Title:
Operating Deflection Shape Analysis, An Overview
Source/Author:Robert J. Sayer
Product:
Me'scope
Technology: Vibration
Classification:
Introduction:
Operating Deflection Shape Analysis, hereinafter referred to as ODS Analysis, provides an
animation of the deflected shape of a vibrating system under actual operating conditions. ODS
testing can be performed with either a single channel or multi-channel analyzer, although the use
of single channel techniques limits the scope and affects the accuracy of the analysis. This
discussion shall be limited to multi-channel ODS methods.
Multi-channel ODS testing not only provides the animated deflected shape at the operating
frequency of the equipment but also provides animated deflection shapes at any other frequency.
Single channel ODS is limited to the animation of the deflection shape at lx operating speed of
the equipment or integer multiple thereof. Multi-channel ODS testing can be used to define the
response of a vibrating system to forces produced by drive belts, gear mesh, bearings, turbulence
and other aerodynamic affects which usually do not occur at an integer multiple of the operating
frequency.
The most common reasons for performing ODS testing and analysis include:
Defining areas of structural "weakness", including looseness in bases and bolted joints. This
information can be used to direct the analyst toward the area for which structural/mechanical
modifications will be most effective.
Prove that a natural frequency is actually being excited. If Experimental Modal Analysis (EMA)
testing is performed along with ODS testing and the animated mode shape associated with the
excitation of a resonance matches the animated ODS shape, then it can be concluded that
dynamic forces are present which are exciting a natural frequency.
Identify the nature and source of a troublesome dynamic force produced by the equipment.
Information relative to the magnitude, direction and type of force can be inferred from an ODS.
Obtain a mode shape for a natural frequency. EMA testing is usually the best method for
determining a mode shape associated with a natural frequency. However, there are cases where
tune and/or access restrictions do not permit the performance of a complete modal test.
Test Procedure:
The test procedure for multi-channel ODS analysis is very similar to EMA testing, except that a
reference transducer is used in place of an impact hammer or shaker. An ODS test is performed
while the equipment is operating as opposed to an EMA test which can be performed only if the
equipment is removed from operation. EMA software can be used to curve-fit and animate ODS
data.
Minimum test equipment required to properly perform a multi-channel ODS includes at least two
transducers (accelerometers or velocity transducers); a multi-channel analyzer such as CSI
Wavepak, CSI Consultant, or CSI Model 2400; and all necessary cables and power supplies for
the transducers. A digital tape recorder or multi-plexor is not required but provides a luxury that is
sometimes necessary when field conditions limit the amount of time during which an entire test
procedure must be completed. Case History # 1 presents a case where data collection for an
ODS test containing 64 data points had to be completed in less than two minutes!
The reference transducer remains stationary during the entire test. It should be placed near a
location and in a direction of maximum vibration. Operating data is obtained by roving with a
response transducer amongst all of the other data points. It is important to note that one set of
data should be obtained with the reference and response transducers at the same location and in
the same direction.
The amount of data required to accurately construct an ODS is dependent upon the vibrating
system to be analyzed. An understanding of structural dynamics together with experience in
performing ODS tests is required to define the minimum data necessary for the test. Absent this
experience, it is recommended that the novice analyst obtain as much data as possible. Since the
ODS test is performed with the equipment operating, the extra time required to obtain extra data
should not affect the productivity of the equipment, although the same cannot be said for the
productivity of the analyst.
The Transfer Function of transmissibility ('lps/ips for velocity transducer & g's/g's for an
accelerometer), which is calculated by the multi-channel analyzer, is used to determine the
relative magnitude of the deformation at any data point. Transient peaks and beats do not affect
the ODS since transmissibility is used instead of the actual vibration level. If the vibrating system
is nearly linear, the vibration level at the response transducer will increase or decrease in
proportion to the reference transducer.
Curve Fitting & Animation:
The ODS shape can be developed manually by plotting the transmissibility and phase at each
data point for every frequency of concern. Each data point should contain three pieces of
information; magnitude of transmissibility, direction and phase. The phase relationship between
data points can be obtained from the phase relationship of each of the individual data point
responses with that of the reference transducer.
The time required to manually develop an ODS can be prohibitive, especially for tests containing
a large amount of data and/or for analyses requiring ODS shapes at several different frequencies.
Commercially available EMA software can be used to simplify this procedure. Besides the
obvious advantage of saving time, EMA software also provides an actual moving animation of the
ODS shape which can be rotated in space, zoomed and magnified to provide a closer and more
detailed view of a particular area of concern. It can be superimposed upon the undetected shape
to provide a better perspective of relative motion. EMA software contains many other advantages
that computer generated graphics afford that a manually plotted sketch cannot provide.
However, the analyst must be careful when using the curve fitting subroutines resident in
commercially available EMA software. These program offer a variety of curve fitting algorithms
which can generally be classified as either "single degree-of-freedom" or "multiple degree-offreedom" techniques. ODS curve fitting should be restricted to the use of the "single degree-offreedom" algorithms; coincident, quadrature or complex peak methods. The use of the "multiple
degree-of-freedom" curve fitting techniques, such as the polynomial or global methods, can result
in major errors in the ODS animation.
The problem with using the higher order curve fitting methods is that the Frequency Response
Function of transmissibility, which is sometimes referred to by EMA analysts as a "pseudo" FRF,
contains many "junk peaks" associated with the noise floor of the reference transducer. Peaks
occur in this FRF at frequencies other than that at which maximum response occur since this
"pseudo" FRF is obtained by dividing the frequency spectrum of the response transducer by that
of the reference transducer.
Figure 2 illustrate the problem associated with curve fitting ODS data. It contains a typical FRF
obtained from a transmissibility measurement. The cursor has been assigned to the machine
operating speed (30 Hz) which also corresponds to the maximum response of the vibrating
system. However, peaks in the FRF are present wherever the magnitude of vibration in the
reference transducer were very low, since dividing any number by a much smaller number
provides a large number. The use of the higher order multiple degree-of-freedom curve fitting
methods, which search for peak responses in the FRF, would actually provide an ODS at the
frequency of one of these "junk" peaks instead of at the frequency of concern.
plexers to serve as a common reference signal. Two computers equipped with CSI WAVEPAK
were used to control the multiplexers. A computer program was written which automatically
provided the keystrokes in Wavepak to obtain and archive the Transfer Functions and advance
the multi-plexor. A complete set of ODS data was obtained for each test in approximately 2
minutes! However, it should be noted that the reduction of this data took several man-days.
Figure 2 includes one of the ODS shapes Obtained from the data. This analysis provided some
interesting information regarding the relative motion between the work rolls. This relative motion
can affect the deformation process, resulting in a chatter pattern being established in the finished
coil. The size of the chatter pattern will be dependent upon the frequency of the vibration.
interesting to note that the motor was 180 out-of-phase with all of the other components of the
system
second mode of natural frequency was 15.3 Hz. In this mode shape, the motor is 180 out-ofphase from the rind body rocking action of the rest of the system, which correlated quite well with
the ODS at this frequency.