Selective decimation and analysis of oversampled data

• US 9,778,080 B2
• Filed: 04/15/2014
• Issued: 10/03/2017
• Est. Priority Date: 04/29/2013
• Status: Active Grant

• Current Assignee: Emerson Electric (US) Holding Corporation (Chile) Limitada (Emerson Electric Company)
• Original Assignee: Emerson Electric (US) Holding Corporation (Chile) Limitada (Emerson Electric Company)
• Inventors: Garvey, III, Raymond E., Vrba, Joseph A., Bowers, III, Stewart V., Skeirik, Robert D., Holtmannsptter, Hermann, Medley, Michael D., Steele, Kevin, Mann, Douglas A.
• Primary Examiner(s): Le, Toan
• Application Number:
  US14/252,943
• Publication Number:
  US 20140324367A1
• Time in Patent Office: 1,267 Days
• CPC Class Codes: G01D 18/00 Testing or calibrating apparatus or arrangements provided for in groups G01D1/00 - G01D15/00

Abstract

Useful and meaningful machine characteristic information may be derived through analysis of oversampled digital data collected using dynamic signal analyzers, such as vibration analyzers. Such data have generally been discarded in prior art systems. In addition to peak values and decimated values, other oversampled values are used that are associated with characteristics of the machine being monitored and the sensors and circuits that gather the data. This provides more useful information than has previously been derived from oversampled data within a sampling interval.

Claims

1. A method for processing dynamic measurement data derived from an analog signal generated by an analog sensor in sensory contact with a machine or a process, and displaying processed dynamic measurement data on a display device, the method comprising:

(a) generating the analog signal by the analog sensor in sensory contact with the machine or process;

(b) converting the analog signal into an oversampled digital data stream;

(c) designating sampling interval datasets within the oversampled digital data stream;

(d) displaying a measurement selector on the display device to allow a user to select a measurement to be used during a decimation process, wherein the measurement is selected from one or more of a median value, a mode value, a standard deviation value (SDV), a maximum value, a range value, a minimum value, a root mean square (RMS) value, a statistical scatter value, a momentum value, a variance value, a skewness value, a kurtosis value, a peak shape factor (PSF) characteristic, a parametric-versus-causal (PvC) characteristic, and one or more difference values;

(e) receiving a measurement selection based on operation of the measurement selector by the user;

(f) decimating sequential sampling interval datasets based on the measurement selection to produce scalar values corresponding to each sampling interval dataset;

(g) generating a waveform comprising the scalar values; and

(h) displaying the waveform on the display device.

2. The method of claim 1 further comprising analyzing the waveform to provide an indication of one or more of an impacting condition, a friction condition, a faulty signal condition, a noisy signal condition, an arcing condition, a corona condition, a contact condition, an electrical discharge condition, a flow condition, an erosion condition, a cavitation condition, a slipping condition, a valve operation condition, and a valve closure condition.

3. The method of claim 1 further comprising, after step (b) and before step (c), pass filtering the oversampled digital data stream to obtain machine or process condition information.

4. The method of claim 1 further comprising:

(i) determining a median value and a mean value from each of a plurality of the sampling interval datasets, and determining difference values between the median values and the mean values from the plurality of sampling interval datasets;

(j) for one or more of the sampling interval datasets, comparing the difference value to a threshold limit to determine a likelihood of whether the sampling interval dataset comprises Gaussian normal data or not-Gaussian normal data, wherein the difference value beyond the threshold limit indicates a likelihood of not-Gaussian normal data;

(k) assigning a Gaussian attribute or a not-Gaussian attribute in association with the sampling interval dataset; and

(l) storing the Gaussian attribute or not-Gaussian attribute in association with one or more of the sampling interval dataset and the waveform.

