Population-based learning with deep belief networks

• US 9,842,302 B2
• Filed: 08/26/2015
• Issued: 12/12/2017
• Est. Priority Date: 08/26/2013
• Status: Active Grant

• Current Assignee: Aspentech Corporation
• Original Assignee: Mtelligence Corporation (Aspentech Corporation)
• Inventors: Bates, Alexander B., Kim, Caroline, Rahilly, Paul
• Primary Examiner(s): HENSON, MISCHITA L
• Application Number:
  US14/836,848
• Publication Number:
  US 20160116378A1
• Time in Patent Office: 839 Days
• CPC Class Codes: G05B 19/4184 characterised by fault tolerance, reliability of production system
• CPC Class Codes: G05B 23/024 Quantitative history assessment, e.g. mathematical relationships between available data; Functions therefor; Principal component analysis [PCA]; Partial least square [PLS]; Statistical classifiers, e.g. Bayesian networks, linear regression or correlation analysis; Neural networks
• CPC Class Codes: G06N 20/00 Machine learning
• CPC Class Codes: G06N 3/084 Backpropagation, e.g. using gradient descent
• CPC Class Codes: Y02P 90/02 Total factory control, e.g. smart factories, flexible manufacturing systems [FMS] or integrated manufacturing systems [IMS]

Abstract

A plant asset failure prediction system and associated method. The method includes receiving user input identifying a first target set of equipment including a first plurality of units of equipment. A set of time series waveforms from sensors associated with the first plurality of units of equipment are received, the time series waveforms including sensor data values. A processor is configured to process the time series waveforms to generate a plurality of derived inputs wherein the derived inputs and the sensor data values collectively comprise sensor data. The method further includes determining whether a first machine learning agent may be configured to discriminate between first normal baseline data for the first target set of equipment and first failure signature information for the first target set of equipment. The first normal baseline data of the first target set of equipment may be derived from a first portion of the sensor data associated with operation of the first plurality of units of equipment in a first normal mode and the first failure signature information may be derived from a second portion of the sensor data associated with operation of the first plurality of units of equipment in a first failure mode. Monitored sensor signals produced by the one or more monitoring sensors are received. The first machine learning agent is then and activated, based upon the determining, to monitor data included within the monitored sensor signals.

Claims

1. A computer implemented method of monitoring operation-of equipment and predicting equipment failure in an industrial plant, the method comprising:

in a computer system having one or more processors;

receiving, through a user interface, user input identifying a first target set of equipment including a first plurality of units of equipment, wherein the first plurality of units of equipment is configured with one or more monitoring sensors, and wherein each equipment unit of the first plurality of units of equipment has a first plurality of matching target parameters;

receiving a set of time series waveforms from sensors associated with the first plurality of units of equipment, the time series waveforms including sensor data values;

processing the received time series waveforms to generate a plurality of derived inputs wherein the derived inputs and the sensor data values collectively compose sensor data;

training a first machine learning agent to;

(i) discriminate between first normal baseline data for the first target set of equipment and first failure signature information for the first target set of equipment; and

(ii) detect operational conditions associated with onset of an equipment failure by identifying patterns or relationships within the sensor data;

wherein the first normal baseline data of the first target set of equipment is derived from a first portion of the sensor data associated with operation of the first plurality of units of equipment in a first normal mode, and wherein the first failure signature information is derived from a second portion of the sensor data associated with operation of the first plurality of units of equipment in a first failure mode;

receiving monitored sensor signals produced by the one or more monitoring sensors;

analyzing, by the trained first machine learning agent, data included within the received monitored sensor signals, the analyzing detecting, from the patterns or relationships within the data, an operational condition identifying an equipment failure onset of one or more monitored units of equipment, such that the analyzing by the first machine learning agent results in predicting failure of the one or more monitored units of equipment; and

transferring an indication of the predicted failure to a plant control system and, using the plant control system, controlling the one or more monitored units of equipment to avert the predicted failure based upon the transferred indication.

2. The method of claim 1 further including identifying a second target set of equipment including a second plurality of units of equipment, wherein each equipment unit of the second plurality of units of equipment is characterized by a second plurality of matching target parameters and wherein the second plurality of matching target parameters are a subset of the first plurality of matching target parameters;

wherein the determining includes determining whether a second machine learning agent may be configured to discriminate between second normal baseline data for the second target set of equipment and second failure signature information for the second target set of equipment wherein the second normal baseline data of the second target set of equipment is derived from a portion of the sensor data relating to operation of the second target set of equipment in a second normal mode and wherein the second failure signature information is derived from a portion of the sensor data relating to operation of the second plurality of units of equipment in a second failure mode.

