Data processing: measuring – calibrating – or testing – Measurement system in a specific environment – Mechanical measurement system
Reexamination Certificate
2002-12-04
2004-09-07
Barlow, John (Department: 2863)
Data processing: measuring, calibrating, or testing
Measurement system in a specific environment
Mechanical measurement system
C702S035000, C702S043000
Reexamination Certificate
active
06789025
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of the Invention
The invention generally relates to systems and methods for monitoring the condition of a machine. More particularly, the invention concerns systems and methods for analysis of machine vibration signals to detect and evaluate specific sources of energy contribution to the vibration energy of the machine.
2. Description of the Related Art
It is common for industrial and commercial facilities to operate a large number of machines concurrently, many of which may cooperate in a large interdependent process or system. Despite increasingly efficient maintenance programs, at any time some percentage of the machines develop defects that are likely to lead to machine failure. For example, machines having moving parts (e.g., bearings) experience constant friction that results in wear. It is known that bearing failures are a major cause of motor faults. Bearing damage due to wear may not be apparent, however, absent gross damage or failure of the motor because the bearing's wear site is likely concealed in the motor's assembled state.
Consequently, the use of machine condition monitoring systems has become essential to preventive maintenance of industrial machinery in order to avoid down time or catastrophic failure of machines. Unscheduled plant shutdowns can result in considerable financial losses. Failure of high performance machinery can lead to fatal injury and processing system backup. Typical benefits from a preventive maintenance program include longer periods between machinery shutdowns, evaluation of the condition of machine components without resorting to costly and/or destructive disassembly for visual inspection, and prolonging the machinery's operational life by taking corrective action when developing faults are identified early.
Measurement and analysis of machine vibrations typically includes sensing the machine's vibrations with a transducer that converts the vibration information to electrical signals. The electrical signals are processed so that a history of vibration amplitude over time can be obtained. Data points representing amplitude at a certain point in time may be plotted on a graph of amplitude versus time. This graph is often referred to as the time-domain vibration signature of the machine.
FIG. 1
shows an exemplary graph of time-domain vibration data.
FIG. 1
is a plot of measured acceleration of a point of a machine assembly over a period of about eight seconds. The particular machine from which this data was measured was rotating at 104.98 rpm, so
FIG. 1
shows data over the course of about 15 revolutions. Peak values measured were about 0.025 g.
Rotating and reciprocating components of a machine produce vibrations having a wide range of frequencies. In addition to the time-domain data representation of machine vibrations, the vibrations of a machine, machine component, or other phenomena acting on the machine may be characterized by a plot of vibration energy as a function of vibration frequency. This diagram is commonly referred to as a “frequency spectrum,” “spectral diagram,” or “spectrum plot.”
FIG. 2
shows an exemplary frequency spectrum, which was derived from the time-domain vibration data of FIG.
1
. Although the frequency scale is not illustrated in
FIG. 2
, prominent peaks are seen at about 10-11 Hz (designated as peak
10
) and about 87 Hz (designated as peak
20
).
Sometimes it is useful to derive a “profile plot” of the vibration data.
FIG. 3
shows an exemplary profile plot derived from one revolution of the machine rotor shaft. The data of
FIG. 3
corresponds to about the first 0.57 seconds of the time-domain vibration data of
FIG. 1
, which is the time for one revolution at 104.98 rpm. In a profile plot, the measured acceleration is plotted as the radial distance from a selected angular location on a circle
350
which represents one revolution of the machine rotor. Conventionally, a machine shaft orientation of zero degrees corresponds to the top of the circle
350
. Thus, a profile plot provides a visual representation of the measured acceleration (or velocity, or displacement as may be desired) as a function of the position of the rotating machine shaft or other periodic event associated with the machine.
The frequencies and associated peaks of the vibrations of a specific machine collectively make up the “frequency spectrum” for the machine, also known as the machine's “vibration signature.” A machine's vibration signature varies with, for example, the design, manufacture, application, and wear of its components. The machine's normal operating conditions determine the amplitude of steady (or “normal”) vibration. It is a common practice to obtain a reference frequency spectrum when the machine is known to be in good condition for comparison against future measurements of the machine's frequency spectrum. Such comparison aids in detecting changes in the condition of the machine or its subcomponents. Hence, analysis of a machine's vibration signature provides valuable insights into the condition of the machine.
A technique known as synchronous time averaging (“STA”) has been utilized to detect the “fault energy” contribution of a periodic signal as a means to troubleshoot product quality problems on machinery, such as for example a paper machine. STA is also commonly utilized by data acquisition systems to boost the signal to noise ratio. STA extracts from the time-domain data those signals that are repetitive and synchronized to a physical event, e.g., the rotation of a shaft. Upon the reception of a trigger signal, the data acquisition system acquires N samples at a predetermined sampling frequency F
s
. Hence, the total time for one measurement is T=N/F
s
. Upon the next trigger event, the system acquires and stores another N samples. Using STA, the system then averages the two data sets, on a corresponding sample-by-sample basis, that is, the first sample from the first N samples is averaged with the first sample from the second set of N samples, and so on. This averaging results in a derived time-domain waveform of N averaged data points. The system may in a similar fashion acquire more than two sets of N samples and use STA to produce the averaged waveform. Thus, the total time required for acquiring the data for processing with STA is (N/F
S
)*A, with A being the number of sets included in the averaging.
STA suppresses noise random to the signal synchronized to the trigger event because the noise component of the signal averages out after the system acquires and averages multiple data sets. The periodic signal does not average out because the system acquires the data based on the trigger and thereby starts collection of the N samples in synchronization with the periodicity of the physical event.
By way of example, to apply STA to signals from a machine having a rotating shaft, typically the shaft is configured to provide for a once-per-revolution signal that occurs every time the shaft is in a particular position. Another example is a belt and pulley system. To apply STA to signals from the belt, a trigger provides a once-per-revolution signal that occurs every time the belt is in a particular position. The physical location of the trigger is not critical, as long as it remains constant over each data acquisition cycle. In the context of the profile plot of
FIG. 3
, for example, the trigger event may correspond to the top of the circle
350
. Thus, in these conventional systems, if it is desired to compute STA measurements for shaft related phenomena as well as belt related phenomena using the same sensor to see how each (i.e., the shaft and the belt) phenomenon contribute to the vibration at the point of the sensor, it is typically required to perform two different STA measurements, each with its own set of averages and each with its own trigger.
Notwithstanding the several methods available for component defect detection and/or fault energy contribution by analysis of a machine's frequency spectrum or time-
Barlow John
Le John
SKF Condition Monitoring Inc.
LandOfFree
Cyclic time averaging for machine monitoring does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Cyclic time averaging for machine monitoring, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Cyclic time averaging for machine monitoring will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3212028