5. The method of claim 4 further comprising, in situations in which the comparing step (j) indicates that the dynamic sampling interval dataset is classified as not-Gaussian normal data, further interpreting the dataset characteristics and assigning an attribute indicating a condition that likely resulted in the not-Gaussian normal classification, wherein the condition is selected from the group consisting of an impact, a sensor fault, a circuit fault, a machine operation, a noise, a steady condition, a random event, a systematic event, and a change in environment parameters.

6. The method of claim 4 further comprising, in situations in which the comparing step (j) indicates that the sampling interval dataset comprises not-Gaussian normal data, using the median values determined in step (i) for spectral analysis of the measurement data.

7. The method of claim 4 wherein step (g) comprises generating the waveform using at least one value selected from the sampling interval dataset for which the Gaussian normal or not-Gaussian normal classification is made.

8. The method of claim 1 wherein step (b) comprises oversampling the dynamic measurement data at a sampling rate that is substantially greater than a Nyquist frequency.

9. The method of claim 1 further comprising deriving a characteristic of the machine or a characteristic of the method used to collect the machine vibration data based at least in part on the displayed waveform.

10. The method of claim 1 further comprising:

(i) determining a plurality of standard deviation values, each standard deviation value derived from a corresponding one of the plurality of sampling interval datasets;

(j) determining a plurality of maximum values, each maximum value derived from a corresponding one of the plurality of sampling interval datasets;

(k) determining a plurality of minimum values, each minimum value derived from a corresponding one of the plurality of sampling interval datasets;

(l) determining a plurality of first difference values, each first difference value determined by determining a difference between the maximum value and the minimum value for the corresponding sampling interval dataset;

(m) determining a plurality of second difference values, each second difference value determined by determining a difference between the standard deviation value and the first difference value for the corresponding sampling interval dataset;

(n) comparing one or more of the second difference values to a threshold value to determine a likelihood of whether the dynamic measurement data comprises causal data or Gaussian normal data, wherein second difference values greater than the threshold value indicate a likelihood of causal data; and

(o) in situations in which the comparing step (n) indicates that the dynamic measurement data comprises causal data, assigning a condition that likely resulted in the causal data, wherein the condition is selected from the group consisting of an impact, a sensor fault, a circuit fault, a machine operation, a noise, a steady condition, a random event, a systematic event, and a change in environment parameters.

11. The method of claim 1 further comprising:

step (g) including;

(g1) generating a mid-range waveform comprising a plurality of mid-range values, wherein each of the plurality of mid-range values in the mid-range waveform is selected from a corresponding one of the plurality of sampling interval datasets, wherein the plurality of mid-range values comprise a plurality of median values, a plurality of mean values, a plurality of RMS values or a plurality of mode values;

(g2) generating a maximum-range waveform comprising a plurality of maximum-range values, wherein each of the plurality of maximum-range values in the maximum-range waveform is selected from a corresponding one of the plurality of sampling interval datasets, wherein the plurality of maximum-range values comprise a plurality of absolute maximum values, a plurality of rectified maximum values, a plurality of minimum values, or a plurality of peak-to-peak maximum-to-minimum values;

(g3) generating a statistical scatter waveform comprising a plurality of statistical scatter values, wherein each of the plurality of statistical scatter values in the statistical scatter waveform is selected from a corresponding one of the plurality of sampling interval datasets, wherein the plurality of statistical scatter values comprise a plurality of variance values, a plurality of skewness values, or a plurality of kurtosis values;

(g4) generating a rectified maximum waveform comprising a plurality of rectified maximum values, wherein each of the plurality rectified maximum values in the rectified maximum waveform is selected from a corresponding one of the plurality of sampling interval datasets; and

(g5) generating a combined waveform by

adding values comprising one of the waveforms derived in steps (g1) through (g4) to corresponding values comprising another of the waveforms derived in steps (g1) through (g4), or

subtracting values comprising one of the waveforms derived in steps (g1) through (g4) from corresponding values comprising another of the waveforms derived in steps (g1) through (g4),

wherein the combined waveform represents peaking within the sampling interval datasets.

