Technical Note · Random Vibration Control
Why Did My Random Vibration Test Abort?
What each controller protection is watching for, how the timing of an abort points to its cause, and a symptom-by-symptom guide to getting the test running again without hiding the problem.
Key takeaways
- An abort is a protection working. Most report a real physical condition, so the first job is to read what the controller is telling you.
- When the abort happens is the best clue: loop check problems are wiring and setup; ramp-up problems are usually drive and low-level behavior; full-level problems are often resonance, nonlinearity or shaker limits.
- Sharp resonances and anti-resonances are the most common cause of lines out of tolerance. The fix is usually resolution, loop speed or the fixture, not wider limits.
- Coherence separates a control problem from a hardware problem: a local dip points to rattles and loose joints.
- Never disable control signal loss on an unattended test.
An abort is the controller doing its job
A random vibration controller holds the control accelerometer's spectrum on a reference profile by continuously adjusting the drive (see The Adaptive Control Loop). Around that loop sits a set of protections. Each watches for a condition in which continuing would over-test the article, damage the shaker, or produce a test result that could not be trusted. When one trips, the controller ramps the drive down and stops.
That makes an abort annoying, but rarely random. Knowing which protection tripped, and when, narrows the cause to a few candidates.
| Protection | What it checks | What it protects against |
|---|---|---|
| Loop check | Ambient noise on each input, then the response to a small drive | Open or wrong channels, dead amplifiers, wiring faults, before any real energy is applied |
| Control signal loss | Each control channel's rms against a level stored after the loop check | A loose or fallen accelerometer, a broken cable: the controller would otherwise raise the drive to compensate |
| Alarm and abort lines | How many spectral lines sit outside the alarm and abort tolerance bands | Under- and over-test at particular frequencies |
| Grms alarm and abort | Overall control level against the reference | Gross over- or under-test |
| Limit and abort channels | Response channels against their own limit profiles | Over-testing a sensitive part of the article |
| Shaker limits | Acceleration, velocity, displacement, drive voltage and force | Hitting the stops, overdriving the amplifier, exceeding the force rating |
| Input overload | Each input against its full-scale range | Clipped, invalid data being used for control |
When it happened is the first clue
Group the possibilities by phase. An abort during the loop check is almost always wiring, channel setup or amplifier state. During low-level equalization the usual suspects are lost control signals and parts of the band that need more drive than is available. During the level steps and at full level, resonance, nonlinearity and shaker limits take over.
At start-up: loop check failures
Before applying the profile, the controller measures the ambient noise on each channel with the drive off, then applies a small drive and checks that each control channel responds as expected. This is the cheapest place to find a fault: nothing significant has reached the test article yet.
| What you see | Usual cause | What to check |
|---|---|---|
| Channel too noisy | Ground loop, damaged cable, wrong coupling, or a noisy environment | Cable routing and grounding; the channel's coupling; the maximum-noise setting |
| No response at the control channel | Amplifier off, at zero gain or in fault; accelerometer unplugged; wrong channel assigned | Amplifier state and interlocks; the channel table; cable continuity |
| Response far from expected | Wrong sensitivity entered; wrong units; accelerometer on the wrong axis | Sensitivity and units in the channel table; sensor orientation |
| Loop check needs more drive than allowed | Low system gain: amplifier gain too low, or a heavy payload | Amplifier gain; the maximum loop check drive setting |
Sensitivity errors pass the loop check. An accelerometer entered at the wrong mV/g still responds, just at the wrong level, and the test will run at the wrong amplitude. A quick pre-test check at a low sine level, comparing channels mounted together, catches it (see Transducer Sensitivity Selection).
During ramp-up: drive and control signal loss
The controller learns the system's transfer function, H(f), from drive to control, and shapes the drive so that drive × H matches the reference. Where H is small, at an anti-resonance, the drive must be large. A deep anti-resonance can ask for more drive than the drive limit or the amplifier allows.
The other common ramp-up abort is control signal loss. If an accelerometer comes loose or a cable fails, the control signal drops. A controller without this protection would raise the drive to restore the level, and the shaker would drive the article and fixture far above the test level. Control signal loss detection compares each control channel against a level stored once the channel is well above its ambient noise, and stops the test if it falls away.
At full level: lines out of tolerance
Most random specifications give a tolerance on the control spectrum, often ±3 dB, and the controller adds alarm and abort bands around the reference. The controller counts the spectral lines outside each band; when the count exceeds the allowed number, it alarms or aborts.
A few lines out at the same frequency every run is the signature of a sharp resonance. A resonance with quality factor Q at fn has a half-power bandwidth of fn/Q (see Resonance). If the frequency resolution is coarse compared with that bandwidth, a single line spans the peak and the controller cannot shape the drive finely enough. A Q of 25 at 500 Hz is only 20 Hz wide; 400 lines over 2000 Hz give 5 Hz resolution, enough for four lines across it, while 200 lines give two.
Lines out at the very ends of the profile point the other way: toward the noise floor. Sloped profiles often fall to very low levels at the band edges, where the control signal may be close to the measurement noise (see Dynamic Range Considerations in Vibration Testing).
Is it control, or is it the hardware? Ask the coherence
Coherence measures how much of the control signal is explained linearly by the drive, from 0 to 1 (see Importance of Coherence in Vibration Control). A linear system measured well gives coherence near 1 everywhere. When it drops, the controller is trying to correct something it cannot predict.
A local dip, especially one that appears only as the level rises, usually means a nonlinearity: a loose bolt, a rattling component, a fixture interface that slips. No control algorithm can fix that; a torque wrench can. A broad drop at the top or bottom of the band means the signal there is near the noise floor.
