Vibration time history data (g vs time) from actual field environments has been available for many years. But today, the major challenge facing test engineers is how to replicate this acceleration time history in the lab using a vibration test system.
Most vibration test procedures have relied on some form of approximation in the process of specifying realistic test conditions and are only simulations of the actual service environments. Historically, these test conditions have been less than representative of the actual service environment, primarily due to limitations in vibration test-equipment technology.
Early vibration testing was based on fixed-frequency, sinusoidal motion using shakers driven by simple sine oscillators. As sine controls became more sophisticated, fixed-frequency tests gave way to swept frequency tests, and variable amplitude tests became possible with the introduction of analog servo controls.
As controllers evolved, random vibration entered the picture. Generating many frequencies simultaneously, with randomly varying g levels at each frequency, was achieved using a white noise source to drive the shaker system.
Unlike sine testing, random vibration excited all mechanical resonances simultaneously in the applied vibration spectrum. This represented a major advance in test realism. The introduction of digital controllers extended the capability to approximate the real vibration conditions that exist in field environments.
Yet even after these major advances, a key limiting factor remained. Random- vibration test specifications still were based on field-vibration data acquired using a spectrum analyzer.
To obtain a technically valid measurement of a randomly varying signal, a spectrum analyzer must make many instantaneous signal measurements over time, then combine these multiple measurements using a time-averaging method to produce a statistical approximation of the vibration spectrum. As an inherent part of the spectrum-averaging method, infrequently occurring vibration events are discounted in terms of their contribution to the overall time-averaged spectrum measurement.
This discounting of potentially important—but infrequently occurring vibration events—is an unavoidable result of the spectrum-analysis method. As long as vibration test specifications rely on spectrum-analyzer data alone, infrequently occurring vibration conditions that exist in the actual service environment will be de-emphasized or possibly even lost.
Where’s the Washboard?
Let’s look at a test case that clarifies the basic difference between a spectrum-analysis approximation and a time history measurement of actual road vibration data. A 1995 Ford Bronco was instrumented with four accelerometers placed at widely separated locations on the vehicle. The vehicle was driven at a constant speed over a 100-yard test track (Figure 1).
The test track included a 20-yard gravel segment, a 10-yard washboard segment, and 70 yards of unimproved road including ruts, various gravel sizes, and potholes. The vehicle was driven at approximately 15 mph, corresponding to a total drive time of about 18 s for a complete run.
Vibration signals from four locations on the vehicle were recorded using an Unholtz-Dickie PACS Portable Data Acquisition System equipped with Time Replication Acceleration Control (TRAC) software. This unit can record up to 20 min of real-time field data per test run.
In the interest of brevity, only one accelerometer (channel # 4) will be discussed. It was located on the transfer-case support bracket and measured the vertical motion, perpendicular to the plane of the track. This vibration data was acquired using two methods: spectrum analysis and acceleration time history.
Spectrum Analysis
The vibration spectrum analysis in Figure 2 indicates the average acceleration power spectral density measured at accelerometer # 4 for one complete run of the test track. The overall rms acceleration value for this spectrum measurement equals 0.653 g rms.
The recorded rms acceleration is a measure of the total vibration energy over the entire frequency spectrum. This time-averaged acceleration spectrum indicates higher vibration energy in the low-frequency portion of the spectrum, but there is no way to determine if this spectrum shape was the result of:
Many evenly distributed road irregularities.
Mostly smooth pavement mixed with several huge potholes.
Some combination of these conditions.
The spectrum analysis presents only an average of many instantaneous spectrum measurements made over the entire time interval. Acceleration events that do not occur very often are discounted. Acceleration events that occur many times during the test run dominate the spectrum measurement.
Acceleration Time History
The acceleration time history graph indicates the peak acceleration (g) vs time (seconds) measurement over the entire 100-yard test run (Figure 3). The upper graph of Figure 3 illustrates when specific acceleration events occurred during the 18-s (100-yard) test run.
Vibration levels exceeded 4 g over the washboard section of the track. Considerably lower vibration levels occurred during the initial gravel segment. The last 70 yards of the track also produced moderate instantaneous g levels, but an occasional acceleration spike can be seen in the 10- to 15-s portion of the time history display.
The lower graph in Figure 3 is the fast Fourier transform (FFT) of the Zoom Window positioned at the beginning of the washboard segment of the time history display. The FFT shows the frequency content of the washboard vibration, indicating a fundamental frequency of approximately 6 Hz and several harmonics of 6 Hz. This FFT analysis correlates well to the actual washboard spacing of approximately 36 in.
Spectrum Simulation or Real-Time Data
Let’s say you were assigned the job of designing a new automotive product or determining the cause of customer complaints following a recent product redesign. Would it be preferable to use the spectrum-analysis data from Figure 2 or the acceleration time history data from Figure 3? Based only on the information in Figure 2, would you expect your design to handle the higher-level washboard environment of Figure 3?
Duplicating the Test Track in the Lab
Knowing the actual service environment of a product is an important element in the design process. Whether this data is used to improve a design or develop a new one does not diminish the importance of knowing what vibration the product is exposed to in actual field service.
But measuring the service environment is only half of the solution. The complete solution uses the actual field data to develop test specifications for the product in the lab, whether this means scaling or modifying the data for accelerated life testing, product qualification, proof of design, or design conformance. To reduce design costs and minimize actual field testing, effective implementation of the service environment must be accomplished in the lab.
The acceleration time history data in Figure 3 was acquired using a TRAC software module. The acquired time history display essentially looks like a strip-chart record of the acceleration signal (acceleration vs time). Using a TRAC import utility, the acceleration time history data can be converted into a reference waveform. Then it can be used to duplicate the time history on a shaker system using real-time, closed-loop control.
With TRAC software controlling an electrodynamic shaker, the acquired acceleration time history data from the Bronco was duplicated on a shaker, compensating for fixture and hardware resonances (Figure 4). During the 18-s shaker test, the imported time history waveform was accurately reproduced. Selected portions of the waveform could be cycled and sequenced for enhanced durability testing. The lower zoom display highlights the washboard acceleration spikes.
Summary
Conventional random vibration tests are based on averaged data from spectrum analyzers that, by definition, discounts infrequently occurring events. Advanced methods of data acquisition and real-time, closed-loop shaker control now can perform actual time history vibration testing in the lab. Time history tests replicate the field environment without ignoring individual events, more effectively bridging the gap between the actual vibration environment and laboratory testing.
About the Authors
Michael Garofalo joined Unholtz-Dickie in 1984 as an electrical engineer after attending Western New England College. Currently, he is a systems application engineer based at the UD corporate headquarters. Unholtz-Dickie, 6 Brookside Dr., Wallingford, CT 06492, (203) 265-3929, e-mail: [email protected].
Phil Rogers is an electrical engineer with 25 years experience in vibration test- system applications at Unholtz-Dickie and a 1970 graduate of Yale. Mr. Rogers manages the UD western regional office. Unholtz-Dickie, 1874 S. Pacific Coast Hwy., #700, Redondo Beach, CA 90277, (310) 265-0927.
Copyright 1999 Nelson Publishing Inc.
July 1999