GB2233761A - Bearing defect detector - Google Patents

Abstract

An acoustic system for detecting defects in the bearings of moving railroad cars and identifying the location of these defects has means for opening a window on detection of a wheel passing a particular point such that signals received relate only to the bearing and extraneous signals are excluded. <IMAGE>

Description

IMPROVED ACOUSTIC BEARING DEFECT DETECTOR The present invention relates to bearing defect detectors and in particular to an acoustic system for detecting defects in the bearings of moving railroad cars.

Heretofore, the detection of defects in railway car bearing has relied upon stationary infrared sensing means located along railroad tracks to detect any abnormal heat rise associated with bearing failure in passing railroad cars. While such systems have enjoyed widespread use and an industry-wide reputation for reliability, they suffer from a serious drawback in that they detect a defect only after a damaging heat build-up has occurred within the bearing. Furthermore, this heat build-up often does not occur until a total bearing failure is imminent, thereby normally warranting immediate stopping of the train so that an emergency field repair may be done. As this delays the train until a repair team arrives with the necessary equipment, the total cost can be very high.

From U.S. Patent No. 4,790,190 and pending U.S. patent application Serial No. 168,975 filed on March 16, 1988, it is known that defects in tapered roller bearings, such as those used in railroad cars, produce sounds, during operation, at characteristic frequencies dependent upon the location or type of defect (i.e at bearing cup, cone, or roller), the combination of the size of the wheel and the bearing capacity (frequently encountered combinations on railroads are a 28 inch wheel with a 70 ton capacity bearing, a 33 inch wheel with a 70 ton capacity bearing, and a 36 inch wheel with a 100 ton capacity bearing), and the speed of the train (which, of course, for a given diameter of the wheel, is proportional to the rotational frequency of the wheel).

Additionally, irregularities in the wheel circumference ("flats") produce a characteristic frequency dependent upon wheel rotational frequency.

Thus, for any given train speed, a defective bearing will produce a sound at one or more characteristic frequencies dependent upon the location of the defect in the bearing and the combination of the train speed, wheel size and bearing capacity. Wheel flats will produce a sound at one of three additional characteristic frequencies. Ideally, one need only listen for and detect the characteristic sound frequencies to determine the condition of the bearings of the passing train.

Unfortunately, railroad trains operate in extremely noisy environments. Train noises (such as wheel/rail rubbing, flange/rail squeal, loose equipment and cargo sounds, carbody noises, and clacking of rail joints for tract circuits) and a wind "swish" are low-frequency sounds which tend to camouflage the impact frequencies produced by a defective bearing. Thus, while the production of characteristic impact frequencies in a moving bearing have been known,it has heretofore been impossible to isolate the impact frequencies in a meaningful manner so as to provide useful and reliable information. More importantly, the background noise may trigger defective bearing alarms and it thus becomes important to be able to distinguish between true defective bearing alarms and alarms set off by this background noise.

One object of this invention is to enable the detection of a defective bearing, particularly in on-site railroad applications, before damaging heat build-up has occurred thereby giving repair crews sufficient early warning so as to allow the train to safely run to a place where repairs can be performed conveniently and inexpensively and/or to repair the bearing in its early stages of failure.

Another object is to enable the location of the defective bearing in the train (car number, axle number and axle side) to be determined.

A further object of this invention is to enable the type of bearing defect (i.e, whether the defect is in the bearing cup, cone, or roller) to be identified before repairs begin.

Still a further object of this invention is to enable the detection of defects reliably in a noisy environment such as is common in railroad and other industrial applications.

In analyzing the sounds generated by passing railroad trains, it has been noted that the impact frequencies characteristic of various bearing defects are generated in amplitude-modulated form on an acoustic carrier frequency band which is independent of the speed of the passing train. Thus, while the bearing signature may be a function of train speed, the carrier frequency bands are not.

In accordance with the present invention, we propose placing a microphone beside railroad tracks so as to monitor the sounds emanating from the bearings of a passing railroad train.

Preferably microphones are placed adjacent opposite rails. A series of electronic amplifiers and preliminary filters is used to increase the output of the microphone to a usable voltage level and to filter extraneous frequencies from the resulting electronic signal. This allows the apparatus to operate effectively notwithstanding excessive ambient noise.

The gain of the amplifier is adjusted by the train speed and direction sensor so that the output of the amplifier is relatively constant in spite of significant volume changes due to train speed and direction (in that, under some circumstances, train direction is related to train loads, which is, in turn, related to the volume of noise generated thereby).

