Method and apparatus for performing pre-emptive maintenance on operating equipment

• US 5,710,723 A
• Filed: 04/05/1995
• Issued: 01/20/1998
• Est. Priority Date: 04/05/1995
• Status: Expired due to Fees

• Current Assignee: Dayton T. Brown
• Original Assignee: Dayton T. Brown
• Inventors: Morrelly, C. Kenneth, Hoth, Donald K.
• Primary Examiner(s): Trammell, James P.
• Application Number:
  US08/417,116
• Time in Patent Office: 1,021 Days
• Field of Search: 364/551.01, 364/551.02, 364/552, 364/554, 364/557, 364/558, 364/571.03, 364/578, 364/508, 395/900, 395/902-907, 395/911-915
• US Class Current: 702/181
• CPC Class Codes: G01H 1/003 of rotating machines G01H1/10 takes precedence
• CPC Class Codes: G05B 23/024 Quantitative history assessment, e.g. mathematical relationships between available data; Functions therefor; Principal component analysis [PCA]; Partial least square [PLS]; Statistical classifiers, e.g. Bayesian networks, linear regression or correlation analysis; Neural networks
• CPC Class Codes: G05B 23/0297 Reconfiguration of monitoring system, e.g. use of virtual sensors; change monitoring method as a response to monitoring results

Abstract

A pre-emptive maintenance system for performing maintenance and process assurance on run-critical equipment employs an intrinsic health monitor that includes a number of sensors that are used with an operating equipment or machine to generate a set of intrinsic physical signatures that are products of the primary performance characteristics of the operating equipment. These signatures are nonfunctional parametrics that are consistent and reliable indicators of the normal operation of the equipment and statistical norms are set during an initial learning mode using neural processors and the outputs from the sensors. Thereafter, the sensor outputs are monitored and analyzed using the statistical norms to predict the probabilities of future failures of the operating equipment. By using the intrinsic health monitor an advance in the field of equipment maintenance and process assurance is provided in the approach to maintenance being a pre-emptive procedure.

Claims

1. A system for providing future failure probability information for operating equipment, comprising:

a plurality of sensors arranged proximate the operating equipment and including a primary intrinsic parameter sensor, a dependent operational load parameter sensor, a dependent operational parameter sensor, and an independent operational parameter sensor;

monitoring means located at a same location as the operating equipment and connected to receive data signals produced by said plurality of sensors and including a control processor operating an artificial intelligence algorithm for processing said data signals initially according to a learning procedure and subsequently according to a monitoring procedure whereby energy contents in discrete portions in the data signals produced by said plurality of sensors are evaluated and producing evaluation signals; and

work station means connected to receive said evaluation signals from said monitoring means and including a display for providing a visual display representing said evaluation signals.

2. The system according to claim 1, wherein said work station means is located at a different location than the operating equipment.

3. The system according to claim 1, wherein said work station means is connected to said monitoring means by one of commercial telephone lines, local area network, wide area network, cellular network, satellite communications, or direct wire.

4. The system according to claim 1, wherein said monitoring means includes a second display located at the same location as the operating equipment for providing a visual display representing said evaluation signals.

5. The system according to claim 1, wherein said monitoring means includes:

a reference power supply, and a signal interface unit connected to said primary intrinsic parameter sensor and to said reference power supply for referencing a data signal from said primary intrinsic parameter sensor to a reference voltage from said reference power supply.

6. The system according to claim 5, wherein said monitoring means further comprises a programmable attenuator/amplifier receiving an analog output signal from said signal interface unit and a digital-to-analog converter for converting an output signal from said programmable attenuator/amplifier to a digital signal.

7. The system according to claim 5, wherein said signal interface means provides a digital signal based on the data signal from said primary intrinsic parameter sensor, wherein said monitoring means further comprises a digital system bus connected to said digital signal and wherein said control processor is connected to said system bus for receiving said digital signal and performing a statistical analysis using said artificial intelligence algorithm to produce said evaluation signals.

8. The system according to claim 1, wherein said monitoring means further comprises a a signal interface unit having a digital-to-analog converter for converting at least a data signal from said primary intrinsic parameter sensor to digital data for processing by said control processor.

9. The system according to claim 8, wherein said signal interface unit also produces a digital signal from said vibration sensor and said digital signal is connected to said system bus.

