Preventive vs Predictive Maintenance for SME Factories
Published · 7 min read · IoTronix Solution
In short
Preventive maintenance services a machine on a fixed schedule, by calendar or running hours. Predictive maintenance services it when measurements show wear is starting [1]. Most SME factories should do preventive maintenance on every machine. Then add condition monitoring only on the 2 or 3 machines whose stoppage costs the most. This follows the US Department of Energy guide, which says to use predictive maintenance for critical equipment [1].
Preventive vs predictive maintenance: the short answer
Preventive maintenance is work done on a fixed schedule, by time or machine running hours, to slow wear and make parts last longer [1]. The schedule can follow the calendar or the machine's running hours [2]. Predictive maintenance uses measurements that detect the start of wear, so the job is based on the machine's actual condition instead of a preset date [1].
On an injection moulder, preventive means you grease the toggle and change the hydraulic oil filter after a fixed number of running hours, even if it is still fine. Predictive means a sensor checks the hydraulic pump motor. You change the bearing when its vibration or current starts to go up.
Many factories do neither. They only repair the machine after it breaks down. The US Department of Energy's maintenance guide quotes a study where more than 55% of an average facility's maintenance was still reactive, 31% preventive and 12% predictive [1].
How much each one costs to start
Preventive maintenance is cheap to start. You need a list of machines, the maker's service schedule, a running-hour count and one person to sign off each job. The DOE guide estimates a proper preventive programme saves 12% to 18% over running machines until they break [1]. But it has three problems: it needs manpower, it includes work the machine did not need yet, and sudden breakdowns still happen [1].
Predictive maintenance needs sensors, someone to read the data and some training. The DOE guide says clearly that it is not cheap to start. Much of the testing equipment it describes cost more than US$50,000, and staff training also costs a lot [1]. In return it estimates 8% to 12% savings over preventive maintenance alone [1]. Surveys it cites report industrial averages of 25% to 30% lower maintenance cost, 70% to 75% fewer breakdowns and 35% to 45% less downtime [1]. These are survey averages from a US guide. They are not a promise for your factory.
What data predictive maintenance needs
Predictive maintenance only works with signals that change before the machine breaks down. Four signals cover most moulding, press and extrusion factories:
- Vibration. The most common method for rotating machines [4]. A bearing, pump or gearbox shakes more as it wears. Each reading is compared with that machine's own normal level [4].
- Motor current. Motor current signature analysis reads the motor's electrical signal, with no sensor on the moving parts [3][4]. A screw motor drawing more current than last month on the same job is working harder.
- Temperature. Bearings and windings are the two most common causes of motor failure, and a 10 °C rise in operating temperature can halve the life of a motor winding [5]. Motor bodies, hydraulic oil and barrel heater zones all get hotter before a problem.
- Run hours and cycle counts. This does not show condition, but the other signals depend on it. It changes calendar PM into running-hours PM [2]. It also shows how hard each machine works.
You do not need all four on every machine. The usual practice is handheld checks where fixed sensors cost too much, and fixed sensors on critical machines [4].
When preventive maintenance is enough
Preventive maintenance is enough for wear parts, consumables and machines that have a spare. Use predictive maintenance only for critical machines that break down without warning [1]. The DOE guide uses NASA's reliability centred maintenance method. It groups equipment by how it fails and how important it is [1]:
| Approach | Fits | On your floor (examples) |
|---|---|---|
| Run to failure | Small parts, non-critical or redundant equipment, items unlikely to fail [1] | Spare granulator, conveyor fans, a standby chiller |
| Preventive | Equipment that wears, consumables, known failure patterns, maker's recommendations [1] | Filters, hydraulic oil, heater bands, die springs, toggle grease |
| Predictive | Critical equipment and random failure patterns [1] | Hydraulic pump motor on your busiest moulder, press main drive, extruder gearbox |
Some machines have a spare, a short repair time and cheap parts. For these, preventive maintenance, or even running until it breaks, is the right choice. On minor items, condition monitoring can cost more than the equipment it watches [2]. Predictive maintenance is worth it where one stoppage makes you miss a delivery.