After electronic amplification and preliminary filtering, the signal consists predominantly of the preselected carrier frequency band which, in the presence of the characteristic impact frequencies of bearing defects, will be generated within an envelope of the characteristic impact frequencies. In other words, there is an amplitude-modulated signal with a preselected carrier frequency band within an envelope defined by any characteristic impact frequencies which may be present.

Standard amplitude demodulation techniques are used to extract the envelope in which the carrier frequency band is being transmitted. The extracted frequency signal is processed by an array of low-frequency bandpass filters which pass only the frequency ranges in which the various characteristic impact frequencies occur. The centre frequencies and the bandwidth of these filters are automatically increased with increasing train speed by the use of switch capacitors, responsive to a train speed and direction sensor.

The array of train speed tuned filters thus performs a spectral analysis of the resulting signal.

The predominant frequency values of the spectrum analyzer are set automatically within its filter frequency range to the expected characteristic impact frequency values of a defective bearing.

When a characteristic impact frequency value falls within the filter range of the filter circuitry of the spectrum analyzer,the operators of the apparatus notify the operators of the train in order that appropriate plans for repair may be made.

An embodiment of the present invention will now be described by way of example with reference to the accompanying drawings in which Figure 1 is an elevation of a typical tapered bearing as is used in railroad applications; Figure 2 is a schematic view of acoustic bearing defect detector apparatus, and Figures 3A and 3B show delay gates for a raceway (cup or cone) defect and a roller defect, respectively.

Figure 1 shows a tapered roller bearing assembly 10 such as those commonly used in railroad applications. Components which commonly fail because of rolling surface defects (spalls, Brinell, cracks and pits) are the roller 12, the cup 14 and the cone 16. Upon failure of any component, a characteristic component dependent rotational speed-dependent acoustic impact frequency generates a sound spectrum which amplitude modulates a carrier acoustic signal of rotational speed-inde-pendent frequency contained within an envelope of the spectrum.

It has been found that the carrier signal has a frequency of approximately 10-12 kilohertz, while sample characteristic impact frequencies for a 6.1/2" x 12" bearing assembly (100 tone capacity with a 36" wheel) are listed below: Characteristic Impact Frequencies (Hz) Defect associated with 36 inch wheel 20 mPh 30 mph 40 mph 6.1/2" x 12" bearing assembly (100 ton capacity) roller 14.6 21.9 29.2 cup 32.1 48.1 64.1 cone 39.5 59.3 79.0 The acoustical signal processor 20 shown in Figure 2 receives an input from a horn 22 wherein acoustic vibrations are transduced into an electrical signal. Ideally, this is an electret microphone in a horn assembly with a cavity optimised for the desired bandpass of 10 to 12 kilohertz.A horn is preferred since it provides the directional characteristics needed to isolate and identify defective bearing axle locations without ambiguity.

After extensive experimentation, it has been found that placement of the horn assembly is critical so as to avoid picking up excessive background noise (e.g. wheel/rail rubbing; flange/rail squeal; loose equipment or cargo rattling; car body noise; clacking of rail joints for track circuits). The horn must not be far back so as to pick up excessive background noise but must be close enough to pick up a signature over a horizontal linear distance corresponding to one circumference or rotation of the wheel in order to capture defect repetition rate signatures that generally repeat once per wheel rotation. A wider range than this causes ambiguous results as the horn may monitor different axles simultaneously. A substantially narrower range than this may result in inadequate information for processing.However, a slightly narrower range than this was found advantageous in order to definitely insolate and locate the axle of interest.

Furthermore, the horn should not pick up over a vertical distance above or below the wheel. Therefore, a horn elongated in the horizontal plane and shortened in the vertical plane to collect acoustic vibrations from a 60 degree horizontal angle and a 20 degree vertical angle (i.e.± 30 degrees horizontally and ± 10 degrees vertically from a line perpendicular to the rail) is well adapted to the present purpose. It is preferred to place horn on both sides of the tracks (one horn for each track). The horn should be placed five feet outboard of the rail gauge 19 and collect sounds from a 60 degree horizontal angle, or somewhat over five feet of track. Additionally, the horn should be directed horizontally with its horizontal axis 12"-16" above the rail picking up sound vertically from a 20 degree vertical angle for about two feet; at a level just below the centreline of a wheel.

A horn 22 is placed on both sides of the railroad track so as to monitor the bearings of wheels traversing both rails.