12. The method of claim 1 further comprising:

(i) determining a plurality of mode values, each mode value based on a most frequently repeating value or value range occurring within a corresponding one of the plurality of sampling interval datasets;

(j) determining a plurality of minimum values, each minimum value derived from a corresponding one of the plurality of sampling interval datasets;

(k) determining a plurality of maximum values, each maximum value derived from a corresponding one of the plurality of sampling interval datasets;

(l) determining a plurality of MODE−

MIN difference values, each MODE−

MIN difference value determined by determining a difference between the mode value and the minimum value for the corresponding sampling interval dataset;

(m) determining a plurality of MAX−

MODE difference values, each MAX−

MODE difference value determined by determining a difference between the maximum value and the mode value for the corresponding sampling interval dataset;

(n) determining that the sensor is faulty if the MODE−

MIN difference value is less than a predetermined threshold value over successive sampling interval datasets; and

(o) determining that the sensor is in a saturated condition if the MAX−

MODE difference value is less than a predetermined threshold value over successive sampling interval datasets.

13. The method of claim 12 further comprising generating a scalar value comprising a first set of bits and a second set of bits, wherein the first set of bits comprise the MODE−

MIN difference value, and the second set of bits comprise a relationship attribute value indicating a sensor faulty state or a sensor not faulty state.

14. The method of claim 12 further comprising generating a scalar value comprising a first set of bits and a second set of bits, wherein the first set of bits comprise the MAX−

MODE difference value, and the second set of bits comprise a relationship attribute value indicating a sensor saturated state or a sensor not saturated state.

15. The method of claim 1 further comprising:

step (c) including organizing the sampling interval datasets by sorting data values from minimum to maximum to form a sorted cumulative data distribution; and

(i) for one or more of the sampling interval datasets, determining a median value to be an average of two or more data values immediately above and below an absolute median of the sorted cumulative data distribution.

16. The method of claim 1 further comprising:

(i) determining a plurality of maximum values, each maximum value derived from a corresponding one of the plurality of sampling interval datasets;

(j) determining one or more kurtosis momentum values based on the plurality of maximum values;

(k) determining a shape factor by subtracting an integer value three from at least one of the kurtosis momentum values;

(l) determining that the oversampled dynamic measurement data has a normal distribution when the shape factor equals zero;

(m) determining that the oversampled dynamic measurement data has a peaked distribution when the shape factor is greater than zero; and

(n) determining that the oversampled dynamic measurement data has a flat distribution when the shape factor is less than zero.

17. The method of claim 1 further comprising:

(i) determining a plurality of mode values, each mode value based on a most frequently repeating value or value range occurring within a corresponding one of the sampling interval datasets;

(j) determining a plurality of median values, each median value derived from a corresponding one of the sampling interval datasets;

(k) determining a plurality of MODE−

MED difference values, each MODE−

MED difference value determined by determining a difference between the mode value and the median value for the corresponding sampling interval dataset; and

(l) determining that a stable measurement condition exists if an absolute value of one or more of the MODE−

MED difference values is less than a predetermined threshold value over successive sampling interval datasets.

18. The method of claim 1 further comprising:

(i) determining a plurality of median values, each median value derived from a corresponding one of the sampling interval datasets;

(j) determining a plurality of maximum values, each maximum value derived from a corresponding one of the sampling interval datasets; and

(k) determining a plurality of crest factors, each crest factor determined based on a difference between the median value and the maximum value for the corresponding sampling interval dataset.

19. The method of claim 1 further comprising:

(i) determining a plurality of maximum values, each maximum value derived from a corresponding one of the sampling interval datasets;

(j) determining a plurality of minimum values, each minimum value derived from a corresponding one of the sampling interval datasets;

(k) determining a plurality of MAX−

MIN difference values, each MAX−

MIN difference value determined by determining a difference between the maximum value and the minimum value for the corresponding sampling interval dataset;

(l) determining a plurality of standard deviation values, each standard deviation value derived from a corresponding one of the sampling interval datasets;

(m) determining a plurality of SDV difference values, each SDV difference value determined by determining a difference between the standard deviation value and the MAX−

MIN difference value for the corresponding sampling interval dataset; and

(n) comparing one or more of the SDV difference values to a threshold value to determine a likelihood of whether the dynamic measurement data comprises causal data or Gaussian normal data, wherein difference values greater than the threshold value indicate a likelihood of causal data.

