Who this is for: Industry solution architects evaluating which use cases to pursue, reliability engineers looking for patterns that match their assets, and business leaders who need concrete examples to justify the investment. Find your industry, find your asset, and find your starting point.
The Pattern Finder's Guide
Every organization asks the same question: "What should we predict first?"
The answer is almost never "the most expensive thing." It is "the thing you have the best data for that also hurts when it fails."
We have implemented Maximo Predict across four major industry verticals. In every single one, certain patterns recur. This blog catalogs those patterns so you do not have to discover them from scratch.
Here is a number that matters: Organizations that select their first use case based on data readiness rather than perceived business urgency succeed 3x more often. The most critical asset is not always the best first prediction target.
Manufacturing Use Cases
Manufacturing environments are target-rich for predictive maintenance. Diverse assets, harsh operating conditions, and high cost of downtime.
Rotating Equipment: The Gateway Use Case
Assets: Pumps, fans, blowers, compressors, motors
This is where most organizations start, and for good reason. Rotating equipment fails in predictable, degradation-based patterns. Vibration data is widely available. Failure examples are plentiful.
Named Failure: The Weekend Bearing
A food manufacturer's plant manager told us: "Every bearing failure happens on a Friday night. Every single one." His maintenance records confirmed it -- not because bearings prefer weekends, but because Friday afternoon is when weekend production ramps up, pushing already-degraded bearings past their limit. Predict could see the degradation trend days before the weekend killed the bearing.
Use Case: Pump Bearing Failure Prediction
| Element | Detail |
|---|---|
| Model type | Failure probability (30-day window) |
| Features | Vibration trend, runtime since replacement, days since PM, operating temperature |
| Data sources | Monitor (vibration, temp), Manage (WOs, meters) |
| Threshold | >65% triggers inspection |
| Result | 40% reduction in unplanned pump failures |
Use Case: Motor Winding Degradation
| Element | Detail |
|---|---|
| Model type | RUL estimate |
| Features | Current draw trending, winding temperature, runtime hours, insulation resistance |
| Data sources | Monitor (electrical params), Manage (WOs, inspections) |
| Planning | Schedule replacement during next planned outage |
| Result | Zero unplanned motor failures in first year |
Production Line Equipment
Assets: Conveyors, packaging machines, assembly robots, stamping presses
Production lines fail in patterns tied to cycle counts, wear rates, and environmental conditions.
Named Failure: The Cascading Line Stop
When one station on a packaging line jams, every downstream station stops. A 15-minute fix becomes a 2-hour restart. A pharmaceutical client found that 60% of their unplanned line stops were preceded by detectable cycle time variance 4 to 8 hours earlier.
Use Case: Packaging Machine Jam Prediction
| Element | Detail |
|---|---|
| Model type | Failure probability (8-hour window) |
| Features | Cycle time variance, reject rate trend, film tension, ambient temperature |
| Data sources | PLC data via Monitor, quality system, Manage WOs |
| Action | High probability triggers operator inspection between shifts |
| Result | 25% reduction in unplanned line stops |
HVAC and Facilities
Assets: Chillers, boilers, air handling units, cooling towers
Critical for data centers, hospitals, clean rooms, and any environment where temperature control is non-negotiable.
Use Case: Chiller Compressor RUL
| Element | Detail |
|---|---|
| Model type | RUL estimate |
| Features | Discharge temp, oil pressure, power consumption, runtime, refrigerant levels |
| Data sources | BMS via Monitor, Manage WOs and inspections |
| Planning | Schedule replacements during maintenance windows |
| Result | Zero unplanned chiller outages in critical data center facilities |
Utilities and Energy Use Cases
Long-lived assets, geographically distributed, high reliability requirements, serious consequences for failure.
Transformers
The flagship utility use case. High-value assets with catastrophic failure consequences.
Named Failure: The Silent Degradation
A utility lost a $1.2M power transformer that had passed its routine inspection 6 weeks earlier. Post-mortem revealed insulation degradation that dissolved gas analysis would have caught -- but DGA results were in a spreadsheet that nobody had connected to the asset management system. The data existed. The integration did not.
