Method and apparatus for diagnosing the state of a machine

• US 5,251,151 A
• Filed: 12/24/1990
• Issued: 10/05/1993
• Est. Priority Date: 05/27/1988
• Status: Expired due to Fees

• Current Assignee: The Research Foundation for The State University of New York (State University of New York)
• Original Assignee: The Research Foundation for The State University of New York (State University of New York)
• Inventors: Wright, Selwyn, Benenson, David M., Soom, Andres, Soumekh, Mehrdad, Park, Soon Y., Acharya, Raj S., Demjanenko, Victor
• Primary Examiner(s): Harvey, Jack B.
• Assistant Examiner(s): CHOI, KYLE JAEHUN
• Application Number:
  US07/634,091
• Time in Patent Office: 1,016 Days
• Field of Search: 364/550, 364/576, 364/474.17, 364/508, 364/551.02, 73/654, 381/43, 395/2
• US Class Current: 702/56
• CPC Class Codes: G01H 1/00 Measuring characteristics of vibrations in solids by using direct conduction to the detector G01H9/00, G01H11/00 take precedence

Abstract

A method and apparatus for diagnosing the mechanical condition of a machine. The method of the invention includes the following steps: sensing a reference vibrational signal of the machine; creating a reference set of events from the reference vibrational signal of the machine; saving the reference set of events; sensing a test vibrational signal of the machine; creating a test set of events from the test vibrational signal of the machine; and, comparing the test set of events with the reference set of events to determine the mechanical condition of the machine. The invention further includes a method of extracting a set of events from a vibrational signal of a machine wherein a subset of the events corresponds to a physical action of the machine. The invention also includes a method of determining a set of clustered events for a machine from a plurality of sets of events, wherein each of the sets has a number of events therein. An apparatus is also disclosed for implementing the method of the invention

Claims

1. A method of extracting a set of events from a vibration signal of a machine wherein a subset of said events corresponds to a physical action of said machine, comprising:

sensing a vibrational signal of said machine;

creating a time domain signature from said vibrational signals;

processing said signature to obtain a set of clustered events from a plurality of sets of possible events including in said set of clustered events a subset which corresponds to the physical action of said machine; and

comparing said subset with a reference subset to determine the mechanical condition of said machine.

2. A method as recited in claim 1 wherein said time domain signature is created from said vibrational signal by:

computing an envelope of said vibrational signal; and

,

differentiating said envelope to create said signature.

3. A method as recited in claim 1 wherein said processing step comprises the step of an envelope of said vibrational signal which computing step comprises:

computing a Hilbert Transform of said vibrational signal to obtain a transformed signal;

taking the square root of the sum of the square of the transformed signal and the square of the vibrational signal.

4. A method as recited in claim 1 wherein said vibrational signal comprises a plurality of signals and said processing step comprises the step of computing of an envelope of said vibrational signal which further comprises:

computing an envelope of each of said plurality of signals comprising said vibrational signal thereby providing a plurality of envelopes; and

performing a selective averaging process to form a single envelope which is representative of said plurality of envelopes.

5. A method as recited in claim 1 wherein said processing step comprises the step of computing of an envelope of said vibrational signal which computing step further comprises computing a short-time energy of said vibrational signal to obtain said envelope.

6. A method as recited in claim 1 wherein said processing step comprises the step of computing of an envelope of said vibrational signal which computing step further comprises:

computing an absolute value of said vibrational signal to produce a rectified form of said vibrational signal; and

passing said rectified vibrational signal through a median filter to obtain said envelope.

7. A method as recited in claim 1 wherein said step of processing of said signature to comprises selecting windows within said signature and selecting a predetermined number of times corresponding to a series of largest amplitudes within each of said windows.

8. A method as recited in claim 1 wherein the processing of said signature to obtain a set of possible events wherein a subset of said set corresponds to a physical action of said machine comprises selecting windows within said signature and selecting a predetermined number of times corresponding to a series of largest amplitudes within each of said window.

9. A method as recited in claim 3 wherein said vibrational signal is passed through a bandpass filter prior to computation of said Hilbert Transform.

10. A method as recited in claim 3 wherein said step of computing of an envelope of said vibrational signal further comprises:

obtaining an unfiltered envelope as a result of said step of taking the square root and the square; and

passing said unfiltered envelope through a low pass filter to obtain said envelope.

11. A method of diagnosing a mechanical condition of a machine, comprising:

sensing a reference vibrational signal of said machine;

creating a reference set of events form said reference vibrational signal of said machine by first creating a reference envelope from said reference vibrational signal and then extracting a set of reference times from said reference envelope;

saving said reference set of times;

sensing a test vibrational signal of said machine;

creating a test set of events from said test vibrational signal of said machine by creating a test envelope from said test vibrational signal and extracting a set of times from said test envelope; and

comparing said test set of times with said reference set of times to determine the mechanical condition of said machine.

12. A method as recited in claim 11 wherein said reference set of events is created by sensing a vibrational signal from said machine when said machine is operating normally and creating a time domain signature from said vibrational signal and processing said signal to obtain a first set of clustered events from a first plurality of sets of possible events including in said first set of clustered events a first subset which corresponds to a physical action of said machine and wherein said test set of events is created by subsequently sensing a test vibrational signal from said machine and creating a time domain signature from said test vibrational signal and processing said signature to obtain a second set of clustered events a second plurality of sets of possible events including in said second set of clustered events a second subset which corresponds to a physical action of said machine.

13. A method as recited in claim 11 wherein said comparing of said test set of times with said reference set of times to determine the mechanical condition of said machine comprises determining if a specific test event time occurs within a predetermined time period on either side of a time at which its corresponding reference event occurs.

