This disclosure relates generally to monitoring systems and more specifically to an apparatus and method for identifying health indicators for rolling element bearings.
Rolling element bearings are one of the most common and frequently failing components in rotating machinery. A rolling element bearing often includes the following four subcomponents: bearing balls, a cage, an outer race, and an inner race. The bearing balls typically move during rotation of a machinery element, and the cage typically helps to maintain separation of the balls. Each race is typically a grooved ring, and the bearing balls move between the races.
A crack, indentation, or wear in any one of these subcomponents often requires replacement of the entire rolling element bearing. Not only that, the failure of a rolling element bearing often causes immense secondary damage, such as shaft misalignment, rub, stator insulation failure, and gear failure. Therefore, the monitoring of the condition of a rolling element bearing is often an important activity in many industries.
This disclosure provides an apparatus and method for identifying health indicators for rolling element bearings.
In a first embodiment, a method includes receiving an input signal containing information associated with a rolling element bearing and/or a piece of equipment containing the rolling element bearing. The method also includes decomposing the input signal into a frequency-domain signal and determining at least one family of frequencies corresponding to at least one failure mode of the rolling element bearing. The method further includes generating a reconstructed input signal using the at least one family of frequencies and the frequency-domain signal. In addition, the method includes determining, using the reconstructed input signal, an indicator identifying an overall health of the rolling element bearing.
In a second embodiment, an apparatus includes at least one interface configured to receive an input signal containing information associated with a rolling element bearing and/or a piece of equipment containing the rolling element bearing. The apparatus also includes at least one processing device configured to decompose the input signal into a frequency-domain signal and to determine a family of frequencies corresponding to a failure mode of the rolling element bearing. The at least one processing device is also configured to generate a reconstructed input signal using the family of frequencies and the frequency-domain signal and to determine, using the reconstructed input signal, a value identifying a likelihood that the rolling element bearing is suffering from a defect associated with the failure mode.
In a third embodiment, a computer readable medium embodies a computer program. The computer program includes computer readable program code for decomposing an input signal to produce a frequency-domain signal, where the input signal contains information associated with a rolling element bearing and/or a piece of equipment containing the rolling element bearing. The computer program also includes computer readable program code for determining multiple families of frequencies corresponding to multiple failure modes of the rolling element bearing. The computer program further includes computer readable program code for generating multiple reconstructed input signals, where each reconstructed input signal is generated using at least one family of frequencies and the frequency-domain signal. In addition, the computer program includes computer readable program code for determining, using the reconstructed input signals, (i) an indicator identifying an overall health of the rolling element bearing and (ii) values identifying likelihoods that the rolling element bearing is suffering from defects associated with the failure modes.
Other technical features may be readily apparent to one skilled in the art from the following figures, descriptions, and claims.
For a more complete understanding of this disclosure, reference is now made to the following description, taken in conjunction with the accompanying drawings, in which:
The data collector 106 represents any suitable device, system, or other component for collecting data associated with the equipment 102 or its rolling element bearings 104. For example, the data collector 106 could include a vibration sensor that takes vibration measurements of the equipment 102 or its rolling element bearings 104. The data collector 106 could also include a speed sensor that takes speed measurements of the rate at which an element in the equipment 102 is rotating. The data collector 106 could further include one or more analog-to-digital converters for converting any analog measurements into digital measurements. Any other or additional types of measurements of the equipment 102 or its rolling element bearings 104 could be obtained by the data collector 106.
The processing system 108 represents any suitable processing or computing device or system for detecting problems with one or more rolling element bearings 104 based on data from one or more data collectors 106. The processing system 108 could, for example, include a processor, microprocessor, microcontroller, field programmable gate array, digital signal processor, or other processing or control device 112. The processing system 108 could also include one or more memories 114 storing instructions and data used, generated, or collected by the processing system 108. The processing system 108 could further include at least one interface 116 facilitating communications with the data collector 106, the display 110, or other external component(s). The at least one interface 116 could include an Ethernet or other wired network interface or an IEEE 802.11, BLUETOOTH, or other wireless interface.
