The four maintenance strategies
Most plants use a mix of four strategies, selected based on asset criticality and failure characteristics:
| Strategy | Trigger | Also called | Best for |
|---|---|---|---|
| Reactive (RTF) | Failure occurs | Run-to-fail, breakdown maintenance | Non-critical, low-cost, redundant assets |
| Preventive (PM) | Time or usage interval | Time-based maintenance, scheduled maintenance | Assets with known wear-out patterns |
| Predictive (PdM) | Condition monitoring data | Condition-based maintenance (CBM) | Critical rotating equipment with detectable failure signatures |
| Proactive | Root cause elimination | Precision maintenance, reliability engineering | Chronic failure elimination, new installations |
Preventive maintenance explained
PM intervals are based on manufacturer recommendations, operating experience, or regulatory requirements. Common PM tasks:
- Oil and filter changes at fixed hours (e.g., 2,000-hour oil change on a rotary screw compressor)
- Belt replacement at fixed intervals (e.g., every 12 months)
- Bearing replacement after a set number of operating hours
- Electrical insulation testing annually
The problem with PM: Intervals are set conservatively to cover the worst-case operating conditions. In many cases, components are replaced with significant remaining useful life, wasting money and introducing infant-mortality failures (new components fail disproportionately in the first hours after installation). Studies show that 33–66% of PM tasks either cause no benefit or actively create failures (Nowlan & Heap, MSG-3 analysis).
Predictive maintenance explained
PdM uses sensor data to monitor the actual condition of an asset and generate a maintenance alert only when the condition degrades beyond a threshold. The asset is serviced only when necessary.
Common PdM techniques and the failure modes they detect:
| Technique | Detects | P-F interval (typical) |
|---|---|---|
| Vibration analysis | Bearing defects, imbalance, misalignment, looseness | 2 weeks – 6 months |
| Oil analysis | Wear, contamination, degradation | 1 – 3 months |
| Infrared thermography | Electrical hot spots, insulation failure, bearing heat | 1 day – 3 months |
| Ultrasound | Bearing defects (early), compressed air / steam leaks | 2 weeks – 4 months |
| Motor current analysis | Rotor bar faults, eccentricity, mechanical load | 1 week – 3 months |
Side-by-side comparison
| Factor | Preventive (PM) | Predictive (PdM) |
|---|---|---|
| Trigger | Time / usage interval | Condition threshold crossed |
| Upfront cost | Low | Medium–High (sensors, training) |
| Ongoing labour | Medium (scheduled tasks) | Low (route monitoring + analysis) |
| Parts cost | High (components replaced regardless) | Lower (intervention only when needed) |
| Unplanned downtime | Moderate (some failures between PMs) | Low (failures detected early) |
| Lead time for planning | Low (scheduled in advance) | Higher (need P-F interval awareness) |
| False alarm rate | Zero (no monitoring) | Low–Medium (analyst skill dependent) |
| Failure modes addressed | Age-related wear-out (Weibull bathtub right tail) | Random failure + gradual degradation |
| ROI (well-implemented) | $2–4 per $1 spent | $4–8 per $1 spent |
Decision framework: which strategy to use
Use this logic tree to select the right strategy for each asset:
Step 1: Is the asset critical? (Safety risk, single point of failure, high downtime cost > £5,000/hr?) → If NO: consider reactive or basic PM. If YES: continue.
Step 2: Does the failure have a detectable P-F interval with available technology? → If NO: use PM or redesign. If YES: PdM is appropriate.
Step 3: Is the P-F interval long enough to plan and schedule the repair? (At least 2× the maintenance lead time?) → If YES: implement PdM. If NO: accept the risk or improve detection.
Asset criticality considerations
Always classify assets before selecting a strategy. A simple criticality matrix scores each asset on:
- Safety consequence of failure (1–5)
- Production impact (1–5)
- Maintenance cost of failure (1–5)
- Repair time / lead time for parts (1–5)
Assets scoring 15+ are typically candidates for PdM. Below 8, reactive or basic PM suffices. See our equipment criticality analysis guide.
Cost comparison
The total cost of each strategy includes:
- Prevention costs — labour, parts, sensor equipment
- Internal failure costs — unplanned downtime, secondary damage
- External failure costs — customer impact, safety incidents
Source: Plant Engineering Magazine, 2023 Maintenance Study. Figures include labour, parts, and downtime cost allocation.
Implementing PdM — where to start
- Identify your top 20 critical assets — use criticality analysis to prioritise. These are your first PdM candidates.
- Select the right techniques — vibration for rotating equipment, thermography for electrical, oil analysis for gearboxes and engines.
- Establish baselines — you need at least 3–6 months of "healthy" data to set meaningful alarm thresholds.
- Train your team or contract a specialist — vibration analysis to ISO 18436-2 Category II requires significant training. Consider outsourcing until in-house capability develops.
- Integrate with your CMMS — PdM alerts must generate work orders automatically to close the loop.
FAQs
Rarely. Some tasks cannot be deferred based on condition alone — lubrication intervals, filter changes, safety device testing, and regulatory-mandated inspections all remain on time-based schedules. PdM replaces most component-replacement tasks (bearings, belts, seals) but co-exists with a reduced PM schedule. A mature programme might be 60–70% PdM, 30–40% PM, and near-zero reactive for critical assets.
The P-F interval (Potential failure to Functional failure) is the time between when a failure becomes detectable (P) and when it causes loss of function (F). It is the window you have to act once a PdM alert is raised. If the P-F interval for a bearing defect detected by vibration is 6 weeks, you have 6 weeks to plan and execute the repair — but you must act before the interval expires. Your monitoring frequency must be less than half the P-F interval to reliably catch failures.
The terms are often used interchangeably. Strictly, condition-based maintenance (CBM) triggers an action when a measured parameter crosses a threshold. Predictive maintenance uses trend analysis and modelling to predict when failure will occur — going beyond simple threshold alerting. In practice, most industrial PdM programmes are CBM with trending, rather than true predictive modelling.