3. The method of claim 2 further including identifying a third target set of equipment including a third plurality of units of equipment, wherein each equipment unit of the third plurality of units of equipment is characterized by a third plurality of matching target parameters and wherein the third plurality of matching target parameters are a subset of the second plurality of matching target parameters;

wherein the determining includes determining whether a third machine learning agent may be configured to discriminate between third normal baseline data for the third target set of equipment and third failure signature information for the third target set of equipment wherein the third normal baseline data of the third target set of equipment is derived from a portion of the sensor data relating to operation of the third plurality of units of equipment in a third normal mode and wherein the third failure signature information is derived from a portion of the sensor data relating to operation of the third plurality of units of equipment in a third failure mode.

4. The method of claim 1 wherein the first plurality of matching target parameters include at least one of:

(i) make and model, and (ii) operating context.

5. The method of claim 2 wherein the first plurality of matching target parameters include (i) make and model, and (ii) operating context and wherein the second plurality of matching target parameters include one of (i) make and model, and (ii) operating context.

6. The method of claim 2 wherein the first plurality of matching target parameters include (i) make, model and revision, and (ii) operating context and wherein the second plurality of matching target parameters include (i) make and model and (ii) operating context.

13. The method of claim 1 wherein the determining includes determining whether the first machine learning agent may be configured to discriminate between the first normal baseline data and the first failure signature information with an accuracy exceeding a configurable threshold.

15. The method of claim 2 wherein:

the first plurality of units of equipment are of a first make and model and a first revision, and are associated with a first operational context, and

the second plurality of units of equipment include equipment units that are of the first make and model and the first revision, and are associated with the first operational context and further include other equipment units that are of the first make and model and a second revision different from the first revision, and are associated with the first operational context.

7. A computer implemented method of monitoring operation of equipment and predicting equipment failure in an industrial plant, the method comprising:

in a computer system having one or more processors;

identifying a first target set of equipment including a first plurality of units of equipment, wherein the first plurality of units of equipment is configured with one or more monitoring sensors, and wherein each equipment unit of the first plurality of units of equipment has a first plurality of matching target parameters;

receiving sensor data from sensors associated with the first plurality of units of equipment;

training a first machine learning agent to;

(i) discriminate between first normal baseline data for the first target set of equipment and first failure signature information for the first target set of equipment; and

(ii) detect operational conditions associated with onset of an equipment failure by identifying patterns or relationships within the sensor data;

wherein the first normal baseline data of the first target set of equipment is derived from a first portion of the sensor data associated with operation of the first plurality of units of equipment in a first normal mode, and wherein the first failure signature information is derived from a second portion of the sensor data associated with operation of the first plurality of units of equipment in a first failure mode;

receiving monitored sensor signals produced by the one or more monitoring sensors;

analyzing, by the trained first machine learning agent, data included within the received monitored sensor signals, the analyzing detecting, from the patterns or relationships within the data, an operational condition identifying an equipment failure onset of one or more monitored units of equipment, such that the analyzing by the first machine learning agent results in predicting failure of the one or more monitored units of equipment; and

transferring an indication of the predicted failure to a plant control system and, using the plant control system, controlling the one or more monitored units of equipment to avert the predicted failure based upon the transferred indication.

8. The method of claim 7 further including identifying a second target set of equipment including a second plurality of units of equipment, wherein each equipment unit of the second plurality of units of equipment is characterized by a second plurality of matching target parameters and wherein the second plurality of matching target parameters are a subset of the first plurality of matching target parameters;

wherein the determining includes determining whether a second machine learning agent may be configured to discriminate between second normal baseline data for the second target set of equipment and second failure signature information for the second target set of equipment wherein the second normal baseline data of the second target set of equipment is derived from a portion of the sensor data associated with operation of the second plurality of units of equipment in a second normal mode and wherein the second failure signature information is derived from a portion of the sensor data associated with operation of by the second plurality of units of equipment in a second failure mode.

9. The method of claim 8 further including identifying a third target set of equipment including a third plurality of units of equipment, wherein each equipment unit of the third plurality of units of equipment is characterized by a third plurality of matching target parameters and wherein the third plurality of matching target parameters are a subset of the second plurality of matching target parameters;

wherein the determining includes determining whether a third machine learning agent may be configured to discriminate between third normal baseline data for the third target set of equipment and third failure signature information for the third target set of equipment wherein the third normal baseline data of the third target set of equipment is derived from a portion of the sensor data associated with operation of the third plurality of units of equipment in a third normal mode and wherein the third failure signature information is derived from a portion of the sensor data associated with operation of the third plurality of units of equipment in a third failure mode.

10. The method of claim 7 wherein the first plurality of matching target parameters include at least one of:

(i) make and model, and (ii) operating context.

11. The method of claim 8 wherein the first plurality of matching target parameters include each of (i) make and model, and (ii) operating context and wherein the second plurality of matching target parameters include one of (i) make and model, and (ii) operating context.

12. The method of claim 8 wherein the first plurality of matching target parameters include (i) make, model and revision, and (ii) operating context and wherein the second plurality of matching target parameters include (i) make and model and (ii) operating context.

14. The method of claim 7 wherein the determining includes determining whether the first machine learning agent may be configured to discriminate between the first normal baseline data and the first failure signature information with an accuracy exceeding a configurable threshold.