The display 110 represents any suitable display for presenting information to a user, such as a liquid crystal display (LCD), light emitting diode (LED) display, or cathode ray tube (CRT) display. Any suitable information could be presented to the user. This could include notifications or alarms when problems with the rolling element bearings 104 are detected, reminders that problems with the rolling element bearings 104 were previously detected, and analysis results generated by the processing system 108. The display 110 could represent a stand-alone display or a display in a larger device or system, such as a display in a control room, a display in or coupled to a computing device, or a display in a mobile telephone or other mobile device.
Any suitable connections could be used between the components shown in
In one aspect of operation, the processing system 108 identifies one or more health indicators for one or more rolling element bearings 104. More specifically, the processing system 108 monitors the health of one or more rolling element bearings 104 using signals from the data collectors 106, such as vibration measurements. This technique can be used in a number of industries with a variety of assets or machines that have rolling element bearings, such as pumps, compressors, and turbines. Additional details regarding the specific operations performed by the processing system 108 to monitor the health of and identify problems with one or more rolling element bearings 104 are provided below.
Although
In this example, the bearing health monitoring device 200 receives various input data, such as baseline data 202 associated with normal operation of a rolling element bearing 104 or with a defective bearing 104 (where the defective condition will increase in severity over time). The bearing health monitoring device 200 also receives current or real-time sensor data 204 from one or more data collectors 106′, such as vibration data from an accelerometer or other vibration sensor or speed data from a speed sensor. Any other suitable data could be collected, such as audible noise, voltage readings, or electrical current flowing through a component. The sensor data 204 could represent any suitable data having any suitable bandwidth(s), such as acceleration data having a bandwidth of 0-10 kHz or Once Per Revolution (OPR) data having a bandwidth of 0-5 kHz. The OPR data represents data collected during each revolution of the rolling element bearing 104. The bearing health monitoring device 200 further receives a mathematical model 206 defining the signal patterns associated with different types of faults in a rolling element bearing 104.
At least some of this data is filtered in a pre-processing filter 208, which in this example represents a low-pass filter. Any suitable low-pass filter could be used, such as a Butterworth filter or wavelet-based filters using various wavelets. The filtering helps to produce a signal of a particular frequency bandwidth (such as 0-1 kHz or 0-2 kHz). In particular embodiments, an operator can select the frequency bandwidth depending upon his or her expert knowledge of the system being monitored.
The data is analyzed using a processor core 210 to generate various fault indices identifying the likelihood of different faults and to generate an overall health indicator for a rolling element bearing 104. In this example, the processor core 210 includes one or more processors that include or otherwise provide various functions. These functions include a fast Fourier transform (FFT) analysis function 212, a frequency/frequency bandwidth selection (FFBS) function 214, a signal reconstruction function 216, a statistical feature determination function 218, and a fault severity index determination function 220. The processor core 210 is also able to perform a rule-based diagnostic function 222 and an artificial intelligence-based diagnostic function 224. These diagnostics are performed to generate one or more output values 226.
The bearing health monitoring device 200 operates based on the concept that different types of defects (such as inner race, outer race, ball, and cage defects) have specific frequency characteristics. The energy content of input signals therefore changes depending on the defect or defects affecting a rolling element bearing. The FFT analysis function 212 performs a fast Fourier transform to convert time-domain signals into frequency-domain signals. It can also perform amplitude demodulation so that the effects of a carrier frequency are reduced or eliminated, allowing the modulating frequencies to be highlighted.
The FFBS function 214 isolates one or more frequencies and amplitudes that will be used in the signal reconstruction function 216 to reconstruct one or more signals. For example, the FFBS function 214 could isolate only those frequencies related to inner race, outer race, ball, and cage defect frequencies. Therefore, contributions from other frequency components can be reduced or eliminated. The signal reconstruction function 216 reconstructs one or more time-domain signals using the isolated frequencies and their respective amplitudes.
The statistical feature determination function 218 determines one or more statistical features of each reconstructed signal. In some embodiments, the statistical features may include Root Mean Square (RMS), Kurtosis, standard deviation, mean, skew, correlation dimension (CD) statistic, D statistic (D-Stat), autoregressive (AR) coefficient, or form/crest factor values. The fault severity index determination function 220 determines one or more fault severity index values associated with each reconstructed signal and/or a baseline signal (baseline data 202). The fault severity index values could represent any suitable values, such as the difference between a specified frequency and sideband frequencies.
