventors: Millett, Peter, Akhrarov, Murad, Sharma, Chanchal, Bates, Joe
• Primary Examiner(s): VINCENT, DAVID ROBERT
• Application Number:
  US12/030,437
• Publication Number:
  US 20100030521A1
• Time in Patent Office: 1,336 Days
• Field of Search: 702181-189, 702121-123, 706/12, 706/20, 706 45- 48, 707/737
• US Class Current: 706/45
• CPC Class Codes: G05B 13/0265 the criterion being a learning criterion
• CPC Class Codes: G05B 23/0229 knowledge based, e.g. expert systems; genetic algorithms
• CPC Class Codes: G06F 18/232 Non-hierarchical techniques
• CPC Class Codes: G06N 20/00 Machine learning

Abstract

Process data mining system and method. The system analyzes data from complex process plants or systems and operates in open-book and closed-book modes. In closed-book mode, the system monitors incoming data sets against pre-defined clusters of data values and generates reports, indicating whether incoming data is a match or a no-match with the pre-defined clusters. In open-book mode, the system generates initial clusters, without having a-priori knowledge of the component or process, and also creates clusters “on the fly”, thereby fine-tuning the analysis. A knowledge base encompasses a combination of parameters for a particular component. Clusters are defined within the knowledge base, each cluster representing a particular operating condition. The system expands clusters, within pre-defined limits, or creates new clusters, as needed, in order to accommodate incoming data values. Newly created clusters are then named, so as to indicate the particular operating conditions.

Claims

1. A method of analyzing data relating to parameters of a component in a process plant, a component being a piece of equipment, a location, or a process step, and a process plant being an agglomeration of equipment and/or process steps, the method comprising the steps of:

a) defining a set of data channels as a data vector for a particular component, each data channel in the set carrying data relating to a specific parameter of the particular component, such that a first data channel of the set carries data relating to a first specific parameter and a second data channel transmits data relating to a second specific parameter, the data vector being a discrete transmission of instantly measured data vector of the set of data channels, each discrete transmission being indicative of an operating condition of the particular component;

b) defining a knowledge base that relates to the particular component;

c) receiving the discrete transmission of the instantly measured data vector into the knowledge base;

d) prior to defining a cluster that encompasses a first data vector for a first operating condition of the particular component in the knowledge base, operating the knowledge base in open-book mode;

e) providing a cluster-generating algorithm that generates an initial cluster from the instantly measured data vector; and

f) adding the initial cluster to the knowledge base, the initial cluster subsequently serving as a pre-defined cluster.

2. The method of claim 1, further comprising the steps of:

g) providing one or more pre-defined clusters in the knowledge base, the pre-defined cluster being a pre-defined data vector having pre-defined boundaries for expected values of the data vector from the plurality of data channels in the knowledge base, each one cluster of the one or more pre-defined clusters being indicative of a specific one operating condition, such that, when more than one pre-defined cluster is provided, a first cluster is indicative of a first operating condition and a second cluster is indicative of a second operating condition;

h) providing a cluster-matching algorithm that determines whether one pre-defined cluster of the one or more pre-defined clusters encompass transmitted values of the instantly measured data vector;

i) upon determining that the one pre-defined cluster encompasses the transmitted values of the instantly measured data vector, adding the instantly measured data vector to the one pre-defined cluster; and

j) upon determining that none of the one or more pre-defined clusters encompass the transmitted values of the instantly measured data vector, creating a new cluster having boundaries that encompass the instantly measured data vector.

3. The method of claim 2, further comprising the step of:

k) operating the knowledge base in closed-book mode, for monitoring operation of the particular component;

l) upon determination of the cluster-matching algorithm that the transmitted values of the instantly measured data vector fall within the pre-defined boundaries of the pre-defined cluster, allocating the instantly measured data vector to the pre-defined cluster; and

,

m) upon determination of the cluster-matching algorithm that the transmitted values of the instantly measured data vector fall outside the pre-defined boundaries of the pre-defined cluster, issuing a report on the instantly measured data vector.

