31.2 Distribution Qualification

Back

The term “distribution qualification” denotes the process by which the ability of a product to withstand a

clearly defined distribution environment is established. Dynamic effects on the product due to handling

loads, characteristics of packaging, and excitations under various modes of transportation (truck, rail,

air, and ocean) must be properly represented in the test specifications used for distribution qualification.

If a product fails a qualification test, then corrective measures and subsequent requalification are

necessary prior to commercial distribution. Product redesign, packaging redesign, and modification of

existing shipping procedures might be required to meet qualification requirements.

Often, the necessary improvements can be determined by analyzing data from prior tests. Proper

distribution qualification will result in improved product quality (and associated reliability and

performance), reduced wastage and inventory problems, cost-effective packaging, reduced shipping and

handling costs, and reduced warranty and service costs.

31-1

© 2005 by Taylor & Francis Group, LLC

Random testing can more accurately represent vibrations in distribution environments. Several

characteristics make it superior to sine testing. A sine test is a single-frequency test; thus, only one

frequency is applied to a test object at a given instant. As a result, failure modes caused by the simultaneous

excitation of two or more modes of vibration cannot be realized by sine testing, at least under steady

excitations. On the other hand, in random testing, many frequencies are simultaneously applied to the

test object. Therefore, conditions are more conducive to multiple-mode excitations and associated

complex failures. A comparison of testing with four types of excitation signals is given in Table 31.1.

31.2.1 Drive-Signal Generation

The first step in signal synthesis for driving the exciter is to assign independent and identically

distributed, random phase angles to the digitized spectral magnitude (spectral lines) of the drive

spectrum. The number of lines chosen is consistent with the fast Fourier transform (FFT) algorithm that

is employed and the desired numerical accuracy. The inverse Fourier transform is obtained from the

resulting discrete, complex Fourier spectrum. In general, the signal thus obtained would not possess

ergodicity and Gaussianity.

Stationarity can be attained by randomly shifting the signal with respect to time and summing the

results. The resulting signal would also be weakly ergodic. Ergodicity is improved by increasing the

duration of the signal. To obtain Gaussianity, a sufficiently large number of time-shifted signals must be

summed as dictated by the central limit theorem. Furthermore, because the magnitude of a Gaussian

signal almost always remains within three times its standard deviation (99.7% of the time), Gaussianity

can be imposed simply by windowing the time-shifted signal. The amplitude of the window function is

governed by the required standard deviation of the drive signal. Unwanted frequency components

introduced as a result of sharp end transitions in each time-shifted signal component can be suppressed

by properly shaping the window. This process introduces a certain degree of nonstationarity into the

synthesized signal, particularly if the windowed signal segments are joined end-to-end to generate the

drive signal. A satisfactory way to overcome this problem is to introduce a high overlap from one segment

to the next. Because the processing time increases in proportion to the degree of overlap, however, a

compromise must be reached.

In summary, for a given drive spectral magnitude the drive signal can be synthesized as follows:

1. Assign independent, identically distributed random phase values to the drive-spectral lines.

2. Perform an inverse Fourier transform of the resulting spectrum using FFT.

3. Generate a set of independent and identically distributed time-shift values.

4. Perform a time-shift of the signal obtained in Step 2 using the values from Step 3.

5. Window the time-shifted signals.

6. Join the windowed signals with a fixed overlap.

The resulting digital drive signal is converted into an analog signal using a digital-to-analog converter

(DAC) and passed through a low-pass filter to remove any unwanted frequency components before it is

used to drive the shaker. This procedure is illustrated in Figure 31.1.

TABLE 31.1 Comparison of Test Types

Sine

Testing

Random

Testing

Narrowband

Random Sweep

Broadband

Random Sweep

Simultaneous multimodal (multiresonant)

excitation possible?

No Yes No Yes

Test duration Long Short Long Moderate

Power requirements Low High Low High

Represents a random environment? No Yes Yes Yes

Test system cost Low High Moderate to high High

Overtesting possibility High Low High Low

31-2 Vibration and Shock Handbook

© 2005 by Taylor & Francis Group, LLC

31.2.2 Distribution Spectra

The distribution environment to which a product is subjected depends on several factors. In particular,

one must consider (1) the nature and severity of handling prior to and during shipment, (2) the mode of

transportation (truck, rail, air cargo, ship), (3) the geographic factors, (4) the environmental conditions,

(5) the characteristics of the protective packaging used, and (6) the dynamic characteristics of the

product itself. These factors are complex and essentially random in nature. Laboratory simulation of such

an environment is difficult even if a combination of several types of tests — e.g., vibration, shock, drop,

and thermal cycling — is employed. A primary difficulty arises from the requirement that test

specifications should be simp1e, yet accurately represent the true environment. The test must also be

repeatable to allow standardization of the test procedure and to facilitate evaluation and comparison of

test data. Finally, testing must be cost effective.