The rule-based diagnostic function 222 and the artificial intelligence-based diagnostic function 224 apply various rules to the statistical features and/or the fault severity index values. These rules may include Fuzzy Logic-based rules (such as those having various combinations of membership functions and that apply various aggregation and defuzzification methods). The results of the diagnostics may include a final health indicator 228 for a rolling element bearing 104 and values 230-236 identifying the likelihood of different types of faults.
In some embodiments, the health indicator 228 represents a single value that reflects the health of a rolling element bearing 104, independent of the particular failure mode of the bearing. In other words, the single indicator 228 represents the overall health of the bearing and therefore the likelihood of any fault or defect existing in the rolling element bearing. However, the bearing health monitoring device 200 can also identify the root cause of deteriorating health of a bearing, such as a crack, indentation, or other problem in the outer race, inner race, ball, or cage of the bearing. These are reflected in the values 230-236 generated by the bearing health monitoring device 200. This classification or isolation of faults may facilitate, for example, easy refurbishment of the bearing if required.
The health indicator 228 can act as a severity index (such as a value that varies between 0 and 1) to identify the severity of a failure of a bearing. The health indicator 228 and the related values 230-236 can also be used in any suitable manner. For example, the health indicator 228 and the related values 230-236 could be trended with time to view their changes over a given interval. The health indicator 228 and the related values 230-236 can also be stored in a database to create a historical record of the health for a rolling element bearing 104. The health indicator 228 and the related values 230-236 could further be compared to one or more thresholds to identify if any problems exist. If a problem is found, field personnel or other personnel can be alerted of possible or existing damage to the bearing. In addition, the health indicator 228 and the related values 230-236 could be used to automatically shut down a piece of equipment 102, schedule maintenance, or otherwise take corrective action.
Although
As shown in
A statistical feature and/or a fault severity index value is determined for the reconstructed signal at step 308. This could include, for example, the processor core 210 performing the statistical feature determination function 218 or the fault severity index determination function 220 using the reconstructed signal. A statistical feature and/or a fault severity index value is determined for a baseline signal at step 310. This could include, for example, the processor core 210 performing the statistical feature determination function 218 or the fault severity index determination function 220 using the baseline signal represented by the baseline data 202. A difference between the statistical features and/or a difference between the fault severity index values of the reconstructed and baseline signals is determined at step 312.
A determination is made whether the identified difference is greater than a first (lower) threshold at step 314. If not, a normal condition is detected at step 316, and appropriate action can be taken (such as generating an output indicating that no fault has been detected). If so, a determination is made whether the identified difference is greater than a second (higher) threshold at step 318. If not, a warning condition is detected at step 320, and appropriate action can be taken (such as generating an output warning an operator that a possible fault is developing or has developed). Otherwise, an alarm condition is detected at step 322, and appropriate action can be taken (such as generating an alarm notifying an operator that a fault has developed).
Although
The input signal 402 (after FFT) includes various frequencies 404, including a ball defect frequency, an outer race (OR) defect frequency, an inner race (IR) defect frequency, and a cage defect frequency. These frequencies 404 can be used to identify a set of data 406, which includes amplitudes of the defect frequencies 404 in one or more reconstructed signals. This data 406 is used to produce two statistical values, namely RMS and Kurtosis values 408-410. Also, one or more baseline signals 412 are used to produce two statistical values, namely baseline RMS and Kurtosis values 414-416. The RMS values 408 and 414 can be used to produce normalized RMS values 418, and the Kurtosis values 410 and 416 can be used to produce normalized Kurtosis values 420. In particular embodiments, the normalized values 418-420 can be expressed as:
Here, the “original” signal represents the input signal 402, and the “baseline” signal represents the baseline signal 412. The normalized values 418-420 can then be used in combination to calculate a health indicator (HI) 422 for one or more bearings. As a particular example, Fuzzy Inference can be applied to the normalized values 418-420 to obtain the health indicators 422.
Although
The use of multiple thresholds may allow different courses of action to be taken depending on the severity of the detected problem. For example, the lower threshold could represent a warning threshold, where personnel are simply notified of the possible problem. The upper threshold could represent an alarm threshold, where an alarm is generated and repair or replacement of the bearing is initiated. Of course, other or additional thresholds could be used, and any suitable course of action could be taken for each threshold.