4. The method of claim 1, further comprising the following steps:

n) identifying each pre-defined cluster of the one or more pre-defined clusters within the knowledge base as representative of a particular operating state, wherein a first pre-defined cluster corresponds to a first operating state and a second pre-defined cluster corresponds to a second operating state;

o) comparing the instantly measured data vector with the pre-defined clusters, so as to determine a closest cluster, the closest cluster having cluster boundaries that encompass the transmitted values of the instantly measured data vector;

p) determining a residual value of the instantly measured data vector relative to an expected value of the closest cluster, the expected value being defined by the cluster boundaries and the residual value defining a distance between the instantly measured data vector and the expected value of the closest cluster;

q) determining a match condition if the residual value is within the cluster boundaries; and

r) determining a no-match condition, if the residual value is outside the cluster boundaries.

5. The method of claim 4, further comprising the steps of:

s) providing a pre-defined boundary-expansion tolerance for each of the one or more pre-defined clusters;

t) determining that the residual value of the instantly measured data vector lies within the pre-defined boundary-expansion tolerance of one of the pre-defined clusters; and

u) expanding the knowledge base by automatically expanding the cluster boundaries of said one of the pre-defined clusters, so as to obtain an expanded cluster that encompasses the transmitted values of the instantly measured data vector.

6. The method of claim 5, further comprising the steps of:

v) within the knowledge base, defining all the data channels of the knowledge base as base level parameters;

w) within each of the one or more pre-defined clusters, defining one or more of the base level parameters as being particularly indicative of a particular operating state and defining the at least one or more base level parameters to also be a corresponding one or more key level parameters;

x) within the knowledge base, defining a key-level knowledge base that contains a key-level cluster comprising a selection of the one or more key level parameters, the key-level cluster being indicative of a particular operating state of the component;

y) first processing with the algorithm the key level parameters in the key-level knowledge base to determine a match with the key-level cluster, so as to determine the particular operating state, and then subsequently processing the base level parameters against the pre-defined clusters that correspond to the particular operating state.

7. The method of claim 6, the step of processing the base level parameters comprising the steps of:

z) forming a base-level knowledge base comprising all data channels;

aa) within the base-level knowledge base defining one or more parent clusters and one or more child clusters, wherein a first parent cluster corresponds to a first operating condition and a first child cluster of the first parent cluster corresponds to a particular scenario within the first operating condition;

ab) applying a fingerprinting algorithm to the one or more parent clusters;

ac) searching for a match of the data values of the discrete transmission with the expected value of an individual parent cluster of the one or more parent clusters;

ad) if a match is found with the individual parent cluster, associating a cluster ID of the individual cluster with the transmitted values of the instantly measured data vector and incrementing a cluster counter for the individual parent cluster; and

ae) if a no-match with the one or more parent clusters is found, expanding the base-level knowledge base by generating a new algorithm-defined parent cluster that reflects the transmitted values of the instantly measured data vector.

8. The method of claim 6, the step of processing the key level parameters further comprising the steps of:

af) ascertaining whether the process plant includes a plurality of plants;

ag) allocating a collection of the data channels to a corresponding one plant of the plurality plants; and

ah) sequentially processing a first knowledge base of each plant of the plurality of plants and then a second knowledge base of the each plant.

9. The method of claim 2, further comprising the steps of:

ai) operating the cluster-generating algorithm concurrently in the open-book mode and in a closed book mode, wherein the closed-book mode disallows generation of a new initial cluster;

aj) providing pre-existing data for the knowledge base, so as to provide at least one cluster that is a first pre-defined cluster that has cluster boundaries that are defined by a minimum-maximum tolerance range for each data vector;

ak) comparing the transmitted values of the instantly measured data vector with values in the first pre-defined cluster;

al) determining whether the transmitted values of the instantly measured data vector fall within the cluster boundaries of the first pre-defined cluster;

am) upon determining a match, issuing a match report; and

an) upon determining a no-match, issuing a no-match report.

10. The method of claim 1, the step of defining a knowledge base further comprising the step of:

b1) defining a plurality of knowledge bases, each knowledge base of the plurality relating to one particular component and containing data channels for the one particular component.

11. The method of claim 10 further comprising the steps of:

b2) defining a first knowledge base as comprising a first set of data channels and defining a second knowledge base as comprising a second set of data channels, wherein the first and the second sets of data channels are exclusive of each other.