Use Case: Distribution Transformer Failure Probability
| Element | Detail |
|---|---|
| Model type | Failure probability (12-month window) |
| Features | Age, loading history, DGA results, location characteristics, weather exposure |
| Data sources | Manage (assets, inspections), lab system (DGA), weather API |
| Application | Prioritize replacements in capital budget |
| Result | 30% better targeting of replacement investments |
Switchgear and Breakers
Use Case: Breaker Mechanism Failure
| Element | Detail |
|---|---|
| Model type | Failure probability based on operations and condition |
| Features | Operation count, timing test results, age, manufacturer, contact resistance |
| Data sources | Manage (WOs, inspections), relay data, test equipment |
| Action | Schedule overhaul when probability exceeds threshold |
| Result | 35% reduction in breaker-related outages |
Renewable Energy Assets
Assets: Wind turbines, solar inverters, battery storage systems
Young asset classes with growing data histories and unique challenges.
Use Case: Wind Turbine Gearbox Prediction
| Element | Detail |
|---|---|
| Model type | RUL for gearbox bearings |
| Features | Vibration spectra, oil analysis, power output anomalies, wind speed patterns |
| Data sources | SCADA, condition monitoring system, oil lab |
| Planning | Coordinate replacements with crane availability and low-wind periods |
| Result | Reduced gearbox replacement costs through optimal scheduling |
The unique challenge with renewables: Assets are often in remote locations. The cost of a technician visit is high even when the repair is simple. Prediction accuracy directly translates to reduced truck rolls.
Transportation Use Cases
Mobile assets, varied operating conditions, strict safety requirements, large homogeneous fleets.
Fleet Vehicles
Assets: Trucks, buses, service vehicles
Large fleets of similar vehicles are ideal for predictive models. Hundreds or thousands of similar assets provide abundant training data.
Named Failure: The Roadside Breakdown
A logistics company's average roadside breakdown cost $8,500 in towing, lost load time, and emergency repair premiums. Their fleet of 1,000 trucks averaged 47 roadside breakdowns per month. After deploying Predict on engine failure patterns, that number dropped to 23. That is $200K per month in avoided breakdown costs.
Use Case: Engine Failure Prediction for Truck Fleet
| Element | Detail |
|---|---|
| Model type | Failure probability (30-day window) |
| Features | Engine hours, fault codes, fuel efficiency trend, mileage since overhaul |
| Data sources | Telematics (via Monitor), diagnostic systems, Manage WOs |
| Action | Route high-risk vehicles to depot for inspection |
| Result | 50% reduction in roadside breakdowns |
Rail Assets
Assets: Locomotives, railcars, track components, signaling equipment
Safety-critical with regulatory oversight. Even modest prediction accuracy justifies the investment.
Use Case: Railcar Bearing Prediction
| Element | Detail |
|---|---|
| Model type | Failure probability based on wayside detector data |
| Features | Temperature trending, acoustic signatures, mileage, load patterns |
| Data sources | Wayside detectors, on-board diagnostics, Manage WOs |
| Action | Set out cars for bearing replacement before failure |
| Result | Significant reduction in bearing-related derailment risk |
Aviation Ground Support
Assets: Baggage tugs, belt loaders, pushback tractors
Use Case: Ground Support Equipment Reliability
| Element | Detail |
|---|---|
| Model type | Failure probability by equipment type |
| Features | Usage cycles (flights handled), age, fault codes, days since PM |
| Data sources | Flight ops systems, telematics, Manage WOs |
| Planning | Ensure spare equipment during peak travel periods |
| Result | Improved on-time departure performance |
Oil & Gas and Chemicals Use Cases
High-value assets in hazardous environments. Safety and environmental compliance add urgency to every prediction.
Process Pumps and Compressors
Named Failure: The Turnaround Surprise
A gas processing plant planned a 21-day turnaround. During pre-turnaround inspection, they discovered a compressor needed major repair they had not planned for. The turnaround extended to 31 days, costing $1.4M in additional lost production. Predict models would have flagged the compressor 4 months earlier, giving the turnaround team time to plan.