During transportation, a package is subjected to multi-degree-of-freedom (multi-DoF) excitations

that can include rectilinear and rotational excitations at more than one location simultaneously.

However, test machines are predominantly single-axis devices that generate excitations along a single

direction. Thus, any attempt to duplicate a realistic distribution environment in a laboratory setting can

prove futile.

An alternative might be to use trial shipments. However, because of the random nature of the

distribution environment, many such trials would be necessary before the data would be meaningful.

Therefore, trial shipments are not appealing from a cost – benefit point of view and also because test

control and data acquisition would be difficult. Data from trial shipments are extremely useful, however,

in developing qualification-test specifications and in improving existing laboratory test procedures.

A more realistic goal of testing would be to duplicate possible failure and malfunction modes without

actually reproducing the distribution environment. This, in fact, is the underlying principle of testing for

distribution qualification. For instance, sine tests can reproduce some types of failure caused during

shipment even though the test signal does not resemble the actual dynamic environment; however,

random testing is generally superior.

Test specifications are expressed in terms of distribution spectra in distribution qualification where

random testing is used. Specification development begins with a sufficient collection of realistic data.

FIGURE 31.1 The synthesis of a random drive signal.

Seismic Qualification of Equipment 31-3

© 2005 by Taylor & Francis Group, LLC

Sources of data include field measurements during trial shipments, computer simulations (e.g., Monte

Carlo simulations), and previous specifications for similar products and environments. For best results,

all possible modes of transportation, excitation levels, and handling severities should be included. The

data, expressed as power spectral density (PSD), must be reduced to a common scale — particularly with

respect to the duration of excitation — for comparison purposes. Scaling can be accomplished by

applying a similarity law based on a realistic damage criterion. For example, a similarity law might relate

the excitation duration and the PSD level such that the value of a suitable damage function would remain

constant. Time-dependent damage criteria are developed primarily on the basis of fatigue – strength

characteristics of a test product.

Owing to nonlinearities of the environment, spectral characteristics (frequency content) will change

with the excitation level. If such changes are significant, then they should be properly accounted for. The

influence of environmental conditions, such as temperature and humidity, must be considered as well.

The PSD curves conditioned in this manner are plotted on a log – log plane to establish an envelope curve.

This curve represents the worst composite environment that is typically expected. The envelope is then

fitted with a small number of straight-line segments.

At this point, the PSD curve should be scaled so that the root-mean-square (RMS) value is equal to

that before the straight-line segments were fitted. The resulting PSD curve can be used as the test

specification. Test duration can be established from the time-scaling criterion. If the corresponding test

duration is excessively long, thereby making the test impractical, then the test duration should be

shortened by increasing the test level according to a realistic similarity criterion.

Product overtesting can be significant only if one reference spectrum is used to represent all possible

distribution environments. Shipping procedures should thus be classified into several groups depending

on the dynamic characteristics of the shipping environment and a representative reference spectrum

should be determined for each group. If a range of products with significantly diverse dynamic

characteristics is being qualified, then reference spectra should be modified and classified according to

product type. At the testing stage, a reference spectrum must be chosen from a spectral database depending

on the product type and applicable shipping procedures. Alternatively, a general composite spectrum can

be developed by assigning weights to a chosen set of reference spectra and computing the weighted sum.

Vibration levels in land vehicles and aircraft can range up to several kilohertz (kHz). Ships are known

to have lower levels of excitation. In general, the energy content in vibrations experienced during the

distribution of computer products is known to remain within 20 Hz. Consequently, the test specification

spectra (reference spectra) used in distribution qualification are usually limited to this bandwidth. The

typical specification curve shown in Figure 31.2 can be specified simply from the co-ordinates of the

break points of the PSD curve. Intermediate values can be determined easily because the break points are

joined by straight-line segments on a log – log plane.