20. The method of claim 1 further comprising:

(i) determining a plurality of maximum values, each maximum value derived from a corresponding one of the sampling interval datasets;

(j) determining three or more values immediately before the maximum value for one or more of the sampling interval datasets;

(k) determining three or more values immediately after the maximum value for one or more of the sampling interval datasets;

(l) for one or more of the sampling interval datasets, determining a peak shape factor characteristic based on the maximum value and the three or more values immediately before and the three or more values immediately after the maximum value; and

(m) based on the peak shape factor characteristic, determining a likely causal event associated with the maximum value.

21. The method of claim 1 further comprising:

(i) determining a plurality of maximum values, each maximum value derived from a corresponding one of the sampling interval datasets;

(j) determining a plurality of standard deviation values, each standard deviation value derived from a corresponding one of the sampling interval datasets;

(k) determining a plurality of parametric-versus-causal characteristics, each parametric-versus-causal characteristic derived from a corresponding one of the sampling interval datasets;

(l) determining a plurality of peak shape factor characteristics, each peak shape factor characteristic derived from a corresponding one of the sampling interval datasets; and

(m) based on the maximum values, standard deviation values, parametric-versus-causal characteristics, and peak shape factor characteristics, determining whether one or more of the following conditions exist;

a spall condition exists resulting from roller bearing component fatigue; and

a broken tooth condition exists resulting from fatigue failure of a gear component.

22. The method of claim 1 further comprising:

(i) determining a plurality of minimum values, each minimum value derived from a corresponding one of the sampling interval datasets;

(j) determining a plurality of median values, each median value derived from a corresponding one of the sampling interval datasets;

(k) determining a plurality of mode values, each mode value derived from a corresponding one of the sampling interval datasets;

(l) determining a plurality of mean values, each mean value derived from a corresponding one of the sampling interval datasets;

(m) determining a plurality of standard deviation values, each standard deviation value derived from a corresponding one of the sampling interval datasets;

(n) determining a plurality of peak shape factor characteristics, each peak shape factor characteristic derived from a corresponding one of the sampling interval datasets; and

(o) based on the minimum values, median values, mode values, mean values, standard deviation values, and peak shape factor characteristics, determining whether a sliding friction condition exists resulting from inadequate lubrication.

23. The method of claim 1 further comprising:

(i) determining a plurality of median values, each median value derived from a corresponding one of the sampling interval datasets;

(j) determining a plurality of mode values, each mode value derived from a corresponding one of the sampling interval datasets;

(k) determining a plurality of mean values, each mean value derived from a corresponding one of the sampling interval datasets;

(l) determining a plurality of standard deviation values, each standard deviation value derived from a corresponding one of the sampling interval datasets;

(m) determining a plurality of peak shape factor characteristics, each peak shape factor characteristic derived from a corresponding one of the sampling interval datasets; and

(n) based on the median values, mode values, mean values, standard deviation values, and peak shape factor characteristics, determining whether a smooth running condition exists resulting from proper lubrication.

24. The method of claim 1 further comprising:

(i) determining a plurality of median values, each median value derived from a corresponding one of the sampling interval datasets;

(j) determining a plurality of mean values, each mean value derived from a corresponding one of the sampling interval datasets;

(k) determining a plurality of standard deviation values, each standard deviation value derived from a corresponding one of the sampling interval datasets;

(l) determining a plurality of peak shape factor characteristics, each peak shape factor characteristic derived from a corresponding one of the sampling interval datasets; and

(m) based on the median values, mean values, standard deviation values, and peak shape factor characteristics, determining whether a misalignment condition exists.

