Unplanned downtime in process industries typically costs between $10,000 and $250,000 per hour, according to industry estimates. A single compressor failure in a petrochemical facility triggers a cascade. Typical ranges include 4-6 hours for diagnosis, 24-72 hours for parts, 8-16 hours for repair, plus 2-4 hours for a safe restart. At roughly $75,000-150,000/hour production loss for a mid-sized process facility, operations can face $3-8 million in losses before equipment runs again. Yet maintenance managers face a paradox: spend too much on scheduled maintenance that may be an estimated 30-40% unnecessary, or spend more on emergency repairs when equipment fails between intervals.
This article provides a practical decision-making framework that goes beyond definitions. If you’re a maintenance manager or reliability engineer evaluating your facility’s approach, you’ll find specific guidance on when preventive maintenance remains the right choice, when predictive maintenance delivers superior ROI, and how to strategically combine both. We’re covering real examples from petrochemical and refining environments, technical depth on monitoring techniques, and implementation guidance that accounts for the 18-36 month reality of transitioning between strategies.
Costs, timelines, and technology specifications referenced in this article reflect general North American industry conditions. Dollar figures represent typical ranges in USD. Verify current pricing with vendors and consult qualified professionals for facility-specific recommendations, as individual results vary significantly based on asset profile, implementation quality, and regional factors. Facilities in Canada should verify alignment with applicable provincial regulations, including Alberta Energy Regulator requirements for oil and gas operations.
Here’s the context that matters: with industrial condition monitoring sensors (vibration, temperature) now ranging from $100-500 per monitoring point. Down from $ 500 to $1,000+ a decade ago, the question isn’t whether predictive maintenance works. Decades of data prove it does. The real question is whether your facility has the data infrastructure, asset profile, and organisational readiness to capture its value.
What Is Preventive Maintenance?
Preventive maintenance is a time-based or usage-based maintenance strategy that performs scheduled interventions, including inspections, part replacements, and lubrication, at predetermined intervals regardless of equipment condition. Think of it as the annual physical for your equipment: you show up at the scheduled time, whether you feel sick or not.
The strategy comes in three flavours. Calendar-based maintenance happens on fixed schedules, such as pump seal inspections every 90 days or heat exchanger cleaning every 12 months. Usage-based maintenance triggers work orders based on meter readings: overhaul the compressor every 8,000 operating hours, replace bearings after 50,000 cycles. Condition-based triggers schedule maintenance when specific wear thresholds are reached, though they still follow predetermined parameters rather than real-time analysis.
How Long Has Preventive Maintenance Been the Industry Standard?
Preventive maintenance has been standard since the 1950s, delivering 12-18% cost savings compared to reactive maintenance, according to the U.S. Department of Energy’s O&M Best Practices Guide. These benchmarks align with findings from Natural Resources Canada and apply broadly to North American industrial operations. The approach uses historical data, OEM recommendations, and mean time between failures (MTBF), which measures the average operating time between breakdowns, to establish intervals that typically catch 70-85% of problems before they become catastrophic.
A computerised maintenance management system (CMMS) provides software for scheduling, tracking, and documenting maintenance activities. Platforms like IBM Maximo, Fiix, or UpKeep transform manual tracking processes into streamlined workflows. Pricing varies by vendor and changes frequently, so contact providers directly for current quotes.
An honest perspective: preventive maintenance is often criticised as “wasteful” by predictive maintenance evangelists, but that criticism often comes from vendors selling $50,000+ monitoring systems. For approximately 60-70% of industrial assets, including utility pumps under $5,000, HVAC systems, and standard filtration equipment, scheduled maintenance remains more cost-effective than sophisticated monitoring.
What Is Predictive Maintenance?
Predictive maintenance is a condition-based strategy that uses real-time sensor data to detect equipment anomalies and predict failures before they occur. Instead of changing oil every 3,000 hours because that’s the schedule, change it when analysis shows contamination has actually exceeded acceptable thresholds.
The foundation is condition-based maintenance (CBM), which monitors actual equipment state through sensor data rather than relying on calendar intervals. CBM enables predictive programs because you can’t predict failures without monitoring conditions. We’ll cover how these work together in the hybrid approach section below.
