Automated fault diagnosis device and method

• US 7,010,445 B2
• Filed: 06/13/2003
• Issued: 03/07/2006
• Est. Priority Date: 01/23/2002
• Status: Expired due to Term

• Current Assignee: Computational Systems Incorporated (Emerson Electric Company)
• Original Assignee: CSI Technology Incorporated (Emerson Electric Company)
• Inventors: Robinson, James C., Van Voorhis, J. Brent, Hillard, Jason E., Battenberg, Rexford A.
• Primary Examiner(s): Hoff, Marc S.
• Assistant Examiner(s): SUAREZ, FELIX E
• Application Number:
  US10/460,967
• Publication Number:
  US 20030209077A1
• Time in Patent Office: 998 Days
• Field of Search: 702/39, 702/54, 702/71, 702/75, 702/76, 702180-183, 702/66, 702/59, 702/60, 702/189, 702/190, 235/454, 370/319, 717/138
• US Class Current: 702/77
• CPC Class Codes: G01N 2291/02881 Temperature
• CPC Class Codes: G01P 3/36 Devices characterised by the use of optical means, e.g. using infrared, visible, or ultraviolet light G01P3/68 takes precedence; gyrometers using the Sagnac effect, i.e. rotation-induced shifts between counter-rotating electromagnetic beams G01C19/64

Abstract

A diagnostic apparatus and method for sensing a signal emanating from a machine and determining a periodic machine event based on the emanated signal. The apparatus includes analog to digital processing components for digitizing the sensed signal, producing a digitized signal. The apparatus also includes a transform algorithm, memory, an analysis algorithm, and a display. The transform algorithm transforms the digitized signal to generate power spectral density and autocorrelation data which is stored in the memory. The analysis algorithm includes criteria for analyzing the stored power spectral density and autocorrelation data to determine a period machine event based at least in part upon the autocorrelation data. The display displays the periodic machine event to a user. In the event of a suspected defect or fault, the device alerts the user. The user may then elect to proceed with repairs or undertake advanced analysis of the suspected fault as economy dictates.

Claims

1. A method of diagnosing a fault in a machine by determining a periodic event related to the fault in the machine, the method comprising:

(a) sensing vibration signals emanating from the machine,

(b) transforming the vibration signals to produce first transformed signals,

(c) transforming the first transformed signals to produce second transformed signals having at least real components, and

(d) processing the first and second transformed signals using select criteria to determine the periodic event.

2. The method of claim 1 wherein:

step (b) further comprises using a Fourier Transform function to transform vibration signals to produce the first transformed signals which comprise power spectral density data, and

step (c) further comprises using a Fourier Transform function to transform the first transformed signals to produce the second transformed signals which comprise autocorrelation data.

3. The method of claim 2 wherein step (d) further comprises locating one or more peaks in at least one of the power spectral density data and the autocorrelation data.

4. The method of claim 3 wherein step (d) further comprises performing peak comparisons using at least one peak in at least one of the power spectral density data and the autocorrelation data to determine a periodic event.

5. The method of claim 4 wherein step (d) further comprises applying select criteria to the comparisons to determine (1) whether a peak in the autocorrelation data is a fundamental frequency peak, (2) whether the peak in the autocorrelation data is within a number of bins of an exact frequency required, and (3) whether there are a select number of harmonics of the fundamental frequency peak.

6. The method of claim 3 wherein step (d) further comprises:

(d1) searching side bands of one or more peaks in the autocorrelation data based on a select peak frequency in the power spectral density data, and

(d2) determining a periodic event based at least in part on the searched side bands.

7. The method of claim 3 wherein step (d) further comprises:

(d1) limiting a number of peaks in the autocorrelation data,

(d2) comparing at least one peak in the limited autocorrelation data with at least one peak in the power spectral density data, and

(d3) determining the periodic event based at least in part on a result of the comparison.

8. The method of claim 3 wherein step (d) further comprises:

(d1) searching for a peak in the power spectral density data which represents a periodic event and

(d2) verifying that the peak represents a periodic event based at least in part on the autocorrelation data.

9. The method of claim 1 wherein step (d) further comprises:

(d1) examining the real components of the second transformed signals with a peak location algorithm, and

(d2) selecting a number of the real components of the second transformed signals when;

(1) a peak is not located in at least one of a first set of frequency bins;

(2) a first peak at a first peak frequency has an amplitude greater than a number multiplied by a largest amplitude peak in a selected bin; and

(3) a second peak is located at a second peak frequency which is related to the first peak frequency.

10. The method of claim 1 wherein step (d) further comprises:

(d1) storing peak magnitude data of the first transformed signals in a number of frequency bins,

(d2) determining a number of largest magnitude peaks from the stored peak magnitude data with a peak location algorithm,

(d3) locating a first peak within the number of largest magnitude peaks which corresponds to a periodic event of the system, and

(d4) determining whether the first peak satisfies select criteria.