Additionally, each horn assembly 22 includes a test tone generator (not shown) which generates a test tone equal to a typical modulation carrier frequency (11.1. kHz) which is modulated sequentially to simulate a range of defects and possible train speeds. The test tone generator on one side of the track is used to generate test tones to calibrate the equipment on the other side of the track during idle periods.

A preamplifier is incorporated into the horn assembly to amplify the bearing acoustical signature and to prevent the broad range of background noise from overloading later amplifying and processing means. Additionally, a heater and protective shutter are added to the assembly to prevent ambient weather conditions from either damaging or altering the response of the microphone.

The shutter opens as the train passes so as to not interfere with the acoustic analysis, but closes thereafter to protect the apparatus from the effects of weather or vandals. Additionally, the heater is turned off as the train passes to prevent the heater from generating alternating current frequency noise.

It will be understood that the circuitry described hereafter is for a single horn 22. Both horns 22 have identical parallel circuitry.

The output of horn 22 is input into the Acoustical Signal Processor 20. The Acoustical Signal Processor 20 extracts the electrical signals related to the resonant metallic sound signatures picked up by the horns 22. The Acoustical Signal Processor 20 includes a bandpass filter 25, demodulator 26, and a spectrum analyzer 27 for each of the signals output from each of the horns 22 positioned alongside rail 1 and rail 2, as shown in FIG.2. The spectrum analyzer 27 is implemented by a matrix array of switch capacitor bandpass filters.

Defects on the cup and cone raceways and on the rollers have distinctive characteristic impact repetition rates that are a function of bearing geometry and the rotational rate of an axle, which in turn depends upon wheel diameter and train speed. Chart A illustrates computed impact repetition frequencies for commonly encountered wheel diameter and bearing size combinations for a train speed of 40 m.p.h.

The spectrum analyzer 27 is pre-programmed to recognise the characteristic cup, cone and roller defect frequencies for each of the commonly encountered wheel diameter and bearing size combinations. These frequencies are proportional to train speed and thus the spectrum analyzer 27 is dynamically controlled by train speed information. Two wheel present transducers A and B, described more fully below, provide information relating to train speed and axle count which is received by circuit 28. Acoustical signatures are correlated with the axle count.

The spectrum analyzer 27 is formed of a matrix of switched CHART A DEFECT FREQUENCIES (Hz) Train Speed 36 inch wheel 33 inch wheel 28 inch wheel 40 MPH 100ton bearing 70ton bearing 70ton bearing Cone 79.0 89.5 105.5 Cup 64.1 73.5 86.6 Roller 29.2 33.8 39.9 lx Rotation Wheel 6.22 6.79 8.00 capacitor bandpass filters including a 3 x 3 array of filters for bearing defects on one rail side, a second 3 x 3 array of filters for bearing defects on the other rail side, and a single 3 x 1 array for wheel defects on either rail side.

Each filter of the spectrum analyzer 27 is an 8-pole Chebyshev bandpass filter having a nominal 1/3 octave bandwidth (3dB bandwidth of +11% of centre frequency). The switched capacitor filters have the unique characteristic of a constant normalised filter passband shape with regard to the centre frequency.

The centre frequency of each filter is determined by a applied clock pulse frequency that is 54.5 times the desired centre frequency. For each bearing defect filter its clock pulse frequency is assigned a constant of proportionality with respect to train speed that is dependent on wheel diameter, bearing size and bearing defect type. Similarly, for each wheel defect filter its clock pulse frequency is assigned a constant of proportionality with respect to train speed that is dependent on wheel diameter. Thus, the clock pulse frequencies of the filters i the matrix are controlled dynamically to be proportional to train speed to thereby effectively tune the spectrum analyzer to the expected defect frequencies.

An exclusive OR network is provided for the output signals of each of the filter arrays of the spectrum analyzer 27. Each exclusive OR network would automatically pass, at any instant in time, the highest output of the associated filter array. For example for one of the 3 x 3 arrays the associated exclusive OR network would have as inputs all nine outputs of the array, these outputs being applied to exclusive OR gate network so that only the highest output would pass through, while the other eight outputs would be blocked. Thus only the highest output and only the bandwidth of the corresponding filter becomes active as a result of the exclusive OR network to exclude noise in the frequency bands of the remaining filter outputs.

The outputs from each of the three exclusive OR networks can each be applied to an individual track of a multi-track analog chart recorder so that the analog outputs of the 3 x 3 matrix of bearing defect filters and the 3 x 1 array of wheel defect filters can be displayed and recorded in real time on an analogue chart recorder and also converted by means of analogue to digital converters, stored in digital memory and printed as an array of numbers representing amplitudes for each bearing and wheel.