The area beneath the PSD curve gives the required mean-square value of the test excitation. The square

root of this value is the RMS value. It is specified along with the PSD curve, even though it can be

determined directly from the PSD curve. An acceptable tolerance band for the control spectrum —

usually ^ 3 db — is also specified. Test duration should be supplied with the test specification.

31.2.3 Test Procedures

Dynamic-test systems with digital control are easy to operate. In menu-driven systems, a routine or mode

is activated by picking the appropriate item from a menu that is displayed on the cathode-ray tube screen.

The system asks for necessary data, and then necessary parameter values are entered into the system.

Typically, the user supplies lower and upper RMS limits for test abort levels, break point co-ordinates of

the reference spectrum, and test duration. The tolerance bands for test spectrum equalization and the

accelerometer sensitivities are also entered. More than one test setup can be stored, a number being

assigned to denote each test.

Preprogrammed tests can be modified using a similar procedure in the edit mode. Any

preprogrammed test can be carried out simply by entering the corresponding test number. Computed

31-4 Vibration and Shock Handbook

© 2005 by Taylor & Francis Group, LLC

results, such as PSD curves and transmissibility functions, are stored for future evaluation. If desired,

these results can be displayed, printed, or plotted with proper annotations and scales while the test is

in progress.

Main steps of a typical test procedure are as follows:

1. Carefully examine the test object and record obvious structural defects, abnormalities, and

hazardous or unsafe conditions.

2. Perform a functional test (i.e., operate the product) according to specifications and record any

malfunctions and safety hazards.

Note: The test may be abandoned at this stage if the test object is defective.

3. Mount the test object rigidly on the shaker table so that the loading points and the excitation axis

are consistent with standard shipping conditions and the specified test sequence.

4. Perform an exploratory test at half the specified RMS level (one fourth the specified PSD level).

Monitor the response of the test package at critical locations including the control sensor location.

5. Perform the full-level test for the specified duration. Record the response data.

6. Change the orientation in accordance with the specified test sequence and repeat the test.

7. After the test sequence is completed, carefully inspect the test object and record any structural

defects, abnormalities, and safety hazards.

8. Conduct a functional test and record any malfunctions, failures, and safety problems.

An exploratory test at a fraction of the specified test level is required for new product models that are

being tested for the first time, or for older models that have been subjected to major design modifications.

Three mutually perpendicular axes are usually tested, including the primary orientation (vertical axis)

that is used for shipping. If product handling during distribution is automated, then testing only the

primary axis is adequate.

When multiple tests are required, the test sequence is normally stipulated. If the test sequence is not

specified, then it can be chosen such that the least-severe orientation (orientation least likely to fail) is

tested first. The test is repeated for the remaining orientations, ending with the most severe one. The

rationale is that with this choice of test sequence, the aging of the most severe direction would be

maximized, thereby making the test more reliable.

FIGURE 31.2 A reference spectrum for the distribution qualification of personal computers.

Seismic Qualification of Equipment 31-5

© 2005 by Taylor & Francis Group, LLC

The test report should contain the following:

1. Description of the test object: Serial number, size (dimensions and weight), product function (e.g.,

system unit, hard drive, power supply, printer, keyboard, mouse, monitor, and floppy disk drive of

a personal computer), and packaging particulars. Descriptive photos are useful.

2. Test plan: Usually standard and attached to the report as an appendix.

3. Test setup: Test orientations, sensor (accelerometer) locations, details of mounting fixtures, and a

brief description of the test apparatus. Photos may be included.

4. Test procedure: A standard attachment that is usually given according to corporate specifications.

5. Test results: Ambient conditions in the laboratory (e.g., temperature, humidity), pretest

observations (e.g., defects, abnormalities, malfunctions), test data (e.g., reference spectrum,

equalized control spectrum, drive spectrum, response time histories and corresponding spectra,

transmissibility plots, coherence plots) and posttest observations.

6. Comments and recommendations: General comments regarding the test procedure and test item

and recommendations for improving the test, product or packaging.

Names and titles of the personnel who conducted the test should be given in the test report, with

appropriate signatures, dates, and location of the test facility.

Tests for distribution qualification can be conducted on both packaged products and those without

any protective packaging, even though it is the packaged product that is shipped. However, the reference

spectra used in the two cases are usually not the same. The spectrum used for testing a product without

protective packaging is generally less severe. Response spectra that are used for testing an unpackaged

product should reflect the excitations experienced by the product during packaging.

Навигация