25. The method of claim 1 further comprising:

(i) determining a plurality of maximum values, each maximum value derived from a corresponding one of the sampling interval datasets;

(j) determining a plurality of standard deviation values, each standard deviation value derived from a corresponding one of the sampling interval datasets;

(k) determining a plurality of parametric-versus-causal characteristics, each parametric-versus-causal characteristic derived from a corresponding one of the sampling interval datasets;

(l) determining a plurality of peak shape factor characteristics, each peak shape factor characteristic derived from a corresponding one of the sampling interval datasets; and

(m) based on the maximum values, standard deviation values, parametric-versus-causal characteristics, and peak shape factor characteristics, determining whether subsurface fatigue cracks exist resulting from resonant vibration of a thermowell in process piping.

26. The method of claim 1 further comprising:

(i) determining a plurality of maximum values, each maximum value derived from a corresponding one of the sampling interval datasets;

(j) determining a plurality of mean values, each mean value derived from a corresponding one of the sampling interval datasets;

(k) determining a plurality of standard deviation values, each standard deviation value derived from a corresponding one of the sampling interval datasets;

(l) determining a plurality of peak shape factor characteristics, each peak shape factor characteristic derived from a corresponding one of the sampling interval datasets; and

(m) based on the maximum values, mean values, standard deviation values, and peak shape factor characteristics, determining whether slip-stick movements are occurring resulting from a coefficient of static friction being exceeded at a loaded interface.

27. The method of claim 1 further comprising:

(i) determining at least one of;

a plurality of minimum values, each minimum value derived from a corresponding one of the sampling interval datasets;

a plurality of median values, each median value derived from a corresponding one of the sampling interval datasets;

a plurality of mode values, each mode value derived from a corresponding one of the sampling interval datasets;

a plurality of standard deviation values, each standard deviation value derived from a corresponding one of the sampling interval datasets; and

a plurality of peak shape factor characteristics, each peak shape factor characteristic derived from a corresponding one of the sampling interval datasets; and

(j) based at least in part on the values determined in step (i), determining whether one or more of the following conditions exist;

occurrence of a partial discharge in vicinity of a high-voltage electrical apparatus; and

occurrence of a leaking condition that produces fluid turbulence in vicinity of a leaky orifice under pressure.

28. The method of claim 1 further comprising:

(i) determining at least one of;

a plurality of median values, each median value derived from a corresponding one of the sampling interval datasets;

a plurality of mode values, each mode value derived from a corresponding one of the sampling interval datasets;

a plurality of standard deviation values, each standard deviation value derived from a corresponding one of the sampling interval datasets; and

a plurality of peak shape factor characteristics, each peak shape factor characteristic derived from a corresponding one of the sampling interval datasets; and

(j) based on the values determined in step (i), determining whether an intermittent fault condition exists in a three-phase electrical power line.

29. The method of claim 1 further comprising:

step (f) including, for one or more of the sampling interval datasets, generating multiple scalar values representative of the dynamic measurement data within the dataset;

(i) based on the multiple scalar values, determining one or more characteristic values indicative of a feature, quality or characteristic of the dynamic measurement data within the dataset; and

(j) saving the multiple scalar values and the one or more characteristic values in association with an identifier for the dataset.

30. The method of claim 1 further comprising:

(i) determining a plurality of first statistical scalar values, each first statistical scalar value derived from a corresponding one of the plurality of sampling interval datasets;

(j) determining a plurality of second statistical scalar values, each second statistical scalar value derived from a corresponding one of the plurality of sampling interval datasets;

(k) determining based on one or more of the first statistical scalar values that the machine or process is in a first state and not in a second state; and

(l) determining based on one or more of the second statistical scalar values that the machine or process is in the second state and not in the first state.

