Method for identifying root cause failure in a multi-parameter self learning machine application model

• US 9,223,667 B2
• Filed: 11/08/2012
• Issued: 12/29/2015
• Est. Priority Date: 11/09/2011
• Status: Active Grant

• Current Assignee: Avo Multi-AMP Corporation (Thyssen Bornemisza Family)
• Original Assignee: ABet SKF
• Inventors: Sanford, Matthew Barber, Wiedenbrug, Ernesto Jorge
• Primary Examiner(s): Bonzo, Bryce
• Assistant Examiner(s): Mehrmanesh, Elmira
• Application Number:
  US13/671,842
• Publication Number:
  US 20130117607A1
• Time in Patent Office: 1,146 Days
• Field of Search: 714/26, 714/47.2, 714/47.3
• US Class Current: 1/1
• CPC Class Codes: G06F 11/25 Testing of logic operation, e.g. by logic analysers
• CPC Class Codes: G06N 20/00 Machine learning

Abstract

A method for identifying root cause failure in a multi-parameter self learning machine application model is presented. At least one multi-function sensor having the capability to measure at least one of a voltage and current of the machine application model is provided. The method includes measuring voltages and currents of a multi-phase load with the multi-function sensors in a passive manor and calculating at least one of a time-varying variable KW, PF, kVAr, or Z out of the measured voltages and currents. The method further provides calculating a first, second or third order derivative of the time-varying variable and classifying segments of at least one of the time-varying variables depending on a state. Then, a step of choosing at least one of the variables and learning their normal behavior is undertaken. Finally, normal behavior is compared to a pattern difference and a root-cause meaning to the pattern difference is identified.

Claims

1. A method for identifying root cause failure in a multi-parameter self learning machine application model comprising:

providing at least one multi-function sensor having the capability to measure at least one of a voltage and current of the machine application model;

measuring voltages and currents of a multi-phase load with the multi-function sensors in a passive manner which includes sensing existing voltages and currents;

storing and accepting the measured voltages and currents into memory by a computer controlled analog to digital converter;

calculating at least one of a time-varying variable KW, PF, kVAr, or Z out of the measured voltages and currents;

calculating at least one of a first, second or third order derivative of the at least one time-varying variable;

classifying segments of at least one of the time-varying variables depending on a state;

choosing at least one of the calculated time-varying variables and learning their normal behavior;

comparing the normal behavior to a pattern difference;

identifying a root cause meaning to the pattern difference;

plotting Min, Max and Median values of the measured voltages and currents in a candlestick chart format; and

determining the directions with which the measured voltages and currents are heading using the Min, Max and Median values.

2. The method of claim 1, wherein the multi-function sensor further comprises the capability of measuring vibration.

3. The method of claim 1, wherein the measured voltages and currents are watched 24 hours a day, 7 days a week by a reading device.

4. The method of claim 2, wherein a data base disposed on the reading device comprises an array to store the data.

5. The method of claim 1, further comprising an auto refining feature to improve the resolution of a segment width and increasing confidence of an issued alarm by adding a statistical evaluation of a preceding time-window being outside the cases seen defined during the learning phase.

6. The method of claim 1, wherein the method of identifying root cause failure in the self learning machine application model is applicable to a motor.

7. The method of claim 1, wherein the method of identifying root cause failure in the self learning machine application model is applicable to a generator.

8. The method of claim 1, wherein the method of identifying root cause failure in the self learning machine application model takes into consideration condition and behavior of the machine.

9. The method of claim 1, wherein the method of identifying root cause failure in the self learning machine application model takes into consideration condition and behavior of a load applied to the machine.

10. The method of claim 1, wherein any disturbances straying from usual and healthy mode of operation are identified and may introduce signatures that can be monitored and diagnosed on the voltage and current signals.

11. The method of claim 1, wherein measuring the voltages and currents of the multi-phase load in a passive manner includes sensing only existing voltages and currents.

12. The method of claim 1, wherein both time and frequency domain data is captured and stored.

13. The method of claim 12, wherein the stored data is used to compare against a respective boundary of a representative segment to detect an eminent system failure.

14. A method for identifying root cause failure in a multi-parameter self learning machine application model comprising:

providing at least one multi-function sensor having the capability to measure at least one of a voltage and current of the machine application model;

measuring voltages and currents of a multi-phase load with the multi-function sensors in a passive manner;

calculating at least one of a time-varying variable KW, PF, kVAr, or Z out of the measured voltages and currents;

calculating at least one of a first, second or third order derivative of the at least one time-varying variable;

classifying segments of at least one of the time-varying variables depending on a state;

choosing at least one of the calculated time-varying variables and learning their normal behavior;

comparing the normal behavior to a pattern difference; and

identifying a root cause meaning to the pattern difference,

wherein the derivatives of the voltages include a newly created variable for instantaneous voltage phasor that provides both amplitude and phase.

15. A method for identifying root cause failure in a multi-parameter self learning machine application model comprising:

providing at least one multi-function sensor having the capability to measure at least one of a voltage and current of the machine application model;

measuring voltages and currents of a multi-phase load with the multi-function sensors in a passive manner;

calculating at least one of a time-varying variable KW, PF, kVAr, or Z out of the measured voltages and currents;

calculating at least one of a first, second or third order derivative of the at least one time-varying variable;

classifying segments of at least one of the time-varying variables depending on a state;

choosing at least one of the calculated time-varying variables and learning their normal behavior;

comparing the normal behavior to a pattern difference; and

identifying a root cause meaning to the pattern difference,

wherein the derivatives of the currents include a newly created variable for instantaneous current phasor that provides both amplitude and phase.

16. A method for identifying root cause failure in a multi-parameter self learning machine application model comprising:

providing at least one multi-function sensor having the capability to measure at least one of a voltage and current of the machine application model;

measuring voltages and currents of a multi-phase load with the multi-function sensors in a passive manner;

calculating at least one of a time-varying variable KW, PF, kVAr, or Z out of the measured voltages and currents;

calculating at least one of a first, second or third order derivative of the at least one time-varying variable;

classifying segments of at least one of the time-varying variables depending on a state;

choosing at least one of the calculated time-varying variables and learning their normal behavior;

comparing the normal behavior to a pattern difference; and

identifying a root cause meaning to the pattern difference,

wherein the derivatives of the voltages and/or currents include a newly created variable for instantaneous impedance that provides both amplitude and phase.

17. A method for identifying root cause failure in a multi-parameter self learning machine application model comprising:

providing at least one multi-function sensor having the capability to measure at least one of a voltage and current of the machine application model;

measuring voltages and currents of a multi-phase load with the multi-function sensors in a passive manner;

calculating at least one of a time-varying variable KW, PF, kVAr, or Z out of the measured voltages and currents;

calculating at least one of a first, second or third order derivative of the at least one time-varying variable;

classifying segments of at least one of the time-varying variables depending on a state;

choosing at least one of the calculated time-varying variables and learning their normal behavior;

comparing the normal behavior to a pattern difference; and

identifying a root cause meaning to the pattern difference,

wherein the derivatives of the voltages and/or currents include a newly created variable for frequency that provides a delta angle of instantaneous voltage phasor or current phasor versus time.