Why Manufacturing Plants Struggle with PdM
Manufacturing facilities have a reliability problem that looks different from power generation or oil and gas. The equipment fleet is diverse — CNC machines, hydraulic presses, conveyors, packaging lines, HVAC systems, compressors, boilers, and building utilities — all under one roof. Maintenance budgets are tight. Production pressures demand uptime. And maintenance teams are often stretched thin, reacting to breakdowns rather than preventing them.
The result is predictable: 30-50% of maintenance work orders in a typical manufacturing plant are reactive. Emergency repairs cost 3-5x more than planned work when you account for premium parts pricing, overtime labor, expedited shipping, and lost production. A well-executed predictive maintenance (PdM) program breaks this cycle, but only if it’s designed for the realities of a manufacturing environment.
Step 1: Equipment Criticality Assessment
You cannot monitor everything. Nor should you. The first step is identifying which equipment matters most.
Use a criticality matrix that scores each asset on:
- Production impact — Does failure stop the line? Reduce throughput? Or is there redundancy?
- Safety risk — Could failure injure someone or create a hazardous condition?
- Environmental risk — Could failure cause a release or compliance issue?
- Repair cost and lead time — What does it cost to fix, and how long does it take to get parts?
- Failure frequency — How often has this equipment failed historically?
Score each factor on a 1-5 scale. Multiply production impact by 3 (it’s the dominant driver in manufacturing), safety by 2, and the others by 1. Total scores above 25 are critical. Between 15-25 are important. Below 15 are general.
This exercise typically identifies 15-25% of assets as critical — and those are the ones that get PdM attention first.
Step 2: Match Technologies to Failure Modes
PdM isn’t a single technology. It’s a collection of condition monitoring techniques, each suited to specific failure modes. Selecting the wrong technology for a given failure mode wastes money and generates false confidence.
Vibration Analysis
Best for: rotating equipment — motors, pumps, fans, gearboxes, spindles. Detects imbalance, misalignment, bearing defects, gear mesh problems, looseness, and resonance. Minimum equipment speed for reliable results: 100 RPM (below that, use ultrasonic methods). This is the backbone technology for most manufacturing PdM programs.
Infrared Thermography
Best for: electrical distribution (panels, MCCs, switchgear, bus duct), mechanical equipment (bearings, couplings, conveyor rollers), process equipment (steam traps, insulation, refractory). Fast and non-contact. Covers large areas quickly. Requires trained operators who understand emissivity, load conditions, and environmental influences.
Oil Analysis
Best for: gearboxes, hydraulic systems, compressors, large bearing reservoirs — any equipment with an oil sump. Detects wear metal generation, lubricant degradation, contamination (water, dirt, process fluids). Monthly sampling on critical systems, quarterly on non-critical.
Ultrasonic Testing
Best for: compressed air leak detection, slow-speed bearing monitoring, steam trap assessment, electrical discharge detection. Low cost of entry, fast payback on leak surveys alone.
Motor Testing
Best for: electric motor insulation and rotor condition. Offline testing (megger, surge, hi-pot) during outages. Online testing (motor current analysis) during normal operation. Particularly valuable for critical motors without installed spares.
Step 3: Build Routes and Schedules
Organize monitoring activities into routes — logical groupings of measurement points that a technician covers in a single session. Routes should follow a physical path through the plant to minimize walking time.
Scheduling guidelines for manufacturing:
- Critical rotating equipment: monthly vibration, monthly thermography
- Important rotating equipment: monthly vibration, quarterly thermography
- Electrical distribution: quarterly thermographic survey under load
- Hydraulic systems and gearboxes: monthly oil samples for critical, quarterly for non-critical
- Compressed air system: semi-annual ultrasonic leak survey
- Steam system: annual steam trap survey
Total data collection effort for a mid-size manufacturing plant (200-500 monitored assets) typically requires 40-60 hours per month. That’s roughly one quarter of a full-time position for collection, plus analysis time on top of that.
Step 4: Analysis and Action — Closing the Loop
Data collection without analysis is the single most common PdM program failure. Plants invest in hardware, train technicians to collect data, and then let the data sit in software databases untouched. A monthly vibration route takes 8 hours to collect. Analysis of that route takes 4-8 hours depending on the number of assets and complexity. Budget that analysis time explicitly.
Alarm Management
Set initial alarm thresholds based on ISO standards and equipment manufacturer guidelines. Adjust thresholds after 6-12 months based on actual baseline data. Alert thresholds should trigger investigation. Danger thresholds should trigger work order generation. Track alarm rates — if your vibration program generates fewer than 5% alerts per route, your thresholds might be too high. If it generates more than 20%, they’re too low (or you have a lot of problems).
Work Order Integration
PdM findings must feed directly into the maintenance planning and scheduling process. When analysis identifies a developing fault, a work order should be generated in the CMMS with a recommended repair action, a severity level, and a timeframe for completion. The planning team then schedules the work into the next available maintenance window.
This integration is where many programs break down. PdM technicians identify problems and issue reports, but nobody owns the follow-up. Assign a single person — reliability engineer, maintenance planner, or maintenance supervisor — as the owner of PdM-generated work orders. Track completion rates. If PdM findings aren’t getting addressed, the program will lose credibility with operations and management.
Step 5: Demonstrate Value — Because You’ll Need To
Manufacturing management cares about costs, uptime, and throughput. Your PdM program needs to speak that language. Track and report:
- Avoided failures — Document every instance where PdM identified a developing problem that was corrected before failure. Estimate the cost of the avoided failure (downtime x production rate + repair cost for emergency vs. planned).
- Cost avoidance ratio — Total avoided failure costs divided by total PdM program costs. Healthy programs achieve 5:1 to 10:1 ratios.
- Reactive maintenance percentage — Track emergency work orders as a percentage of total. Starting point is typically 30-50%. Target is less than 10% within 3 years.
- Mean Time Between Failures (MTBF) — For critical assets, track MTBF before and after PdM implementation. Expect 2-3x improvement within the first two years on monitored equipment.
Report quarterly to plant leadership. Monthly reports get ignored. Annual reports lose the connection between PdM activities and results. Quarterly keeps the program visible and maintains support for the resources it requires.
Common Mistakes in Manufacturing PdM Programs
After working with dozens of manufacturing plants, the same mistakes come up again and again.
Starting too big. Trying to monitor 500 assets on day one overwhelms the team and produces poor-quality data. Start with 30-50 critical assets. Do those well for 6 months. Then expand.
Relying on a single technology. Vibration analysis is powerful but doesn’t catch everything. Electrical failures, lubrication problems, and fluid power issues need their own monitoring approaches. A multi-technology program catches a broader range of failure modes.
Underinvesting in training. PdM technologies require skilled analysts. Budget for ISO 18436 certification for vibration analysts, ASNT certification for thermographers, and appropriate training for oil analysis interpretation. A $40,000 vibration analyzer in the hands of an untrained technician produces noise, not information.
No management support. PdM programs that exist because one enthusiastic engineer championed them will die when that person leaves. Build management support by demonstrating value consistently and integrating PdM into the maintenance workflow, not running it as a side project.
A manufacturing PdM program that follows these steps — criticality assessment, technology matching, disciplined data collection, rigorous analysis, and demonstrated value — will transform maintenance from a reactive cost center to a proactive reliability function. The timeline is typically 18-24 months to reach maturity. The results last as long as the discipline holds.