1
Use of Tach from Vibration to Estimate Bearing Spall Length
Eric Bechhoefer
1
, Omri Matania
2
, and Jacob Bortman
3
1
GPMS International, Inc, Waterbury, VT, USA
eric@gpms-vt.com
2,3
Ben-Gurion University of the Negev, Isreal
omrimatania@gmail.com
jacbort@bgu.ac.il
ABSTRACT
Vibration analysis is often used for bearing fault diagnostics.
Envelope analysis or other cyclo-stationary processes can
capture a fault feature or condition indicator that is correlated
to the spall length. However, no study has defined a process
for estimating spall length on real-world data. The problem is
that the spall length is a time-domain property of the signal.
This paper generates a synthetic tachometer signal from the
fault itself. It is synchronous to the rolling element exit from
the spall, allowing for a time-domain representation of
waveform using the time synchronous average. From this, an
estimate of the length of the bearing fault can be determined.
1. HEALTH AND USAGE MONITORING SYSTEM
OBJECTIVES
One of the objectives of a Health and Usage Monitoring
System (HUMS) is to identify degrading drivetrain
components, such as rolling element bearings. Rolling
element bearing failures form one of rotating equipment's
most critical failure modes. Even though vibration analysis
has been successfully used for bearing fault detection and
diagnostics for years, it typically does not estimate the spall
length of the bearing. Some studies have used simulation or
notched bearing faults on a test stand to estimate spall length,
but few methods have been used to estimate spall length in
real-world systems.
An estimate of the spall length would provide insight into the
degrading reliability of a drivetrain as the fault propagates.
This would improve the timeliness of scheduling a
maintenance action. In this paper, the time domain features
associated with the rolling element (RE) entry and exit into
the spall can be used to estimate the spall length.
The RE–spall interaction generates an impact event in the
time domain. These features are typically not visible due to
the superposition of other various signal sources (gear mesh,
shaft modes) within the gearbox. While a bearing signature
may have a peak-to-peak value of 0.3gs for a fault, gear mesh
could be 10x this value. One signal separation method in the
time/angular domain is the time-synchronous average (TSA).
However, the TSA is ineffective in extracting these features
associated with a bearing because there is typically no
tachometer on the bearing cage. The bearing elements are
asynchronous to the shaft they support because of the RE
non-Hertzian contact with the inner and outer races.
Eps (1991) observed acceleration of the RE associated with
the RE passage into and out of the spall. This consisted of a
step response for the RE entering the spall and an impulse
response for the RE when exiting the spall. The step response
(entry into the spall) had less frequency content than the
impulse response. The impulse response of RE on the exit of
the spall is broadband. The impact generates a high-
frequency, cyclo-stationary response, as was reported by
Antonni (2009).
Building on Eps, Kogan et al. (2017), in "A new model for
spall-rolling-element interaction," developed a simulation
model of the interaction of the RE entry and exit from the
spall. This suggests that using acceleration data, it would be
possible to estimate the spall length as a time domain feature.
The length of the spall would be the difference in time from
the entry and exit into the spall. This time difference is then
multiplied by the shaft rate and the circumference of the race
(inner or outer) to give an estimate of the spall length.
From the bearing’s equation of motion, the fault frequency is:
!
!"
"
#
$
%
!
#$%&'
&
'()
*
()
*
*
+
,-.
/
0
1
2
(1)
!
+"
"
#
$
%
!
#$%&'
&
'3)
*
()
*
*
+
,-.
/
0
1
2
(2)
Eric Bechhoefer et al. This is an open-access article distributed under the
terms of the Creative Commons Attribution 3.0 United States License,
which permits unrestricted use, distribution, and reproduction in any
medium, provided the original author and source are credited.