11. The method of claim 10 wherein step (d) further comprises:

(d5) determining whether the first peak at a first frequency is located in a first frequency bin of a first set of frequency bins,

(d6) determining whether an amplitude of the first peak is greater than a number multiplied by a largest amplitude peak located in a second frequency bin of the first set of frequency bins,

(d7) determining whether a second peak is located at a second frequency which is about two times the first frequency of the first peak and which second frequency is located in a second set of frequency bins, and whether an amplitude of the second peak is greater than a number multiplied by the largest amplitude peak located in the second frequency bin, and

(d8) determining the periodic event based at least part on the first peak.

12. The method of claim 11 wherein step (d) further comprises:

(d9) determining whether a third peak at a third frequency has an amplitude greater than about 0.90 multiplied by the amplitude of the first peak,

(d10) determining whether a fourth peak at a fourth frequency which is about two times the third frequency has an amplitude greater than about 0.90 multiplied by the amplitude of the first peak, and

(d11) determining the periodic event based at least in part on the third peak.

13. A device for diagnosing a fault in a machine by determining a periodic event of the machine, the device comprising:

a sensor for sensing a vibration signal emanating from the machine, the sensor operable to convert the sensed vibration signal into an electrical signal,

signal processing components for digitizing the electrical signal to produce a digitized signal and for processing the digitized signal, and

firmware including at least one transform algorithm for successively transforming the digitized signal to produce first and second sets of transform data, where the second set of transform data is a transform of the first set of transform data, and at least one analysis algorithm for analyzing the first and second sets of transform data to determine the periodic event based on select criteria applied to the first and second sets of transform data.

14. The device of claim 13 wherein the at least one transform algorithm further comprises a Fourier Transform algorithm for transforming the digitized signal to produce power spectral density data.

15. The device of claim 14 wherein the firmware further comprises algorithmic structure for applying the Fourier Transform algorithm to the power spectral density data to produce autocorrelation data and for conditioning the autocorrelation data to produce conditioned autocorrelation data.

16. The device of claim 15 wherein the at least one analysis algorithm further comprises a peak fit algorithm for fitting a curve to the power spectral data and the conditioned autocorrelation data, and a peak location algorithm for locating one or more peaks in the power spectral data and the conditioned autocorrelation data.

17. The device of claim 16 wherein the at least one analysis algorithm further comprises a sideband search algorithm for searching side bands of one or more peaks in the conditioned auto correlation data based at least in part on a select peak frequency of the power spectral density data, and for determining a periodic event based at least in part on the searched side bands.

18. The device of claim 16 wherein the at least one analysis algorithm further comprises a peak comparison algorithm for comparing at least one peak in the conditioned autocorrelation data with at least one peak in the power spectral density data, wherein search criteria applied to a result of the peak comparison determines whether a periodic event exists.

19. The device of claim 18 wherein the search criteria further comprises instructions for determining whether the at least one peak in the conditioned autocorrelation data is a fundamental frequency peak, whether the at least one peak in the conditioned autocorrelation data is within a number of bins of an exact frequency required, and whether there are a select number of harmonics of the fundamental frequency peak.

20. The device of claim 16 wherein the at least one analysis algorithm further comprises a peak comparison algorithm for limiting a number of peaks in the conditioned autocorrelation data and for comparing at least one peak in the limited autocorrelation data with at least one peak in the power spectral density data, wherein a result of the comparison is used to determine whether a periodic event exists.

21. The device of claim 16 wherein the at least one analysis algorithm further comprises a peak search algorithm for searching for a peak in the power spectral density data which represents a periodic event and for verifying that the peak in the power spectral density data corresponds to a periodic event based at least in part on the conditioned autocorrelation data.

22. An apparatus for diagnosing a fault in a machine, the apparatus comprising:

a sensor for sensing a vibration signal emanating from the machine,

analog to digital processing components for digitizing the sensed vibration signal to produce a digitized signal,

at least one transform algorithm for transforming the digitized signal to generate power spectral density data and autocorrelation data from the digitized signal,

memory for storing the power spectral density data and auto correlation data,

at least one analysis algorithm for analyzing the power spectral density and autocorrelation data, the analysis algorithm including criteria for determining a periodic event of the machine based at least in part on the autocorrelation data, and a display for displaying the periodic event to a user.

23. The apparatus of claim 22 wherein the at least one transform algorithm further comprises a Fourier Transform function for producing the power spectral density data when the Fourier Transform function is first applied to the digitized signal and for producing autocorrelation data when the Fourier Transform function is thereafter applied tote power spectral density data, and wherein the apparatus determines the periodic event based at least in part on one or more peaks in at least one of the power spectral density data and the autocorrelation data.