This is the preferred method and has been used extensively. Two techniques are available: (1) a general technique based on recording and analyzing the data obtained to develop the spectral responses through out the unit being screened; and (2) a simplified technique wherein overall gRMS level readings are obtained at the different sites to determine if some components are either overstressed or understressed.
The development of a random vibration stress screen is predicated on tailoring the input to achieve an acceptable response throughout the unit being screened. A vibration survey is the most logical and straightforward way to determine these responses. The spectral responses from selected accelerometer sites identify the frequencies where high responses or damping occur. The input vibration level at appropriate frequencies can then be tailored to eliminate undesired high or low responses.
The vibration survey configuration should replicate the configuration for the Proposed screen.
The item must be representative of the product to be screened. It should be possible to mount accelerometers internally within the item . It should be permissible to accumulate vibration time on the item
The vibration survey should be conducted at an input
random vibration level of 2-3gRMS, which is 6 to 10 dB below the baseline
screening level of Table 3-3
. A low level sine vibration sweep can also be used to obtain a very good picture of resonance responses across the desired spectrum.
The survey should be Performed for each input axis or combination of input axes specified for the screen. For instance, a screen performed by the sequential excitation of three orthogonal axes requires three surveys. A screen performed as the combination of a dual axes excitation and a single axis excitation requires two surveys. A triaxial input screen requires one survey.
The controller, control strategy, and the number and location of control accelerometers should be the same as for the proposed screen.
The excitation system used for the survey should be the same as for the screen.
The fixture, slip-plate, and head expander used for the survey should be the same as for the screen. It is good practice to perform a vibration survey on the mounting fixture only prior to the item survey.
Selection of Measurement Locations
In an exhaustive survey, vibration response would be measured at each component, wire connection, mounting screw, etc., within the item to be screened. This clearly is neither feasible nor desirable. What is desirable is to measure vibration responses at locations throughout the item that are representative of responses at a majority of the potential failure locations. Approximately 20 locations should suffice for mapping most items.
An example of an item is shown schematically in Figure
to illustrate the selection of response locations. The item is an electronics card box with cable connectors and a time meter mounted on the front panel, and transformers mounted on the rear panel. There are 11 standard cards spread throughout the box: four heavier, stiffer cards are located in the center and an encased, thick module is located at the rear. The cards and module have connectors on the bottom which mate with the motherboard at the bottom of the box.
The three measurement locations on the cards are
depicted in Figure
. The locations indicated by "X" are suggested for a rectangular PWA with components mounted uniformly over the surface, supported along the short edges and a connector on the bottom. If the top of the PWA is supported by compression of a rubber gasket on the lid, the locations depicted by "O" would perhaps be a better choice. A square PWA equally supported on each edge could be sufficiently mapped with two locations: one in the center and the other at the middle of one edge. Obviously there are many location choices within this example, and within other items that differ significantly from the example. This illustrates mapping of the entire volume and indicates that engineering judgment must be exercised in the selection of measurement locations.
Accelerometers should be small enough that they can be
mounted in the chosen location and light enough that they do not alter the
dynamic characteristics of the item. In most surveys a mix of accelerometer
types can be used. In the example shown in Figure
, relatively large, heavy accelerometers could be used to measure the acceleration input to the connectors/time meter on the front panel. Similar accelerometers could be used on the rear panel at diagonally opposite corners of the transformer. Medium size and weight accelerometers could probably be used on the motherboard and stiffer PWAs. The standard PWAs normally require the smallest, lightest accelerometers available so as to not alter the dynamic characteristics and to fit available mounting space.
The acceleration in three orthogonal directions must
be known for each chosen measurement location. This does not mandate a
triaxial measurement at each location. A measurement from another location may
be substituted for one of the triaxial measurements if the response is judged
to be the same over the frequency range of interest. As an example, triaxial
response at the three measurement locations depicted on the card in Figure
can be acquired by using five accelerometers. A single accelerometer is needed at each of the three measurement locations with the sensitive axis oriented perpendicular to the plane of the PWA. The in-plane response should be the same for all locations on the PWA and can be acquired by placing the two accelerometers wherever there is adequate space. Triaxial measurements will be required only if the contractor has equipment capable of triaxial excitation.
Accelerometers should measure the input to components
or parts, not the response of a particular component or part In the example
shown in Figure
, this means placing accelerometers on the PWAs, not the components, and on the front and rear panel, not the parts mounted to the panels .
It is assumed that the control and response acceleration data will be recorded and played back to a spectrum analyzer for data analysis. Alternatively, if the spectrum analyzer has enough data channels, the data analysis could be performed "on-line," obviating the need to record and later play back data for spectral analysis. If a recorder is not readily available, or if the number of available channels is limited, the survey can be accomplished in segments by analyzing the response of each available accelerometer and moving the accelerometer to another location or direction. In most cases this can be done quite efficiently with minimum impact to the overall survey.
Data Acquisition Equipment
The data acquisition system, i.e., accelerometers, signal conditioners, and recorder system, should have sufficient dynamic range to observe and record the response accelerations. The system should be compatible within itself and with the data analysis equipment.
The recorder speed should be sufficient to obtain the desired frequency response for the acquired data.
For the first data acquisition run in each survey, all control accelerometers should be recorded along with the response accelerometers. For all remaining data acquisition runs in each survey, one control accelerometer should be recorded with the response accelerometers. The control accelerometer should remain the same for all remaining runs to validate repeatability in case of questionable response data.