31. A method for collecting and processing oversampled vibration data collected by at least first and second vibration sensors attached to a mechanical structure used in the processing of a material, wherein the mechanical structure is operable to transmit vibrational energy from the material to the first and second vibration sensors, wherein the oversampled vibration data comprises a plurality of sampling interval datasets, wherein each sampling interval dataset corresponds to a particular sampling interval, the method comprising:

(a) positioning the first vibration sensor in a first location on the mechanical structure and positioning the second vibration sensor in a second location on the mechanical structure;

(b) receiving vibrational energy at the first vibration sensor, wherein the vibrational energy was generated by an event that occurs within the material being processed and traveled through the mechanical structure to the first vibration sensor;

(c) the first vibration sensor generating a first vibration signal based on the vibrational energy;

(d) oversampling the first vibration signal to generate first oversampled vibration data comprising a plurality of first sampling interval datasets;

(e) for each of the plurality of first sampling interval datasets, determining one or more first scalar values selected from the group consisting of a maximum value, a minimum value, a mean value, a median value, a mode value, a standard deviation value, a maximum-to-minimum range value, a kurtosis value, a skewness value, and a wavelength value;

(f) based on the one or more first scalar values, determining one or more first characteristic values that provide an indication of an event type;

(g) generating a first timestamp value representative of a time at which the vibrational energy generated by the event was received at the first vibration sensor;

(h) receiving the vibrational energy at the second vibration sensor, wherein the vibrational energy traveled through the mechanical structure to the second vibration sensor;

(i) the second vibration sensor generating a second vibration signal based on the vibrational energy;

(j) oversampling the second vibration signal to generate second oversampled vibration data comprising a plurality of second sampling interval datasets;

(k) for each of the plurality of second sampling interval datasets, determining one or more second scalar values selected from the group consisting of a maximum value, a minimum value, a mean value, a median value, a standard deviation value, a maximum-to-minimum range value, a kurtosis value, a skewness value, and a wavelength value;

(l) based on the one or more second scalar values, determining one or more second characteristic values that provide an indication of the event type;

(m) generating a second timestamp value representative of a time at which the vibrational energy generated by the event was received at the second vibration sensor;

(n) comparing the one or more first characteristic values to the one or more second characteristic values to determine that the event type indicated by the one or more first characteristic values is the same event type as indicated by the one or more second characteristic values;

(o) determining a time difference between the first timestamp and the second timestamp; and

(p) based at least in part on the time difference, determining a location of the event relative to the first vibration sensor and the second vibration sensor.

32. The method of claim 31 wherein the mechanical structure comprises one or more of a pipe, a vessel, a mill, a crusher, a grinder, and a metal cutting tool.

33. The method of claim 31 wherein the event is selected from the group consisting of an impact event, a recoil event, a cavitation event, a void formation event, a material movement event, a dry contact friction event, a boundary lubrication friction event, a fluid leak event, a fluid turbulence event, a surface erosion event, a chatter event, a resonance event, and a buckling event.

34. An apparatus for collecting and processing machine or process vibration data comprising:

at least one vibration sensor attached to a machine, the at least one vibration sensor generating at least one analog vibration signal having a maximum frequency of interest, F_MAX, which is greater than an event frequency of events occurring in the machine or the process;

at least one analog-to-digital converter for oversampling the at least one analog vibration signal at a sampling rate of at least seven times F_MAX to generate a plurality of sampling interval datasets, each corresponding to a particular sampling interval;

a decimation module having a plurality of parallel field programmable gate arrays comprising;

a first field programmable gate array for receiving the plurality of sampling interval datasets and determining a first scalar value from each sampling interval dataset, the first scalar selected from the group consisting of a maximum value, a minimum value, a median value, a mode value, a mean value, a standard deviation value, a parametric-versus-causal value, an operational condition value, and a peak shape factor value; and

a second field programmable gate array for receiving the plurality of sampling interval datasets and determining a second scalar value from each sampling interval dataset that is different from the first scalar value, the second scalar selected from the group consisting of a maximum value, a minimum value, a median value, a mode value, a mean value, a standard deviation value, a parametric-versus-causal value, an operational condition value, and a peak shape factor value.