Condition threshold alarms
Fixed alarms on individual sensor readings, such as vibration RMS or bearing temperature, that flag an asset once it crosses a set point.
Part of the recommended approachTechnique: Rules Engine
Condition threshold alarms are the first layer of unplanned-downtime prevention, checking each incoming sensor reading against a set point reliability engineers already agree marks the start of trouble.
They catch what a technician on a walk route would catch by eye: bearing temperature running hot, vibration RMS climbing on a rotating shaft, or oil particulate counts rising ahead of a scheduled change.
This is the standard starting point on any instrumented line, because these alarms give technicians and planners an explainable, auditable floor that no model score alone provides.
They stay in place even after a failure risk model is added, since a threshold on a named sensor can say exactly why it fired in a way a blended score cannot.
They are the wrong tool once degradation only shows up as a combination across sensors, since no single threshold on a single sensor will ever catch a joint pattern.
Retune them on a fixed schedule rather than leaving them at their commissioning values, since equipment condition and setpoints both move over an asset’s life.
Strengths for this problem
- Catches known failure signatures immediately, such as bearing temperature climbing past the point that has historically preceded seizure.
- Every alarm is explainable in the terms a technician already uses: which sensor, which threshold, which asset.
- Cheap to stand up against instrumentation that is already collecting vibration, temperature, current, or particulate data.
Limitations for this problem
- Only catches degradation patterns someone has already characterized and set a threshold for.
- Thresholds set at commissioning drift out of date as equipment ages or setpoints change, producing false alarms or missed ones.
- Does not combine weak signals across sensors, so a fault that only shows up as a joint pattern across vibration and current goes uncaught until a single threshold is crossed on its own.
Data needed
- Sensor fields: vibration RMS, bearing temperature, spindle motor current, and oil particulate counts per asset.
- Asset criticality rating, to decide whether an alarm routes to an immediate response or the next planned maintenance window.
- Threshold change history, so a recalibration is distinguishable from a genuine shift in equipment condition.
How it is measured
- Firing rate and precision per individual threshold, not the alarm set as a whole.
- Share of confirmed failures that were preceded by an alarm with enough lead time to act.
- Rate of threshold changes made without a documented justification.
Pitfalls
- Leaving thresholds uncalibrated for years after commissioning, so alarms fire on drift rather than genuine degradation.
- Writing a threshold around one past failure, which stops adding value once that specific failure mode has been designed out of the line.
- Loosening a threshold to cut alarm volume without checking what real failures it would then miss.
Human oversight
A maintenance planner or reliability engineer reviews every alarm before scheduling a response, and no alarm stops a line or triggers a teardown automatically.