10. The system according to claim 8, wherein said monitoring means further comprises a neural processor connected to said control processor for performing said learning procedure using statistical analysis to establish a set of stable relationships between data from said primary intrinsic parameter sensor and data from said dependent operational load sensor, said dependent operational parameter sensor, and said independent operational parameter sensor, respectively.

11. The system according to claim 8, wherein said control processor includes means for performing a summary operating equipment failure probability based on fuzzy logic.

12. The system according to claim 11, wherein said control processor operates according to an inference engine.

13. The system according to claim 1, wherein said monitoring means includes a programmable bandpass filter receiving an analog data signal from said primary intrinsic parameter sensor for segmenting the data signal into said discrete portions and producing respective outputs for said discrete portions, said outputs being used in said learning procedure for acquisition of a base energy distribution and said outputs being used in said monitoring procedure for acquisition of a comparative energy distribution.

14. The system according to claim 1, wherein said primary intrinsic parameter sensor includes a vibration sensor.

15. A method of evaluating operating equipment to determine a maintenance requirement, comprising the steps of:

sensing a plurality of operating parameters of said operating equipment, including sensing ambient temperature, equipment operating load, and vibration and producing respective data signals;

performing a learning operation on the data signals using statistical processing;

deriving a plurality of operating norms based on results of said statistical processing;

utilizing statistical inference and fuzzy logic for analyzing the data signals based on the derived plurality of operating norms and producing a failure probability conclusion;

transmitting the failure probability conclusion to a location remote from the operating equipment; and

displaying the transmitted failure probability conclusion.

16. The method according to claim 15, wherein the step of transmitting includes transmitting the failure probability conclusion over one of commercial telephone lines, local area network, wide area network, cellular network, or satellite communications.

17. The method according to claim 15, comprising the further step of digitally converting the failure probability conclusion prior to the step of transmitting.

18. The method according to claim 15, comprising the further step of performing anti-aliasing filtering on the data signals before the step of utilizing statistical inference and fuzzy logic.

19. The method according to claim 15, wherein the step of displaying includes determining within which one of a plurality of quantitative ranges the transmitted failure probability conclusion falls and displaying each quantitative range in a respective different color.

20. A method of determining maintenance requirements for an operating system including at least one machine, comprising the steps of:

obtaining operating data from the system during operation including obtaining vibration data, operating load data, and ambient temperature data;

forming obtained data into a raw data table;

repeating the steps of obtaining and forming over a first set period of time to produce hourly data;

storing the produced hourly data;

performing learning mode calculations from the stored hourly data, including calculating upper and lower confidence intervals using stored hourly data representing a second set period of time;

determining data values within the upper and lower confidence levels evidencing correct system operation;

performing monitoring mode calculations by comparing present raw data with the data values evidencing correct system operation; and

displaying to a user results of the step of comparing.

21. The method according to claim 20, including the further step of checking the operating load data for stability and setting a stability flag in the raw data if successive quantities of the operating load data vary one from another more than a predetermined amount.

22. The method according to claim 21, including the further step of checking the raw data for the presence of the stability flag and, upon the presence of the stability flag, excluding the raw data having the stability flag from the produced hourly data.

23. The method according to claim 20, wherein the first set period of time is set to one hour and the second set period of time is set to twelve hours.

24. The method according to claim 21, wherein the step of obtaining vibration data includes obtaining vibration data over a plurality of discrete frequency bands.

25. The method according to claim 20, including the further step of checking whether said at least one machine is on and setting an off flag in the raw data if it is determined that the at least one machine is not on.

26. The method according to claim 25, including the further step of checking for the presence of the off flag and, upon the presence with the off flag, excluding the raw data having the off flag from the produced hourly data.

27. A system for providing future failure probability information for operating equipment, the system comprising:

a plurality of sensors arranged proximate the operating equipment and including a primary intrinsic parameter sensor, a dependent operational load parameter sensor, and an independent operational parameter sensor; and

monitoring means located where the operating equipment is located and being connected to receive data signals produced by said plurality of sensors and including a control processor operating an artificial intelligence algorithm consisting of a learning process and a monitoring process and for producing future failure monitoring signals from said data signals.

28. The system according to claim 27, wherein said monitoring means includes a programmable bandpass filter receiving an analog data signal from said primary intrinsic parameter sensor for segmenting energy contained in said analog data signal into discrete portions for use in obtaining a first energy distribution in said learning process and a comparative, second energy distribution in said monitoring process.