Trouble-shooting a vibration test system requires a thoughtful approach when something goes wrong. A vibration test is a controlled test; the user determines the test to be performed from field measurements or specified requirements and programs them into some type of controller. It is the job of the controller to overcome the characteristics of all of the individual test system components and produce a drive signal that results in the desired test level at all frequencies. Since it is a closed-loop system, it will behave like any feedback loop and a problem anywhere in the loop will appear at all other points as the controller tries to correct, making troubleshooting difficult; it is usually best to break the loop and test the controller by itself and the rest of the system separately.
The
obvious first question in the troubleshooting process is to ask when was the
last time it worked if the system has been previously used, and what has been
changed since then, and try to repeat what last worked. Shaker systems are
expensive and easy to damage without a cautious approach, and this cannot be
stressed too much! There are many factors that can affect the outcome of a
test, and here are a few of the common ones:
1
) CONTROLLERS: Whenever there is a
problem running a vibration test the controller is always the first thing
questioned, but it is rarely the source of the problem. The safest way to start
is to check out the controller with a self-test or closed-loop test performed by
simply connecting the drive output to the Channel #1 input and starting the
test. If it cannot successfully complete this simple test you can be assured
that it will not run a real test and the problem lies within the controller.
This should be run before being connected to the shaker system.
If the closed-loop test passes, you can assume the problem probably lies elsewhere and the common approach is to drive the amplifier with a 200 Hertz sine source and slowly increase the gain until a tone is heard from the shaker or significant armature current is indicated; then one of the following factors is probably the problem, which need to be investigated one at a time:
2
) SYSTEM LIMITS: All vibration systems have physical and
electrical limits beyond which the equipment will be damaged; these safety
limits should be set up in the System limits menu; the three limits and their
causes are as follows:
A ) Displacement- The peak-to-peak maximum travel of the armature before it hits the stops, or preferably the limit switches, expressed in Mils or MM; this is the physical limit of travel of the shaker and is only valid if the center of motion remains stationary and does not creep. A test specified as 0.5” D.A. means 500 mils p-p ( Double Amplitude ).
B ) Velocity- When the armature coil is moving, it generates a back EMF voltage; this velocity voltage must be overcome by the amplifier, and thus the maximum voltage of the amplifier determines the velocity limit, expressed in Inches per second ( IPS ) or MM/sec.
C ) Acceleration- Either the physical limit where the coil may come apart into a slinky, or the maximum current rating of the amplifier, whichever comes first, usually expressed in normalized units of G-peak or m/sec^2.
3) ACCELEROMETERS: Placing control accelerometers on a flimsy test article or fixture will result in a failed test. I recommend always placing the channel #1 control accelerometer on the armature table center if possible and others on the fixture; use average or extremal control as desired, but rely on channel #1 as your main source of control. It is best to use only channel #1 for control and use other channels for monitoring, at least initially until the behavior of the other channels is established. Most tests now use ICP accelerometers which have a built-in charge amplifier and are powered by a current source on the signal line, typically around 4 MA with an open circuit voltage of 18 VDC or higher; a very common problem is that the current is under software control and the user has not enabled it, and thus there is no feedback; setting the wrong sensitivity is another common problem, as are bad cables. The output of the control accelerometer can be “tee”d to a scope to monitor the response; the response should correspond with what is expected for the current drive level.
4
) FIELD SUPPLY: Electro-dynamic shakers utilize a field coil
to generate the magnetic field, except for smaller ones of 100# force or
smaller, which use permanent magnets. Power levels of several hundreds of volts
and amps are common on larger sizes, and if the field supply is somehow not
turned on, large currents can flow through the armature with very low
accelerations being noticed. The field supply should always be interlocked so
that the amplifier cannot drive without it being present.
5
) FIXTURE DESIGN: The design of a good test fixture is as
difficult as an armature design itself and it’s importance is frequently
under-estimated. Flexing and various resonant modes will limit the test
capabilities. I recommend starting with a solid block of aluminum or preferably
magnesium and milling out un-needed sections; machined flat surfaces with
adequate mounting hardware are a must. A modal analysis should be done in
advance to avoid surprises during testing. Even well-designed fixtures will
have difficulty above 2000 Hertz and transmitting the drive through the D.U.T.
mounts is even more difficult. Mounting the control accelerometer on the shaker
table will yield the best results; averaging multiple accelerometers mounted on
the fixture is the second choice. In my opinion all armatures and test fixtures
should be made of magnesium and efforts made to reduce the Q of resonances,
either by physical design or applying damping materials wherever possible.
6 ) NOISE: The noise level in a system will determine
the lowest level test that can be run, no matter whether it be Sine or Random;
noise can be reduced by using a short cable between the controller and
amplifier and tightly coupling their grounds, or adding an isolation
transformer in extreme cases. Noise can also be picked up in the accelerometer
base and cabling and can be reduced with an isolation mount for the
accelerometer. Noise is almost always worse with switching amplifiers and
increases with the size. High sensitivity accelerometers will help and 100 MV/G
or higher should be used for low-level testing. Grounding the controller and
all instrumentation to the amplifier may help.
7
) AMPLIFIER DISTORTION: Rarely even considered, distortion in the
amplifier will limit the available dynamic range for the test; for example 1%
distortion will put a 40 Db limit on most random tests. The least signal that a
controller can put out at any given frequency is zero; if the amplifier
generates it’s own signals from distortion components that are 40 Db below the
desired one, then that establishes the lower limit of the test.
8
) SHAKER DISTORTION: Shakers typically have high distortion at
low frequencies due to the flexure spring constant and body motion, but this does
not usually pose a problem unless it is a low frequency test. The axial
resonance is a far worse problem since any distortion components at that
frequency will be multiplied by the Q of the resonance; thus an amplifier with
1% distortion and a shaker with a resonance Q of 100 ( 40 Db peak ) may result
in 100% acceleration distortion when driven at 1/3 or 1/5 of the resonant
frequency. This will again reduce the available dynamic range. If the shaker
armature has rotary or cross-axis motion ( and they all do, sometimes as great
as the desired axial motion ), the test article will be excited in various ways
and the result will be un-controllable motion since the controller did not
generate it.
9
) TEST PROFILE SHAPE AND
FILTERING: Some tests have spectral
shapes that are difficult to run as they use up all the computational power of
the DSP in the controller. In particular, shapes with steep or vertical slopes
should be avoided since they require extremely high-order filters to simulate.
Gentle shapes without peaks and notches will be easier to run.
All sampled-data systems utilize the discrete Fourier transform, or DFT to accomplish the spectral analysis; the DFT suffers from it’s very nature that it is discrete, not continuous, and that errors result from quantization and finite starting and stopping points. These errors tend to smear the results and are most noticeable when the spectrum has steep slopes and sharp transitions. Volumes have been written about various filtering ( windowing ) methods to lessen the errors, and it is both art and science choosing the best one for a given application. In general simple smooth tests may perform adequately with a rectangular filter, which is really no filter at all; tests with complicated shapes may benefit from one of the popular mathematical Hamming, Hanning or Blackman filters. I have found that the Hanning filter frequently yields the best results.
Rev
2/12/08