Although
fi; fi±fr; fi±2fr; fi±fb; fi±2fb
2fi; 2fi±fr; 2fi±2fr; 2fi±fb; 2fi±2fb
. . .
nhfi; nhfi±fr; nhfi±2fr; nhfi±fb; nhfi±2fb
where fi denotes the particular defect frequency, fr denotes the frequency of rotation, and fb denotes sideband frequencies. Also, nh denotes the number of harmonics to be included in the family, which can be computed by dividing the upper limit of the frequency band of the filtered input signal (received from the pre-processing filter 208) by fi. Note that a frequency band, such as a band of 1 Hz or 1.5 Hz, could be used with and centered on each frequency in a family of frequencies. This can help to compensate for slight fluctuations in the defect frequency, such as fluctuations caused by supply line fluctuations, manufacturing errors, installation errors, speed changes, slippage of the bearing balls in the bearing, and changes in load, shaft weight, or defect condition.
With this in mind, the bearing health monitoring device 200 decomposes the filtered input signal into the frequency domain using any suitable technique, such as a Fourier series, FFT (as is shown in
where x(t) denotes the reconstructed signal, Ai denotes the amplitude of the ith frequency, and n denotes the number of frequencies being combined. Using signal reconstruction can help in improving the signal-to-noise ratio of the input signal being analyzed.
At this point, statistics (such as RMS and Kurtosis values) can be calculated using the reconstructed signal. An example of these statistics is shown in
The computed statistics can be normalized with respect to a baseline reconstructed signal, which allows the bearing health monitoring device 200 to operate uniformly across different pieces of equipment, such as equipment with different capacities. A feature fusion technique can then be applied to the normalized values, such as Fuzzy Logic, Dempster-Shafer, or Bayesian feature fusion, to determine the final value of the health indicator 228.
As shown in
Note that any other or additional rules could be used here. For example, a rule could specify that a normalized RMS value greater than four results in an alarm condition (regardless of the normalized Kurtosis value). Also, other or additional conditions could be specified using appropriate controls. For instance, if an alarm appears three consecutive times, the alarm can be marked as being confirmed.
Although
As shown in
The bearing health monitoring device 200 then goes through a series of checks to validate the received configuration data. For example, the bearing health monitoring device 200 verifies whether the desired number of samples N is less than a maximum number of samples (NS) at step 902 and, if not, outputs a warning instructing a user to reduce the desired number of samples at step 903. The bearing health monitoring device 200 verifies whether the upper frequency limit FH is less than half the desired sampling frequency FS at step 904 and, if not, outputs a warning instructing the user to reduce the upper frequency limit FH at step 905. The bearing health monitoring device 200 verifies whether the upper frequency limit FH is less than the natural frequency FN of an accelerometer at step 906 and, if not, outputs a warning instructing the user to reduce the upper frequency limit FH at step 907.
If and when validated configuration data is obtained, the bearing health monitoring device 200 collects accelerometer data at step 908 and OPR data at step 909. The OPR data is processed as shown in
The filtered accelerometer data is also analyzed to calculate statistic values (such as RMS values) for the accelerometer data at step 911. A determination is made whether the computed statistics indicate that the accelerometer's maximum rating has been exceeded at step 912. If so, a message is presented instructing the user to replace the accelerometer at step 913. Otherwise, the velocity and displacement of the equipment being monitored, as well as the RMS value of the velocity, are determined at step 914. These values can be determined using the filtered accelerometer data.
The computed RMS value of the velocity is compared against one or more threshold values associated with the appropriate class of machine at step 915. Example thresholds associated with different classes of machines are shown in the table 1000 in
The OPR data collected at step 909 is processed as shown in
The filtered accelerometer data is subject to FFT analysis at step 920, which converts the data from the time domain into the frequency domain. At this point, a determination is made whether all of the FoF frequencies are less than 1 kHz at step 921. If so, the maximum amplitude at each FoF frequency (within a ±1 Hz window) is determined at step 922. If not, a determination is made whether all of the FoF frequencies are less than 2 kHz at step 923. If so, the maximum amplitude at each FoF frequency (within a ±1.5 Hz window) is determined at step 924. Otherwise, the maximum amplitude at each FoF frequency (within another window) is determined at step 925. The amplitudes and frequencies are used to form a matrix at step 926.