Documentation for the data acquisition should include the following information:
- Screen identification
- program name - item name screening station - recorder - engineer - date - excitation system
- Channel information
- accelerometer identification - accelerometer serial number - accelerometer sensitivity - charge amplifier gain - charge amplifier serial number
- Run information
- run identification - frequency range and level of excitation
The full scale g level of each channel should be estimated for each survey location prior to performing the data recording. This calculation or estimate will significantly reduce the instrumentation error caused by noise threshold or saturation.
A calibration signal, preferably a sine wave representing the full scale g level of the instrumentation, should be placed on each tape data channel. The run identification should note the voltage level, equivalent g level, and frequency of the calibration signal. The calibration should be recorded for at least two minutes after any changes in the patching of charge amplifiers to the recorder, and at any time that there is a question as to whether the input gains have been adjusted since the previous run.
It is also desirable for a broadband, approximately white noise, random signal to be recorded. The frequency range of the noise signal should extend over the frequency range of the excitation and its voltage amplitude should be within the dynamic range of the recorder. This signal, coming from one source, should be recorded simultaneously on all active data channels at the beginning of each run for a period of one minute. Record the true RMS voltage level of this signal during playback. This signal permits the frequency response of each channel and the transfer function between any two channels to be measured. Any discrepancies that are found can be compensated for during analysis.
Data Recording and Review
The minimum duration for recording of data should be the time necessary to calculate acceleration spectral density (ASD) functions over the desired frequency range, using 50 averages. This minimum time will vary, depending on the analysis block size and bandwidth, the number of channels processed simultaneously, and the analyzer computational speed. The entire run should be recorded if the screen is a nonstationary process. The data should be reviewed after the run to confirm that the amplitudes are appropriate, that the waveforms appear reasonable, and that the data segment is properly identified. The gain setting of each channel should also be verified
The end result of the vibration survey should be a collection of ASD functions on a mass storage device available for "massaging." ASD functions should be calculated for all control and response accelerometers.
Data Analysis Equipment
It is recommended that the data processing be performed by playing back the recorded data to a digital Fourier spectrum analyzer. The analyzer should have the capability to calculate ASD functions, label the functions, and store the functions and labels on a mass storage device such as disk or tape. Additionally, the analyzer should be able to retrieve a stored ASD, integrate the function over selected frequency ranges to obtain gRMS values, and print the gRMS values.
Data Analysis Parameters
ASD functions should be calculated with 50 averages.
An analysis bandwidth of approximately 5 Hz should be used for ASD calculation
over the frequency range of 20 Hz to 2000 Hz. Alternatively a constant
percentage bandwidth analyzer may be used if the bandwidth does not exceed
Each ASD function should be stored with a unique identifier. A data analysis log should record the run information and analysis parameters shown in Table 4-2
The following vibration survey procedure assumes that data is recorded on analog tape and played back to a spectrum analyzer for ASD calculation. The procedure can be modified for use with an on-line spectral analysis system.
The procedure also assumes that the excitation system is an electrodynamic shaker. For other types of excitation systems, not all steps will be relevant.
1. Record the calibration signal on all data channels of the tape recorder.
2. Record the white noise on all data channels of the tape recorder.
3. Attach any accelerometers and cables that require special treatment (disassembly of unit, cleanroom facilities, obstructions when installed in the fixture, etc.) to the unit.
4. Create or retrieve input specification on the controller.
5. Mount fixture to shaker table.Torque to specified values.
6. Mount control accelerometer(s) to fixture and patch to controller and data acquisition system.
7. Perform vibration dry run(s) to fixture and patch to the controller and data acquisition system.
8. Mount unit in fixture. Torque to specified values .
9. Attach remainder of response accelerometers and cables for this data run (attach accelerometers and cables for all runs if available).
10. Patch response accelerometers for this run to data acquisitions system.
11. Tap check all accelerometers to verify that they are properly patched to the input of the tape recorder and that all instrumentation functions properly.
12. Install all lids, covers, and unit cabling that will be on during screening.
13. Perform vibration run, recording all data.
14. Verify that the recorded data is valid before proceeding to the next run.
15. Repeat steps 9 through 14 for remaining groups of response accelerometers.
16. Repeat steps 4 through 15 for additional surveys, if applicable.
17. Analyze recorded data to obtain ASD functions. Label and store functions on the mass storage device for later retrieval and "massaging."
Compare vibration survey response spectra against allowable stress limit criteria applicable to the assembly under evaluation. Subsequent engineering analyses may result in appropriate hardware modifications to remove vibration screening concerns. In addition, tailoring of the input spectrum is a viable alternative for reducing response maxima to within allowable stress limits. However, because extensive tailoring can adversely affect the ability to stimulate defects throughout the entire assembly, it should be viewed as the exception, not the general rule. Where warranted, temporary stiffening or damping of the assembly should be considered to eliminate the need for tailoring.
The simplified vibration survey technique is a modification of the general technique. The general technique is based on recording and analyzing the data obtained to develop the spectral responses throughout the unit being screened, but there are many situations where neither the equipment nor the associated trained personnel are available to do this. For these situations the general technique can be modified so that only overall gRMS level readings are obtained at the different accelerometer sites throughout the unit being screened. Comparing these overall gRMS values will determine if some components are either overstressed or understressed due to structural resonances or damping, respectively. There is some risk that responses peculiar to random vibration may be missed.
If the mean RMS response derived from multiple
locations on an assembly are within +6, -3 dB of the input excitation level,
no tailoring may be required. Since overall level is only a crude indication
of spectral response, if the responses for individual locations differ
appreciably from the mean RMS level, a vibration response survey should be
conducted at an excitation level of 6 to 10 dB below the baseline screening
level of Table
(2 to 3gRMS).