Outputs exceeding preset levels would trigger alarms. Such levels may be amplitude, area under the curve, etc.

Since wheels with severe flats or other tread defects can generate intense broadband acoustical noise that may stimulate an output from a bearing defect filter, the system can be made to ignore bearing alarms for a specified number of axles ahead of, or after the detection of a flat wheel.

A rather useful real time analogue display and recording is obtained by applying the output from the exclusive OR network of one of the 3 x 3 arrays for one rail side to one track of the analogue chart recorder and the exclusive OR network output of the 3 x 3 array for the other rail side to a second track of the analogue chart recorder.

An important feature is a window by which a "true" alarm i.e. an alarm that is actually triggered by a bearing defect on a specified axle, can be distinguished from "false" alarms caused by background noise. If an alarm is within the window then it is associated with the specified axle with the corresponding bearing defect identified by the spectrum analyzer filtering circuit 27, otherwise it is presumed not to be a bearing defect and ignored by the system.

Wheel presence transducers A and B are spaced apart and positioned on the track side to define a gate or window. The two horns 22 are located perpendicular to the rails and midway between transducers A and B. The window defined by transducers A and B corresponds to the detection of a particular wheel/bearing/axle set within the window. The window duration is issued to determine train speed. Whether transducer A is followed by transducer B or vice versa determines direction. By counting the number of windows from the head end of the train it is possible to keep track of the number of axles passing through the zone defined by transducers A and B.

As a defective bearing passes the trackside horn 22 corresponding to the side of the train the bearing is on, the output signal of the horn 22 reaches a peak value when the bearing is in the A to B zone due to the horizontal directivity of the horn 22.

Therefore due to the directivity characteristics of the horns 22 and where they are placed, the location of a defective bearing with an abnormally high acoustical signature can be determined in terms of trainside and axle count. The distance between wheel/axle sets can be determined from train speed and the time interval between successive windows. Axle spacing patterns are used to separate axles into those grouped on trucks, and the trucks on cars, so that defect location can then be defined by car count, lead or trail truck on the car, axle position on the truck and axle side.

The propagational delay time of each defect filter in the spectrum analyzer 27 is inversely proportional to the absolute value of its bandwidth which in turn is proportional to its centre frequency which may be controlled to be proportional to the train speed. Therefore the propagation delay time of a filter, that is the time delay between when an acoustical signature is received by a trackside acoustical sensor and the time a processed resulting signal is output from the filter, is inversely proportional to train speed.

Since the distance travelled by the train for a given time interval is proportional to train speed, each filter of the spectrum analyzer 27 has a processing delay that is a constant displacement with respect to train movement relative to the location of the horns 22. The propagational delay distance constants of the filters are inversely proportional to their programmed centre frequency to train speed proportionality constants. The higher the relative defect frequency the shorter the delay distance.

In the preferred embodiment, the switched capacitor filters have a delay distance of 3 to 4 feet downstream from the location of the trackside horns 22 for the 36 inch wheel/100 ton, 33 inch wheel/ 70 ton and 28 inch wheel/70 ton cone and cup (raceway) filters.

For the 36 inch wheel/too ton, 33 inch wheel/70 ton, and 28 inch wheel/70 ton roller filters, the delay distance lies in the range of 7 to 9 feet.

Referring now to Figs. 3A and 3B the wheel presence transducers A and B are preferably spaced two feet apart. As shown in Fig.

32A, when a wheel moves from A to B, a raceway defect window is initiated which starts when the wheel is at A and ends either when the wheel is seven feet downstream from A or when a second wheel reaches A, which ever occurs first. For roller defect detection, as shown in Fig.3 B, the window begins when a wheel is six feet downstream of A and ends when the wheel has moved twelve feet downstream of A or when a second wheel passes and moves six feet downstream of A, whichever occurs first. A similar approach is used for train movement from B to A with B as the reference.

Thus false alarms are minimized as are errors in identifying the location of the bearing defect by providing a window for each cone and cup (raceway) defect and also a second window for each roller defect thereby straddling the corresponding propagation delay distances with the associated window. False alarms are those alarms that are not associated with bearings and which may be caused by wheel squeal, wheel defects, and carbody noise.