At this point, the values in the matrix are analyzed to determine fault indicators for individual fault modes in
In Table 1, the “otherwise” condition is met if at least one of the normalized values is above its lower threshold and below its upper threshold.
Similarly, for an outer race defect, a signal is reconstructed using all outer race defect frequencies at step 931, and statistical features (such as RMS and Kurtosis values) are determined for the reconstructed signal at step 932. Normalized RMS and Kurtosis values are determined at step 933, which may involve using RMS and Kurtosis values for a baseline signal. The likelihood of an outer race defect is then determined using the normalized RMS and Kurtosis values at step 934. For example, the likelihood of an outer race defect could be determined as shown in Table 2.
In Table 2, the “otherwise” condition is met if at least one of the normalized values is above its lower threshold and below its upper threshold.
Further, for a bearing ball defect, a signal is reconstructed using all ball defect frequencies at step 935, and statistical features (such as RMS and Kurtosis values) are determined for the reconstructed signal at step 936. Normalized RMS and Kurtosis values are determined at step 937, which may involve using RMS and Kurtosis values for a baseline signal. The likelihood of a ball defect is then determined using the normalized RMS and Kurtosis values at step 938. For example, the likelihood of a ball defect could be determined as shown in Table 3.
In Table 3, the “otherwise” condition is met if at least one of the normalized values is above its lower threshold and below its upper threshold.
In addition, for a cage defect, a signal is reconstructed using all cage defect frequencies at step 939, and statistical features (such as RMS and Kurtosis values) are determined for the reconstructed signal at step 940. Normalized RMS and Kurtosis values are determined at step 941, which may involve using RMS and Kurtosis values for a baseline signal. The likelihood of a cage defect is then determined using the normalized RMS and Kurtosis values at step 942. For example, the likelihood of a cage defect could be determined as shown in Table 4.
In Table 4, the “otherwise” condition is met if at least one of the normalized values is above its lower threshold and below its upper threshold.
The values in the matrix formed at step 926 are also analyzed to determine the overall health indicator for the bearing as shown in
The normalized values are then “fuzzified” using one or more Fuzzy Logic membership functions at step 947. This could include fuzzifying one or more Fuzzy Logic output member functions at step 949 and providing rules for one or more Fuzzy Logic input member functions at step 950. Example Fuzzy Logic input member functions 1100 and 1120 are shown in
Although
In some embodiments, various functions described above are implemented or supported by a computer program that is formed from computer readable program code and that is embodied in a computer readable medium. The phrase “computer readable program code” includes any type of computer code, including source code, object code, and executable code. The phrase “computer readable medium” includes any type of medium capable of being accessed by a computer, such as read only memory (ROM), random access memory (RAM), a hard disk drive, a compact disc (CD), a digital video disc (DVD), or any other type of memory.
It may be advantageous to set forth definitions of certain words and phrases used throughout this patent document. The term “couple” and its derivatives refer to any direct or indirect communication between two or more elements, whether or not those elements are in physical contact with one another. The terms “application” and “program” refer to one or more computer programs, software components, sets of instructions, procedures, functions, objects, classes, instances, related data, or a portion thereof adapted for implementation in a suitable computer code (including source code, object code, or executable code). The terms “transmit,” “receive,” and “communicate,” as well as derivatives thereof, encompass both direct and indirect communication. The terms “include” and “comprise,” as well as derivatives thereof, mean inclusion without limitation. The term “or” is inclusive, meaning and/or. The phrases “associated with” and “associated therewith,” as well as derivatives thereof, may mean to include, be included within, interconnect with, contain, be contained within, connect to or with, couple to or with, be communicable with, cooperate with, interleave, juxtapose, be proximate to, be bound to or with, have, have a property of, have a relationship to or with, or the like.
While this disclosure has described certain embodiments and generally associated methods, alterations and permutations of these embodiments and methods will be apparent to those skilled in the art. Accordingly, the above description of example embodiments does not define or constrain this disclosure. Other changes, substitutions, and alterations are also possible without departing from the spirit and scope of this disclosure, as defined by the following claims.
This application claims priority under 35 U.S.C. §119(e) to U.S. Provisional Patent Application No. 61/085,637 filed on Aug. 1, 2008, which is hereby incorporated by reference.
Number | Date | Country | |
---|---|---|---|
61085637 | Aug 2008 | US |