Monitoring slow speed machinery using integrator and selective correction of frequency spectrum

• US 5,646,350 A
• Filed: 01/23/1996
• Issued: 07/08/1997
• Est. Priority Date: 01/23/1996
• Status: Expired due to Term

• Current Assignee: Computational Systems Incorporated (Emerson Electric Company)
• Original Assignee: Computational Systems Incorporated (Emerson Electric Company)
• Inventors: Robinson, James C., Canada, Ronald G.
• Primary Examiner(s): Dombroske, George M.
• Assistant Examiner(s): Noori, Max H.
• Application Number:
  US08/590,340
• Time in Patent Office: 532 Days
• Field of Search: 73/579, 73/599, 73/602, 73/658
• US Class Current: 73/602
• CPC Class Codes: G01M 13/028 Acoustic or vibration analysis
• CPC Class Codes: G01N 2291/2693 Rotor or turbine parts
• CPC Class Codes: G01N 29/12 by measuring frequency or resonance of acoustic waves measuring frequency or resonant frequency of mechanical vibrations or acoustic waves in general G01H1/06, G01H3/04, G01H13/00; acoustic resonators G10K11/04; vibration or shock testing of structures G01M7/00
• CPC Class Codes: G01N 29/40 by amplitude filtering, e.g. by applying a threshold or by gain control
• CPC Class Codes: G01N 29/42 by frequency filtering or by tuning to resonant frequency
• CPC Class Codes: G01N 29/48 by amplitude comparison

Abstract

A method for nonintrusive monitoring of slow speed machinery incorporates a high sensitivity accelerometer capable of detecting low level mechanical vibrations, and a standard data collector/analyzer for processing the accelerometer output in accordance with a slow speed technology (SST) methodology that is implemented by firmware or software. To prevent the high frequency components from overwhelming the low frequency components of the accelerometer output as a result of dynamic range loss during processing of the accelerometer output by the data collector/analyzer, the SST methodology converts the accelerometer output from the acceleration domain to the velocity domain with an analog integrator that amplifies the low frequency components of the accelerometer output while attenuating the high frequency components. The analog integrator distorts the low frequency vibration components as well as the electrical noise introduced by the accelerometer. The distorted signal is then digitized by an A/D converter and transformed to the frequency domain by means of a fast Fourier transform. Within the frequency domain, mechanical vibrations (significant peaks) are separated from the noise and other non-vibration induced transients by fitting the transformed data set to a threshold noise floor that is calculated based on a statistical variance of a most probable noise floor of the accelerometer. All peaks/amplitudes greater than the threshold noise floor are considered significant and retained as indicative of mechanical vibrations occurring within the machinery, while all peaks below the threshold noise floor are considered insignificant and eliminated from further consideration.

Claims

1. A method of processing vibration signals to produce information corresponding to mechanical vibration in slow speed machinery, the method comprising the steps of:

producing a vibration signal measured in acceleration units corresponding to mechanical vibration generated by slow speed machinery, said vibration signal including amplitudes corresponding to frequencies below a pivot frequency, amplitudes corresponding to frequencies above the pivot frequency, and noise;

distorting the vibration signal by analog integration so that amplitudes corresponding to frequencies below the pivot frequency are amplified and amplitudes corresponding to frequencies above the pivot frequency are attenuated, thereby producing a distorted vibration signal measured in velocity units;

digitizing the distorted vibration signal during a predetermined time interval to produce a time series of digital vibration signals measured in velocity units;

producing a first preliminary output in the frequency domain corresponding to the time series of digital vibration signals;

calculating a most probable noise floor;

comparing the amplitudes of the first preliminary output to the most probable noise floor;

eliminating amplitudes in the first preliminary output that are below the most probable noise floor to produce a second preliminary output; and

correcting the second preliminary output by a correction function that inversely corresponds to the distorting of the analog integration to produce a final output corresponding to mechanical vibration generated by the slow speed machinery.

2. The method of claim 1, wherein said calculating step further comprises the steps of:

determining a plurality of minimum amplitudes in a plurality of segments of the first preliminary output;

fitting minimum amplitudes of the first preliminary output with a quadratic function to produce a quadratic fit curve; and

adding one standard deviation to the quadratic fit curve at all frequencies to produce the most probable noise floor.

3. The method of claim 1 wherein said step of producing a first preliminary output comprises the step of converting the time series of digital vibration signals to the frequency domain to produce a distorted vibration spectrum constituting the first preliminary output.

4. The method of claim 3, wherein said converting step further comprises the steps of:

transforming the time series of digital signals to the frequency domain by means of a fast Fourier transform to produce a series of spectra; and

producing a single spectrum from the series of spectra, said single spectrum having a plurality of amplitudes and frequencies with each of said amplitudes corresponding to a unique one of the frequencies, said amplitudes corresponding to mechanical vibration produced by the slow speed machinery.

5. The method of claim 4, further comprising averaging the series of spectra to produce an averaged spectrum as the single spectrum.

6. The method of claim 1, further comprising the step of calculating a threshold noise floor based on a statistical variance of the most probable noise floor.

7. The method of claim 6 wherein said statistical variance is a three sigma statistical variance of the most probable noise floor.

8. The method of claim 1, further comprising the step of eliminating a predetermined range of frequencies and corresponding amplitudes from the distorted vibration signal.

9. The method of claim 1, further comprising the step of eliminating a predetermined range of frequencies and corresponding amplitudes from the vibration signal.

10. The method of claim 1 wherein said distorting step comprises integrating the vibration signal to produce an integrated signal and integrating the integrated signal to produce a double integrated signal constituting the distorted vibration signal as measured in displacement units.