Even the best factories still use all three. The DOE guide puts their split at under 10% reactive, 25% to 35% preventive and 45% to 55% predictive [1]. Predictive adds to preventive. It does not replace it.
Comparison: preventive vs predictive maintenance
Preventive maintenance is cheap and suits every machine. Predictive maintenance costs more. It pays back on the 2 or 3 machines whose stoppages cost the most [1].
| Preventive | Predictive | |
|---|---|---|
| What triggers the job | Calendar date or running hours [1] | Measured condition: vibration, current, temperature [1] |
| Cost to start | Low: a machine list, a schedule, a checklist | Sensors, data and training [1] |
| Estimated saving | 12% to 18% over run to failure [1] | 8% to 12% over preventive alone [1] |
| Main weakness | Unneeded work; sudden failures still happen [1] | Upfront spend; savings hard for management to see [1] |
| Best for | Wear parts with known intervals [1] | Critical machines with random failures [1] |
| Data you need | Machine list, maker's intervals, breakdown log | All of that, plus sensor readings and a baseline per machine |
Step by step for a moulding or press factory
Start with a breakdown log and running-hour PM on every machine. Then add condition sensors on the 2 or 3 machines whose stoppages cost the most [1].
- Log every breakdown for a month. Machine, start time, end time, cause, part changed. Not on WhatsApp. This shows which machines stop and why.
- Count running hours per machine. Move PM from the calendar to running hours [2], so a moulder on three shifts gets serviced more often than one on a single shift.
- Rank machines by the cost of a stop. The DOE guide says to define how critical each machine is and what its downtime costs before choosing the mix [1]. The downtime calculator changes hours stopped into RM.
- Add condition monitoring on the top 2 or 3. Vibration and temperature on pump motors and gearboxes, current on the main drive. Set alerts against each machine's own baseline [4].
- Review every month. If an alert came before a breakdown, add monitoring to the next machine. If not, change the alert limits or keep that machine on PM only.
For scale, Siemens estimates the world's 500 biggest companies lose 11% of revenue, about US$1.4 trillion a year, to unplanned downtime [6]. Your own downtime cost decides how many of these steps you need.
Find out which machines need sensors
Book a free 30-minute Factory Paperwork Check. We visit your factory, look at how breakdowns and PM are recorded today, and show you your 3 most time-wasting processes.
Questions
What is the difference between preventive and predictive maintenance?
Preventive maintenance services a machine on a fixed schedule, by calendar or running hours. Predictive maintenance services it when measurements such as vibration, current or temperature show wear is starting [1].
What is preventive maintenance and why is it important?
It is planned work on a fixed schedule, by time or running hours. It slows wear and makes machines last longer [1]. The DOE guide estimates it saves 12% to 18% compared with running machines until they break [1].
Is condition-based maintenance the same as predictive maintenance?
They are very similar. Both decide the job from the machine's measured condition, not a fixed schedule [1]. Predictive maintenance also estimates when the work should be done [3].
Do I need sensors for predictive maintenance on old machines?
You need some measurement, but not a data port. IoTronix clips sensors onto old machines for motor current, temperature and run/stop. Some people call this an IoT setup. For you, it means the machine records the readings, not your staff.
Which machines should I monitor first?
The 2 or 3 machines whose stoppages cost you the most. Rank them by downtime cost, as the DOE guide advises [1]. Start with vibration and temperature on pump motors and gearboxes, and current on the main drive [4].
What is CMMS software and do I need one?
A computerised maintenance management system (CMMS) keeps the machine list and creates PM work orders by calendar or running hours [1]. With a few machines, a spreadsheet is enough. When several technicians and shifts share the work, an app makes sure breakdown and PM records are not lost.
Sources
- [1]U.S. Department of Energy, Federal Energy Management Program (prepared by PNNL). Operations & Maintenance Best Practices: A Guide to Achieving Operational Efficiency, Release 3.0, 2010-08
- [2]Wikipedia. Preventive maintenance
- [3]Wikipedia. Predictive maintenance
- [4]Wikipedia. Condition monitoring
- [5]ABB Review. Motor efficiency: Focus on optimizing lifetime performance on motors, 2007
- [6]Siemens AG. The True Cost of Downtime 2024, 2024