Core Predictive Maintenance Techniques
Vibration analysis measures oscillatory patterns in rotating equipment, such as pumps, compressors, and turbines, to detect imbalances, misalignments, or bearing degradation. Technicians monitor frequency signatures: 1x RPM indicates imbalance, 2x RPM suggests misalignment, and bearing defect frequencies reveal component wear. When vibration velocity exceeds thresholds, typically 0.16-0.25 in/sec peak (Zone B/C boundary per ISO 10816-3 for Group 2 machinery), triggering investigation, and values above 0.25 in/sec peak (Zone C/D) requiring immediate attention, the system generates work orders. Note: Thresholds vary by machine class, power rating, and foundation type; consult ISO 10816-3 for specific equipment classifications.
Infrared thermography uses thermal imaging to identify abnormal heat signatures in electrical systems and mechanical equipment. Temperature differentials follow NETA MTS severity classifications: when compared with similar components, 4-15°C indicates a probable deficiency requiring scheduled repair, while differentials exceeding 30°C require immediate action.
For electrical systems specifically, power quality monitoring provides an additional layer of early fault detection by tracking voltage fluctuations, harmonics, and power factor changes.
Note: Different thresholds apply when using ambient temperature as a reference; consult NETA MTS Table 100.18 for complete severity criteria.
Oil analysis examines lubricant samples for contamination, wear particles, and chemical degradation, providing 4-12 weeks’ advance warning of internal component wear. Key indicators include ISO cleanliness code changes, wear metal concentrations exceeding baseline limits (varies by equipment type. For example, some gearboxes alarm at 70-100 ppm iron, while others may tolerate higher levels; trending is typically more important than absolute values), and viscosity shifts greater than 10% from baseline.
Ultrasonic analysis detects high-frequency sounds above 20 kHz associated with leaks, electrical discharge, and early-stage bearing defects. This technique is particularly valuable for slow-speed equipment (typically 120-600 RPM, depending on the application) where vibration analysis may be less effective.
Digital Infrastructure Requirements
This is where vendors get uncomfortable: predictive maintenance requires significant digital infrastructure that takes 6-18 months to implement properly. Facilities need sensors at monitoring points, data historians such as OSIsoft PI or open-source alternatives like InfluxDB, analytics platforms, and integration with CMMS for work order generation. Software and platform costs vary widely and change frequently, so request current quotes from vendors based on your specific asset count and requirements.
A note on realistic timelines: If someone tells you predictive maintenance is “plug and play” or “up and running in 30 days,” they’re either selling something or haven’t implemented it in a real facility. Sensor installation takes 2-4 weeks. Network configuration takes another 2-4 weeks. System integration requires 4-8 weeks. Baselining equipment takes 3-6 months. Training teams to interpret alerts takes 3-6 months. Budget 12-18 months from kickoff to reliable operation.
Key Differences Between Preventive and Predictive Maintenance
The core distinction comes down to what triggers maintenance action. Preventive follows fixed intervals: time passes, or usage accumulates; a work order is generated; the technician executes. Predictive responds to equipment condition: sensors detect anomalies, analytics confirm trends, and a work order is generated for the specific problem.
Factor | Preventive Maintenance | Predictive Maintenance Maintenance Trigger | Time/usage intervals | Real-time equipment condition Data Source | Historical MTBF data | Continuous sensor monitoring Implementation Cost | $5,000-50,000 | $75,000-500,000+ Annual Operating Cost | $50-150/asset | $100-300/asset Cost Savings vs. Reactive | 12-18% | Industry benchmarks suggest 25-35% Best Application | Stable failure patterns | Variable failure modes Infrastructure Required | CMMS | CMMS + IIoT + Analytics Implementation Timeline | 2-6 months | 12-24 months
Note: Cost figures represent typical North American ranges in USD and vary based on facility size, asset complexity, and vendor selection. Verify current pricing before budgeting.
Resource requirements differ significantly. Preventive programs need technicians who follow checklists, and most facilities already have this capability. Predictive programs need those technicians plus engineers who can interpret condition data and distinguish genuine signals from sensor noise. That expertise takes 6-12 months to develop internally or costs $150-250/hour for third-party analysts.
