Condition monitoring works because most machines do not fail without warning. They change first: they vibrate differently, run a little hotter, draw more current for the same job. The skill is knowing which signal shows which problem, and how early.
Vibration
Vibration is measured with accelerometers or velocity sensors, usually on the bearing housings.
What it catches
- Unbalance: a strong component at the running speed (1×).
- Misalignment: components at 1× and 2× the running speed, often in the axial direction too.
- Looseness: many harmonics of the running speed.
- Bearing damage: energy at higher frequencies, visible in specialised analysis such as the envelope spectrum, often before the machine sounds or feels different.
- Gear problems: components at gear-mesh frequencies and their sidebands.
How early: among the earliest warnings for rotating machines, often weeks before a bearing fails, but not always.
Caveats
- The trend matters more than a single reading. Compare the machine with its own normal.
- Overall vibration levels have standards: ISO 20816 defines zones by machine class and size. Use them as a reference, not as the only rule.
- Sensor position and mounting change the reading. A sensor that is moved changes the baseline.
Temperature
What it catches: overload, poor lubrication, blocked cooling, failing bearings and, in motors, hot windings. Thermal equipment shows its health in temperature spread and in how much power it needs to hold a setpoint.
How early: later than vibration. Temperature reacts to friction after the damage is already significant, and thermal mass slows the response.
Caveats
- Compare at the same load and ambient temperature. A hot day is not a fault.
- The rate of rise is often more telling than the value.
- Winding temperature protects the motor, but if it trips, the line has already stopped.
Motor current
What it catches: the electrical load of the motor shows the mechanical load of the machine.
- Current rising for the same output points to friction, a blocked flow or a worn component.
- Unequal current between phases points to supply or winding problems.
- Detailed current signature analysis can reveal rotor problems and some mechanical faults, but needs more specialised work.
How early: moderately early for gradual mechanical degradation, and immediate for electrical problems.
Caveats: the current depends on the process load, so interpret it against output, speed and torque.
Process signals: pressure, flow, cycles
- Pumps: a pump that delivers less flow or pressure at the same speed than it used to is wearing or cavitating. Compare against its performance curve.
- Compressors: shorter load cycles with faster pressure drop are a typical sign of a leak or worn valves; oil temperature and vibration add detail.
- Conveyor drives: current rising while belt speed slips points to belt tension or a bearing on a roller.
Putting signals together
A single signal gives hints; several signals together give confidence.
| Signal | Typical problem it shows | How early | Main caveat |
|---|---|---|---|
| Vibration | Bearings, unbalance, misalignment, looseness | Early | Needs good sensor placement and a baseline |
| Temperature | Overload, lubrication, cooling, late bearing damage | Later | Depends on load and ambient |
| Motor current | Mechanical friction, electrical faults | Medium | Depends on process load |
| Pressure and flow | Pump wear, cavitation, leaks | Medium | Compare with the performance curve |
| Cycle counts | Leaks, valve wear | Medium | Needs the machine's normal pattern |
Always add context: load, speed, ambient temperature. Without it a normal change in operation looks like a fault, and a real fault can hide behind a change in operation.
What monitoring cannot see
- Sudden failures such as a shorted winding or a broken shaft from a shock load.
- Faults in parts that are not measured. If there is no sensor near a component, nothing will warn you about it.
- Problems in the process itself that are not visible in the machine's own signals.
Good condition monitoring is honest about this. It reduces surprises on the failure modes that give warning; it does not remove all failures.
Where to start
Pick one critical machine, find out which of these signals you already have in your controller or drive, add vibration where it is missing, record normal behaviour for a couple of weeks and review every alert with the person who knows the machine best.