14. A method of diagnosing a mechanical condition of a machine as recited in claim 11 wherein said test set of events comprises a first clustered set of events from a plurality of machine operations and said reference set of events comprises a second clustered set of events from a plurality of machine operations.

15. A method of determining a set of clustered events for a machine from a plurality of sets of events, wherein each of said sets has a number of events therein, from said machine comprising:

acquiring a set of vibrational signals where each vibrational signal comprises a plurality of events of said machine;

processing each vibrational signal to obtain a set of events for each vibrational signal and a plurality of sets of events for said machine; and

,

clustering similar events from each of said plurality of sets to determine a set of clustered events for said machine.

16. A method of determining a set of clustered events for a machine from a plurality of sets of events from said machine as recited in claim 15 wherein said clustering of similar events from each of said plurality of sets to determine a set of clustered events for said machine comprises:

determining groupings of events from said plurality of sets of events using various possible orderings of the plurality of sets of events; and

,

determining a set of clustered events from the most common groupings of events.

17. A method of determining a set of clustered events for a machine as recited in claim 16 wherein the determining of groupings of events is accomplished by;

a) selecting a first set from said plurality of sets and assigning each individual event within said set to be a first element of an associated group set, and defining the point in time at which each individual event occurs to be a group center within its associated group set;

b) for each associated group set and its group center, select as its next element an individual event from a next set of said plurality of sets such that said individual event occurs closest in time to said group center with respect to all other events from said next set;

c) determining new group centers for each of said group sets based upon a function of all elements in each group set; and

,

d) repeating steps b) and c) above for all remaining sets of said plurality of sets.

18. A method of diagnosing a mechanical condition of a machine, comprising:

sensing a single reference vibrational signal of said machine;

creating a reference set of events and a reference envelope from said single reference vibrational signal of said machine;

saving said reference set of events and said reference envelope;

sensing a single test vibrational signal of said machine;

creating a test envelope from said test vibrational signal of said machine; and

,

comparing said test envelope with said reference envelope to determine the mechanical condition of said machine.

19. A method of diagnosing a mechanical condition of a machine as recited in claim 18 wherein said comparing is accomplished by:

computing a cross-correlation over a fixed-size region between said test envelope and said reference envelope at each discrete point in time in a predetermined interval about said reference events; and

,

recognizing an event in the test envelope corresponding to each of the reference events when the cross-correlation is below a predetermined value at some discrete time in the predetermined interval, wherein said machine is considered to be operating abnormally when at least a predetermined number of said events is not so recognized.

20. A method of diagnosing a mechanical condition of a machine, comprising:

sensing a reference vibrational signal of said machine;

creating a reference envelope from said reference vibrational signal of said machine;

saving said reference envelope;

sensing a test vibrational signal of said machine;

creating a test envelope from said test vibrational signal of said machine; and

,

comparing said test envelope with said reference envelope to determine the mechanical condition of said machine wherein said comparing comprises determining if at least a predetermined number of resolution ratios computed from said test envelope and said reference envelope exceed a predetermined threshold.

21. An apparatus for extracting a set of events from a vibrational signal of a machine wherein a subset of said events corresponds to a physical action of said machine, comprising:

means for sensing a vibrational signal of said machine;

means for creating a time domain signature from said vibrational signal;

means for processing said signature to obtain a set of clustered events from a plurality of sets of possible events including in said set of clustered events a subset which corresponds to a physical action of said machine; and

means for obtaining from said subset information representing the mechanical condition of said machine.

22. An apparatus for diagnosing a mechanical condition of a machine, comprising:

means for sensing a reference vibrational signal of said machine;

means for creating a reference set of events from said reference vibrational signal of said machine including means for creating a reference envelope from said reference vibrational signal, and means for extracting a set of reference times from said reference envelope;

means for saving said reference set of times;

means for sensing a test vibrational signal of said machine;

means for creating a test set of events from said test vibrational signal of said machine including means for creating a test envelope from said test vibrational signal, and means for extracting a set of reference times from said test envelope; and

means for comparing said test set of times with said reference set of times to determine the mechanical condition of said machine.

23. An apparatus for determining a set of clustered events for a machine from a plurality of sets of events, wherein each of said sets has a number of events therein, from said machine comprising:

means for acquiring a set of vibrational signal where each vibrational signal comprises a plurality of events of said machine;

means for processing each vibrational signal to obtain a set of events for each vibrational signal and a plurality of sets of events for said machine; and

,

means for clustering similar events from each of said plurality of sets to determine a set of clustered events for said machine.

24. An apparatus for diagnosing a mechanical condition of a machine, comprising:

means for sensing a single reference vibrational signal of said machine;

means for creating a single reference set of events and a reference envelope from said reference vibrational signal of said machine;

means for saving said reference set of events and said reference envelope;

means for sensing a single test vibrational signal of said machine;

means for creating a test envelope from said test vibrational signal of said machine; and

,

means for comparing said test envelope with said reference envelope to determine the mechanical condition of said machine.

25. An apparatus for diagnosing a mechanical condition of a machine, comprising:

means for sensing a reference vibrational signal of said machine;

means for creating a reference envelope from said reference vibrational signal of said machine;

means for saving said reference envelope;

means for sensing a test vibrational signal of said machine;

means for creating a test envelope from said test vibrational signal of said machine; and

,

means for comparing said test envelope with said reference envelope to determine the mechanical condition of said machine wherein said comparing comprises determining if at least a predetermined number of resolution ratios computed from said test envelope and said reference envelope exceed a predetermined threshold.