What rarely gets discussed: predictive maintenance creates different organisational demands. Instead of “do this task every Tuesday,” teams respond dynamically to unpredictable condition alerts. That flexibility requires cultural change, moving from “schedule compliance” to “condition response” metrics, which an estimated 60-70% of facilities underestimate.
When Preventive Maintenance Remains the Better Choice
Not everything needs condition monitoring. Predictive maintenance purists hate hearing this, but it’s true: for approximately 50-70% of industrial assets, scheduled maintenance makes more economic sense.
Stable, predictable failure patterns favour preventive approaches because monitoring adds cost without adding information. Components with consistent wear curves, such as air filters that need replacement every 2,000-4,000 hours, V-belts lasting 12-18 months, and mechanical seals lasting 24-36 months, don’t benefit from continuous monitoring. You know the filter clogs after roughly 3,000 hours. Vibration monitoring won’t tell you anything new.
Why stable failure patterns favour time-based maintenance: these follow predictable degradation curves where physics don’t change. A paper filter clogs as particulate accumulates, and no sensor predicts this better than operating hour counts. Result: scheduled replacement captures 90% or more of problems.
Low-criticality assets fall into the same category. That 3-HP utility pump serving a non-critical cooling loop? If it fails, you switch to backup in 10 minutes. Installing $800 worth of sensors to monitor a $1,500 pump that causes zero production impact isn’t optimisation. It’s a waste.
When Should Facilities Choose Preventive Over Predictive Maintenance?
Choose preventive maintenance when assets have stable failure patterns, low criticality with failure costs under $25,000, limited data infrastructure, or regulatory-mandated inspection intervals. These conditions describe approximately 50-70% of industrial assets.
Limited data infrastructure presents practical constraints. Without IIoT capability, historian systems, or analytics platforms, implementing predictive maintenance requires building that foundation first. That means significant investment and 12-24 months before monitoring a single asset.
A reality check: there’s no shame in running a preventive-heavy program. Scheduled maintenance has delivered documented cost savings for more than 70 years. The question isn’t “preventive or predictive” but “where does each make sense?”
When Predictive Maintenance Delivers Superior ROI
Predictive maintenance shines when failures are expensive, exceeding $25,000 per incident, unpredictable with intervals varying more than 50% from the mean, and detectable through monitoring with signatures appearing 2+ weeks before failure. That describes roughly 15-25% of a typical facility’s equipment.
Critical rotating equipment, including compressors, turbines, and large pumps over 100 HP, represents the classic use case. A centrifugal compressor costing $1.5-3 million causes substantial production losses during unplanned outages, often $75,000-150,000 per hour, depending on facility output. Vibration analysis catches bearing degradation 4-12 weeks before failure, providing enough time to order parts and schedule repairs during planned downtime.
Why data-driven maintenance delivers for rotating equipment: failures follow a progressive degradation pattern, producing measurable vibration signatures. Bearing wear increases friction, friction creates vibration at specific frequencies, and that vibration grows predictably over weeks. Result: industry experience suggests 70-85% of rotating equipment failures are detectable 30+ days in advance.
High-consequence failures justify monitoring even on less expensive equipment. An $8,000 control valve might not seem worth monitoring until valve failure triggers an emergency shutdown, potentially causing $200,000 or more in lost production.
Variable failure patterns render scheduled intervals entirely ineffective. Some equipment fails unpredictably due to stress corrosion, intermittent electrical faults, or process-induced degradation, with failure rates varying by 200-300% depending on feedstock quality. Condition monitoring addresses this by scheduling maintenance when data indicates actual need.
Calculating Predictive Maintenance ROI
Annual Monitoring Cost: Sensors (typically $100-300/year amortised) + Platform ($100-250/asset/year) + Analyst time ($200-400/year) = approximately $400-950/asset/year
Avoided Unplanned Downtime: (Failure rate) × (Repair hours) × (Downtime cost) × (Detection rate)
Consider this example scenario: monitoring a critical compressor costs approximately $8,500 annually. If historical data shows one failure every 3 years, averaging 72 hours at $75,000/hour, that represents roughly $5.4 million per failure, or $1.8 million annualised. Assuming an 80% detection rate for bearing-related failures, the potential avoided cost is $1.44 million annually, against a $8,500 monitoring investment. In this scenario, ROI approaches 169x with payback under 3 days.
