Vibration alarm limits sound simple, but the wrong setup creates noise that wastes time. Therefore, the goal is to catch real machine change early without turning your dashboard into a constant warning light. At DVA Industrial Solutions Inc., we see false alerts most often when teams skip baselining, mix operating modes, or copy generic limits across different assets.
Why false alerts happen in the first place
Most false alerts come from normal process variation being treated like a fault. That is to say, the machine is doing exactly what it always does, but the alarm threshold is too tight, applied to the wrong measurement, or compared against the wrong operating condition. Moreover, limits often get set right after a repair or startup when vibration behavior is still settling. If the initial data is unstable, the alarm will be unstable too.
Another common cause is using a single overall value to represent a complex machine. For example, overall velocity might rise slightly with load, yet specific defect frequencies remain clean. However, if the system only alarms on overall velocity, you will chase conditions that are not failures. After that, teams lose trust in monitoring and start ignoring alarms, which is the worst outcome.
Vibration alarm limits start with baselines
Alarm limits should be built from real operating data, not from a number in a handbook. Therefore, start by collecting a baseline set that represents normal operation: steady load, normal flow, typical product, and stable temperatures. In other words, your baseline should match the way the asset truly runs most of the time.
Aim for multiple baselines if the asset has distinct modes. For instance, a fan might run at different speeds, or a pump might switch between duty points. Similarly, a compressor may show different vibration behavior at each stage of throughput. If you only store one baseline, the alarms will trigger whenever the process shifts, even if the machine is healthy.
A practical approach is to define “operating states” and baseline each one. After that, apply limits that only compare data within the same state. This single step removes a huge portion of false alerts without hiding real defects.
Pick the right measurements before you pick the limits
The measurement you alarm on matters as much as the threshold. Therefore, avoid relying on only one overall value. Use a small set that reflects different failure types, while staying easy to maintain.
Overall levels for fast screening
Overall velocity or acceleration can work as a first screen, but it should not be the only trigger. For example, overall values are sensitive to process changes and structural resonance. Consequently, use them with wider limits and pair them with “confirming” indicators.
Frequency based indicators for defect confidence
Specific frequency bands or identified fault frequencies are better for confirming defects. That is to say, a bearing defect frequency, vane pass, gear mesh, or 1X running speed trend gives clearer meaning than a single overall value. If your program includes advanced diagnostics, predictive maintenance PdM workflows help tie alarms to actionable steps rather than guesses.
Start up and shutdown behavior needs its own rules
Transient conditions can create short spikes that look scary but are normal. Therefore, do not alarm the same way during ramps as you do at steady speed. Use separate limit logic and time filters for ramps, and consider targeted support like start up and shutdown monitoring technical assistance when machines are sensitive to resonance zones.
Use a tiered system instead of one hard threshold
Single thresholds create “on or off” thinking, and that drives false investigations. Therefore, set at least two tiers:
Alert level
Alert should mean “watch this trend” rather than “stop the machine.” In other words, it is a signal to check repeatability, confirm operating state, and review the spectrum.
Alarm level
Alarm should mean “high confidence change” with evidence. Consequently, you should require either persistence over time, a second confirming indicator, or a clear spectral signature before escalating.
A simple method is to set Alert around the upper edge of normal variability and set Alarm higher based on risk tolerance. After that, tune with real history, not opinions.
Add persistence, deadbands, and rate of change to reduce noise
Even good limits can chatter if the value hovers near the threshold. Therefore, add control logic that reflects how machines behave.
Persistence timers
Require the condition to exist for a set time or a set number of samples. For example, “3 out of 5 readings” avoids single sample spikes.
Deadbands and reset logic
If an alarm triggers at 6.0 mm/s, don’t reset it at 5.99. Instead, add a reset point lower than the trigger point. That is to say, use hysteresis so you do not bounce between states.
Rate of change alarms
Some failures show as a rapid trend, not a high absolute value. Therefore, consider alarming on change per day or per week. This catches new damage early, while ignoring steady, harmless variability.
Validate limits after balancing or structural changes
After maintenance, vibration patterns change, even when the result is good. Therefore, review baselines whenever you balance, align, replace bearings, or change foundations. If a rotor is corrected using dynamic field balancing, the new normal may be lower, and the old limits may now be too high to catch future drift.
Likewise, if you discover a resonance or a mode shape issue, you may need different sensors, points, or alarm bands. In that case, modal and ODS analysis helps separate structural response from true forcing, so your alarms track machine condition instead of the structure’s quirks.
A practical setup checklist you can apply today
- Define operating states and baseline each state.
- Choose 2 to 4 indicators: one overall, one or two frequency based, and one trend or rate of change.
- Set two tiers: Alert for review, Alarm for action.
- Add persistence, deadbands, and ramp logic.
- Re baseline after major maintenance or process changes.
- Document the response steps so each alarm has a clear next action.
- If you need help standardizing across sites, start with the methods and tools from DVA Industrial Solutions Inc. to keep limits consistent and defendable.
FAQs
How many alarm levels should we use for vibration monitoring?
Two levels usually work best. Therefore, an Alert level prompts review and trend checks, while an Alarm level triggers action when evidence is stronger and repeatable.
Should we use ISO limits as our alarm thresholds?
ISO guidance can be a starting point, but it should not be the final threshold. In other words, build limits from your own baseline data and operating states.
How do we stop alarms during startup from causing panic?
Use separate logic for ramps. Therefore, apply time filters, speed based rules, and steady state thresholds that only activate once the machine reaches normal operating speed.
What is the best way to handle machines that run at different loads?
Create multiple baselines by operating state. Consequently, the monitoring system compares like with like, which reduces false alerts when the process changes.
When should we rebuild our alarm limits?
Rebuild after major repairs, balancing, alignment, bearing changes, or structural modifications. Therefore, your limits continue to match the real “normal” behavior of the asset.