Use Case: Compressor Multi-Failure-Mode Prediction
| Element | Detail |
|---|---|
| Model type | Failure probability for multiple failure modes |
| Features | Vibration signatures, discharge temperature, suction pressure, seal system data |
| Data sources | Monitor (process data, vibration), Manage WOs |
| Action | Feed predictions into turnaround planning |
| Result | Extended run time between turnarounds by 15% |
Heat Exchangers
Use Case: Exchanger Fouling Prediction
| Element | Detail |
|---|---|
| Model type | RUL (days until cleaning required) |
| Features | Temperature approach trending, flow rates, days since cleaning, crude blend |
| Data sources | Process historian, Manage cleaning records |
| Planning | Schedule cleaning during maintenance windows |
| Result | Optimized cleaning intervals, improved energy efficiency |
Pipelines
Use Case: Pipeline Corrosion Growth Prediction
| Element | Detail |
|---|---|
| Model type | Predicted metal loss over time |
| Features | In-line inspection measurements, soil conditions, cathodic protection readings, operating pressure |
| Data sources | ILI vendor data, CP monitoring, SCADA, Manage integrity records |
| Planning | Prioritize digs and repairs by predicted risk |
| Result | Better allocation of integrity management budget |
The Five Reusable Solution Patterns
Despite industry differences, these patterns recur everywhere.
Pattern 1: Failure Probability for Work Triggering
The pattern: Calculate failure probability. Trigger work when threshold exceeded.
Score asset daily ──> Compare to threshold ──> Create WO if exceededApplicable to: Any asset type with sufficient failure history.
Key success factor: Well-calibrated thresholds that balance false alarms and missed failures.
Pattern 2: RUL for Capital Planning
The pattern: Estimate remaining life. Feed into capital planning and budgeting.
Estimate RUL ──> Sort by projected failure year ──> Allocate budgetApplicable to: Long-lived, high-value assets (transformers, major equipment).
Key success factor: Multi-year prediction horizon with appropriate uncertainty bounds.
Pattern 3: Condition-Based PM Optimization
The pattern: Adjust PM intervals based on predicted risk.
Score risk ──> High risk: accelerate PM | Low risk: defer PMApplicable to: Assets with established PM programs and varying risk profiles.
Key success factor: Approval workflow for PM changes and outcome tracking.
Pattern 4: Fleet-Level Risk Ranking
The pattern: Score entire population. Rank by risk for resource allocation.
Score all assets ──> Rank by probability ──> Focus on top NApplicable to: Large fleets of similar assets.
Key success factor: Portfolio visualization in Health and clear resource allocation rules.
Pattern 5: Multi-Model Ensemble
The pattern: Combine predictions from multiple models for comprehensive risk.
Model A (bearing) ──┐
Model B (seal) ──┼──> Combined risk score ──> Prioritized action
Model C (motor) ──┘Applicable to: Complex assets with multiple failure modes.
Key success factor: Appropriate weighting of individual model scores.
Adapting Use Cases to Your Context
You have seen the examples. Now adapt them to your reality.
Step 1: Assess What You Actually Have
Forget the ideal data scenario. What data exists today?
- Work orders with failure codes? How consistent?
- Meter readings? How frequent?
- Sensor data? Connected to asset IDs?
- Enough failure examples? For which failure modes?
Step 2: Understand What Actually Fails
Pull your corrective work order data and analyze:
- What asset types generate the most corrective work?
- What failure codes appear most frequently?
- What failures cause the most downtime and cost?
- Which failures show degradation patterns (not random)?
Step 3: Match Failures to Data
For each candidate failure mode, ask:
- Do I have 30+ historical examples?
- Are they coded consistently?
- Is there condition data (sensors, inspections) that precedes these failures?
- Can I act on a prediction (parts available, maintenance window exists)?
Step 4: Start With Your Quick Win
The use case that scores highest across all four dimensions -- data availability, failure frequency, detectable patterns, and actionable outcomes -- is your starting point.
Not the most expensive failure. Not the most politically important. The one most likely to succeed and prove the concept.
The 5 Commandments of Use Case Selection
- Data readiness beats business urgency. The best use case has the best data, not the biggest failure cost.
- Homogeneous fleets are gold. 200 identical pumps beat 200 different assets every time.
- Degradation failures are predictable. Random failures are not. Know the difference.
- Thirty failures is the minimum. Below that, you are guessing, not modeling.
- Start with one industry pattern. Prove it. Then replicate.
Find your pattern. Validate your data. Build your first win.
Next in the series: Part 8: Best Practices, Governance, and Adoption -- The human side of making predictive maintenance stick.
This is Part 7 of the MAS Predict series by TheMaximoGuys. [View the complete series index](/blog/mas-predict-series-index).
TheMaximoGuys | Enterprise Maximo. No fluff. Just results.
The Maximo Guys