For a comprehensive breakdown of how to build a financial case for predictive maintenance investments, see our complete guide to predictive maintenance cost savings.
Individual facility results vary significantly based on asset criticality, failure history, detection accuracy, and implementation quality. Not every asset pencils out this clearly. The discipline is doing the math honestly rather than assuming monitoring always pays.
The Hybrid Approach: Combining Preventive and Predictive Strategies
Every vendor presentation glosses over this: real facilities don’t choose between preventive and predictive. They use both, strategically allocated across asset classes. Industry experience indicates that most mature facilities use hybrid approaches combining preventive and predictive strategies based on asset criticality and failure characteristics.
Reliability-centred maintenance (RCM) is a systematic framework for determining the most effective strategy for each asset based on function, failure modes, and consequences. RCM analysis, typically requiring 8-16 hours per asset class with a qualified facilitator, asks: how can this equipment fail, what happens when it fails, and what strategy addresses those failure modes most effectively?
In Alberta and other Canadian jurisdictions, maintenance strategies and reliability analyses for regulated facilities may require review by a licensed professional engineer. Verify requirements with APEGA or your provincial engineering association.
Asset Criticality Classification
Criticality A (10-15% of assets): Production-critical with greater than $100,000 failure consequences. Strategy: Continuous predictive monitoring. Examples: main compressors, critical pumps over 200 HP.
Criticality B (20-25% of assets): $25,000-100,000 failure consequences. Strategy: Periodic condition assessments, including monthly vibration routes and quarterly thermography. Examples: secondary process equipment, large motors.
Criticality C (40-50% of assets): Less than $25,000 consequences, predictable wear. Strategy: Preventive maintenance on fixed schedules. Examples: auxiliary pumps, standard HVAC.
Criticality D (15-25% of assets): Minimal impact, low repair cost. Strategy: Run-to-failure. Examples: redundant utility equipment, items under $2,000.
A petrochemical facility with 1,000 assets might have 120 in Category A, 230 in Category B, 450 in Category C, and 200 in Category D. This distribution is typical. However, it varies by facility type and industry. Recommendations to monitor all assets typically overlook economic realities. Effective programs match monitoring investment to the consequences of failure.
Transitioning from Preventive to Predictive
Facilities attempting wholesale transformation usually fail. Industry experience suggests that phased implementations typically outperform wholesale transformation approaches, which often struggle to achieve projected ROI within the first few years. A phased approach works better.
Phase 1 (Months 1-6): Install monitoring on 5-10 pilot assets with known problems. Build analyst capability. Keep PM programs running in parallel. Budget: typically $50,000-150,000.
Phase 2 (Months 6-18): Expand to remaining Criticality A assets. Reduce PM frequency where 6+ months of condition data support longer intervals. Budget: typically $100,000-$300,000.
Phase 3 (Months 18-36): Extend assessments to Criticality B assets. Integrate predictive triggers into CMMS workflows. Budget: typically $150,000-$400,000.
Why Does the Parallel Running Period Matter?
Predictive maintenance detects different failures than preventive maintenance. PM catches wear-out failures such as seals and bearings with predictable degradation. PdM catches random failures, including electrical faults and contamination that don’t follow schedules. Until you validate that condition monitoring catches the problems PM was preventing, don’t eliminate those tasks.
How Much Does It Cost to Implement Predictive Maintenance?
The following ranges represent typical North American industry pricing in USD. Software and hardware costs change frequently, so contact vendors directly for current quotes based on your specific requirements.
Sensors: Typically $50-$500 per monitoring point. Basic accelerometers generally run $180-400. Wireless transmitters run $400-700. Typical rotating equipment setup costs $800-$1,500 installed.