If an alarm exists just prior to the beginning of a window and continues to exist during that window then it is classified as a false alarm probably caused by background noise. Furthermore, bearing alarms in the vicinity of a wheel defect alarm can be negated. Thus bering defect alarms eight axles ahead of and four axles after a defective wheel alarm are negated.

A television camera (not shown) activated by a

bring alarm within a corresponding window may be aimed to televise the car and axle having the defect, only when the alarm is generated within the window but not if the alarm is generated before and continues after the window starts.

Claims (24)

1. Apparatus for acoustically scanning bearings of a train or other vehicles moving through a scanning zone for defects during operation of the vehicle, the apparatus comprising means for transducing acoustic vibrations to produce an electrical signal foreach of the bearings, filtering means for filtering frequency components from the electrical signal means for analyzing the said frequency components, and means for identifying particular bearing defects wherein means are provided for generating a window during which an individual bearing is scanned and means to determine if a processed signal is produced by that bearing.

2. Apparatus according to Claim 1 wherein the processed signal is indicative of particular bearing defects.

3. Apparatus according to Claim 2 further comprising means for positioning the window with respect to the scanning zone for said particular defects.

4. Apparatus according to Claim 1 further comprising means for determining if said processed signal is at a level for generating an alarm for a particular bearing defect.

5. Apparatus according to Claim 3 wherein the window is positioned spatially.

6. Apparatus according to Claim 5, further including a matrix of switched capacitor bandpass filters set within frequency ranges corresponding to particular bearing defects to identify the bearing defect associated with the alarm generated within the electric window.

7. Apparatus according to Claim 1 wherein the means for identifying particular bearing defects includes means for generating an alarm signal if the processed signal occurs within the window.

8. Apparatus according to Claim 7, further comprising means for inhibiting the alarm signal if the processed signal initiates before the window and continues into the window.

9. Apparatus according to Claim 7 wherein the means for generating an alarm measures the peak value of the processed signal within the window.

10. Apparatus according to Claim 7 wherein the means for generating an alarm measures the peak value and the width of the processed signal within the window.

11. Apparatus according to Claim 7 wherein the means for generating an alarm measures the area of the processed signal pulse.

12. Apparatus according to Claim 7 further comprising means for inhibiting the alarm signal for a specified number of axles prior to and/or after the detection of a flat wheel.

13. Apparatus according to Claim 6 wherein the switched capacitor filters have propagational delay times inversely proportional to vehicle speed.

14. Apparatus according to Claim 1 further comprising horns disposed below the centrelines of the wheels to monitor the bearings on the wheels.

15. Apparatus according to Claim 6, wherein the switched capacitor filters includes a first 3 x 3 array of filters for filtering signals detecting particular defects in the bearings of a train on one rail side, a second 3 x 3 array of filters for filtering signals detecting particular bearing defects on the other rail side, and a 3 x 1 array of filters for filtering signals detecting wheel defects on either track side.

16. Apparatus according to Claim 15 wherein each of the array of filters has outputs and further comprising an exclusive OR network for each of the array of filters, the network exclusive OR summing the outputs of its associated array of filters so that the highest output passes through the network while blocking the outputs of the remaining filters.

17. Apparatus according to Claim 16 wherein the exclusive OR network limits the noise bandwidth to that of the highest input signal.

18. Apparatus according to Claim 16 further comprising a multitrack analog chart recorder adapted to receive and display in real time the results from the exclusive OR networks of the first and second 3 x 3 array of filters and said 3 x 1 array of filters on first, second and third tracks of said chart recorder, respectively.

19. Apparatus according to Claim 1 wherein the particular bearing is scanned within the window and a signature associated with the bearing is identified.

20. Apparatus according to Claim 19 further comprising means for determining if the signature is at a level for generating an alarm for its associated bearing defect.

21. Apparatus according to Claim 1 wherein during scanning of the bearings of a train, the window identifies car count, lead or trail truck on the car, axle position on the truck and axle side to determine the defect location.

22. Apparatus for acoustically scanning the bearings of a train or other vehicles, constructed and arranged substantially as hereinbefore described with reference to and as illustrated in the accompanying drawings.

23. A method for monitoring vehicle bearings during operation of the vehicle as it moves through a scanning zone, comprising transducing acoustic vibrations to produce an electrical signal associated with each of the said bearings, filtering frequency components from the said signal and analyzing the said frequency components to identify particular bearing defects, wherein a time window is established during which an individual bearing is scanned for determining if a processed signal is produced by that bearing.

24. A method for monitoring vehicle bearings substantially as hereinbefore described with reference to the accompanying drawings.