Shaker limits and limit channels
Some aborts are not about the spectrum at all. A displacement or velocity limit trip means the profile's low end asks for more stroke or velocity than the shaker has, and a force warning means the moving mass, or the article's resonance, demands more force than the rating (see Shaker Performance Limits). A limit channel that takes control and then aborts is reporting that a response on the article has exceeded its allowed profile, usually at a resonance. Whether to notch the input there is a test-specification decision, not a controller setting to change on the fly.
The troubleshooting table
| Symptom | Likely causes | What to check first |
|---|---|---|
| Abort at the same frequency every run | A sharp resonance or anti-resonance the loop cannot follow; resolution too coarse for the peak | The transfer function; frequency resolution against the half-power bandwidth; loop speed and averaging |
| Abort only at full level, never at low level | Nonlinear behavior: loose bolts, rattling parts, fixture slip, armature or suspension nonlinearity | Torque on all fasteners; coherence at the failing frequency; listen for rattles |
| Lines out at the ends of the profile | The profile is close to the noise floor there; input range too coarse for low levels | Dynamic range: input range, accelerometer sensitivity, profile levels at the band edges |
| Drive limit or clipping reached | Anti-resonance or very low system gain in part of the band; amplifier gain too low | Drive spectrum against the transfer function; amplifier gain; sigma clipping setting |
| Displacement or velocity limit trip | Too much energy at the bottom of the profile for the shaker's stroke | Low-frequency end of the profile against the shaker envelope |
| Force limit warning | Total moving mass underestimated; resonance raises the article's apparent mass | Weigh fixture and expander; check force at the article's modes |
| Control signal loss | Accelerometer loose or fallen; intermittent cable or connector | Mounting, cable strain relief, connectors |
| Limit channel takes control, then aborts | A response exceeds its limit profile, often at a resonance | Whether the limit profile is correct; whether notching is permitted by the specification |
| Grms abort with spectrum in tolerance | Grms tolerance tighter than the line tolerance implies; energy outside the control band | The Grms alarm and abort settings; harmonics or noise above the control bandwidth |
Common misconceptions
| Misconception | Reality |
|---|---|
| "The controller aborted, so the controller is the problem." | Most aborts report a real physical condition: a loose accelerometer, a resonance, a profile beyond the shaker. The abort is the controller doing its job. |
| "Widen the abort limits and it will run." | That hides the symptom and can over-test the article. Fix the cause, or agree on a deviation with the customer. |
| "Turn control signal loss off in a noisy lab." | With it off, a fallen accelerometer leaves the controller driving harder to find a signal that is not there. Use it off only in attended, manual operation. |
| "If it ran at −6 dB it will run at full level." | Nonlinear joints, rattles and shaker limits all scale differently with level. Many aborts appear only in the last steps. |
| "More averaging always gives better control." | More averaging gives a smoother estimate but a slower loop, which struggles to follow sharp resonances and changing systems. |
Quick reference
| Check | Why |
|---|---|
| Read the abort message and note the frequency and phase of the test | The message and the timing usually name the cause |
| Inspect every accelerometer, cable and connector on the control and limit channels | Control signal loss and intermittent channels are the most common hardware cause |
| Re-torque fixture and article fasteners | Loose joints cause nonlinear behavior and low coherence at full level |
| Compare the profile with the shaker envelope | Velocity and displacement limits trip at the low end of the profile |
| Look at the drive spectrum and transfer function | Deep anti-resonances need large drive; check headroom |
| Check resolution against the narrowest resonance | Aim for several lines across the half-power bandwidth fn/Q |
| Check input ranges and sensitivities | Low-level profile ends need dynamic range; wrong sensitivities distort everything |
How Panther helps
- Loop check you can tune: Panther's loop check measures each channel's ambient noise against a maximum-noise setting and limits the loop-check drive to a maximum you set, so an open loop cannot run the drive away. A rigorous mode trades speed for a more thorough check, and an initial threshold test confirms the starting level is not too low.
- Control signal loss detection: in Standard mode Panther raises the drive until every control channel is 6 dB above its ambient noise, stores each channel's rms, and aborts if any control channel later falls more than 3 dB below its stored level. It is designed to catch an accelerometer that has fallen off the test article but is still bouncing on the fixture.
- Alarm and abort by line count: the number of lines allowed outside the alarm and abort tolerances can be set as a count or as a percentage of the lines, with separate Grms alarm and abort levels.
- Limit, abort and RMS-limit channels: limit channels reduce the drive when a response exceeds its profile and hand control back when it recovers, abort channels stop the test outright, and RMS limiting watches up to four frequency bands. The log records which channel was in limit or abort.
- Shaker limits: limits on acceleration, velocity, displacement and drive voltage abort the test if reached, and a maximum-force check using the test equipment weights can warn or prevent the test from starting.
- Peak control: Sigma Clipping limits the drive's peaks to prevent amplifier overloads, and kurtosis control sets the peak content of the drive.
- Headroom for difficult systems: 24-bit acquisition with more than 110 dB of dynamic range gives headroom for low-level profile ends, and selectable resolution, averaging and overlap processing balance detail against loop speed.
References and further reading
- J. S. Bendat and A. G. Piersol, Random Data: Analysis and Measurement Procedures, 4th ed., Wiley.
- C. M. Harris and A. G. Piersol (eds.), Harris' Shock and Vibration Handbook, 5th ed., McGraw-Hill.
- MIL-STD-810H, Method 514.8, Vibration.
- Spectral Dynamics Technical Library: The Adaptive Control Loop; Importance of Coherence in Vibration Control; Resonance; Understanding, Finding, & Eliminating Ground Loops.