Platform/Software: Entry-level options range from $100 to $ 150/asset/month. Mid-market solutions typically cost $50,000-80,000/year for 50-100 assets. Enterprise platforms often cost $150,000-$400,000/year for 200+ assets.
Integration: $15,000-$75,000, depending on system complexity.
Training: Vibration certification typically costs $2,500-$3,500 per person. Thermography Level I certification costs $1,500-$2,500. Certification and licensure requirements vary by jurisdiction. In Canada, verify current requirements with your provincial governing body. In Alberta, certain diagnostic and inspection activities at regulated facilities may require oversight by a professional engineer registered with APEGA. Budget 6-12 months for full team adoption.
Typical Pilot Program: 10-20 assets, approximately $75,000-200,000 initial investment, with 12-24 month payback on assets with greater than $500,000 annual failure risk.
A word on vendor transparency: vendors should provide ballpark pricing after an initial conversation. If a vendor requires multiple sales calls before discussing costs, consider whether that approach aligns with your procurement process. Transparent vendors typically quote within 20% of your basic requirements after understanding them.
Can Preventive and Predictive Maintenance Be Combined?
Yes. Most industrial facilities combine both strategies based on asset criticality. Predictive maintenance monitors high-value, critical assets representing 15-25% of equipment, while preventive maintenance handles standardised assets, approximately 50-70%, with predictable wear patterns. This hybrid approach optimises resources without over-investing in monitoring technology.
Integration with your CMMS manages both scheduled preventive work orders and condition-triggered predictive work orders. When vibration analysis detects bearing degradation, the analytics platform creates a work order scheduled for the next maintenance window, living alongside calendar-based PMs in the same interface.
What Are the Best Predictive Maintenance Techniques for Industrial Equipment?
Vibration analysis is best suited for rotating equipment, including pumps, compressors, and motors. This technique detects imbalance, misalignment, and bearing wear. Use monthly routes for Category B assets and continuous monitoring for Category A.
Infrared thermography works best for electrical systems, insulation failures, and mechanical friction. Schedule quarterly scans for electrical systems and monthly scans for critical equipment.
Oil analysis is ideal for gearboxes, hydraulic systems, and lubricated bearings. Typical cost runs $25-50 per sample. Schedule quarterly for most equipment.
Ultrasonic testing excels at leak detection, slow-speed bearings, and steam traps. Schedule monthly for steam traps and quarterly for bearings.
Motor current analysis is best for electric motor health, including rotor bars and stator windings. Schedule annually for most motors and quarterly for critical units.
The right technique depends on the failure mode. Vibration won’t help with electrical problems; use thermography instead. Thermography won’t catch internal bearing wear, so use vibration. Most comprehensive programs use 3-4 techniques matched to specific failure modes.
Bottom Line
The preventive-versus-predictive debate misses the point. Both work. Preventive delivers 12-18% savings according to DOE benchmarks, and industry experience suggests predictive delivers 25-35%. The question is: where does each make sense? Scheduled maintenance remains right for approximately 50-70% of assets: standardised equipment with predictable failures and costs under $25,000. Condition-based maintenance delivers for 15-25%: critical equipment where real-time data prevents high-consequence failures. Programs achieving the strongest results combine both through systematic criticality classification.
Start by classifying assets by criticality, a process that typically takes 2-4 weeks with a cross-functional team. Identify 5-10 high-value candidates for predictive pilots: equipment with greater than $200,000 replacement cost, greater than $50,000 failure consequences, and clear condition-to-failure relationships. Assess infrastructure honestly: do you have sensor capability, integration pathways, and analytics platforms? If not, budget for a significant investment and expect value in 12-18 months.
The figures and timelines in this framework represent industry benchmarks and typical scenarios. Your facility’s results will depend on asset profile, existing infrastructure, organisational readiness, and current market conditions. Certification and licensure requirements vary by jurisdiction. A professional assessment is recommended before major changes to maintenance strategies. Vista Projects combines four decades of industrial engineering experience with digital transformation expertise to help facilities optimise maintenance strategies. Whether evaluating predictive maintenance feasibility, implementing condition monitoring, or integrating data across asset information management platforms, our approach addresses both technical implementation and organisational change management.