In this guide 6 sections
Count what wears the machine, then raise a service task when the count reaches its limit. Running hours suit engines and compressors, cycle counts suit presses and doors, and motor current shows load. It is a sound first step before any prediction, and it needs only a few signals.
Count the thing that wears
Calendar dates are a blunt trigger. Usage is sharper. A diesel generator that ran for two hours last month needs less than one that ran all week. Hours run tell you more than the calendar does.
Pick the unit that matches the wear. Running hours suit engines, compressors, pumps and gensets. Cycle counts suit presses, injection moulding machines, valves, lift doors and conveyor diverters. Metres or tonnes processed suit crushers and cutters.
- Hour meter on a diesel engine or compressor
- Stroke count on a press or stamping machine
- Open and close count on a door or valve
- Start count on a motor, which stresses contactors
- Total energy from a sub-meter on the supply
Take the signal from where it already exists
Many machines already keep these numbers. A PLC holds a cycle counter, a drive stores run hours in a parameter and an engine controller reports hours on its display. Read the existing register before fitting anything new.
Where nothing exists, derive it. A clamp-on current sensor on one phase shows when the motor is running. A proximity switch on a cam or a pulse from a limit switch gives a cycle. Failing that, wire an hour counter to a relay output.
Check the counter survives a power cut. Some small controllers reset to zero on restart. A reset silently wipes the service history.
Use current draw to tell running from idling
A machine can be powered and doing nothing. A compressor that is switched on but unloaded does not wear like one under full load. Motor current separates the two.
Set a threshold just above the idle reading and count time above it as running time. Set a second, higher band for heavy load and count that separately. You then have loaded hours and unloaded hours, which many service manuals treat differently.
A sudden rise in current at the same load, week after week, is also worth logging. It often points to a worn bearing or a clogged filter before a failure shows.
Turn a count into a task
Store a service limit against each asset: oil change at 500 running hours, a die inspection at a stated number of strokes, a contactor check at a start count. When the live counter passes the limit, create a work order in the maintenance system with the asset tag, the task and the counter value.
Add a lead point. Raise a planned task at nine tenths of the limit, so spares and a technician are ready, and an urgent one if the limit is passed. Close the task with the counter reading at completion, then restart the interval from that figure.
- Asset tag and location
- Counter name, unit and current value
- Service limit and the task it triggers
- Date and counter value at last service
- Technician and parts used
Keep it simple before predicting anything
Prediction needs months of clean data and a record of real failures. A usage trigger needs neither. It replaces guesswork with a rule people can read and challenge. That is enough.
Run it for a full service cycle, then compare. Did tasks arrive too early or too late? Adjust each limit using the technician’s notes. Those notes become the training set if you later want vibration or temperature models.
Start with your five costliest machines. Write down the counter, the limit and where the number lives for each one, and fit nothing until that list is done.
Usage trigger sheet
Print it as a table, one row per asset. Columns: asset tag, wear unit, signal source, register or sensor, service limit, lead point, task raised, last service date, counter at last service and technician’s note. A second block lists counters that reset on power loss.
Open a blank worksheet to printQuestions owners ask
Do I need new sensors to start?
Often not. Check the PLC, drive or engine controller first, since many already hold hours or cycle counts. Add a current sensor or a pulse input only where nothing exists.
How accurate must the counter be?
Close enough to trigger a task before wear matters. A small drift is fine for a service interval of hundreds of hours. A counter that resets without warning is the real risk.
Is this the same as predictive maintenance?
No. This rule fires at a fixed limit. Prediction estimates remaining life from patterns in the data. A usage trigger is the usual stepping stone.
How does the task reach our maintenance software?
The gateway sends the counter value to the maintenance system, which compares it with the limit stored on the asset and opens a work order. Log the asset tag exactly as the system spells it.