Technical Guide

IoT Predictive Maintenance: From Corrective to Condition-Based

Switch to IoT-based condition-based maintenance with vibration, temperature, and current sensors, predictive alerts, and CMMS integration.

IoT-based predictive maintenance enables a shift from reactive maintenance to maintenance driven by the actual condition of the equipment. Instead of waiting for a failure or replacing parts according to a fixed schedule, sensors that monitor vibration, temperature, current, pressure, and process variables detect gradual deviations. Eziwan collects this field data, puts it into context, alerts maintenance teams, and feeds the CMMS to trigger the right interventions at the right time.

The Problem

The industry is often caught between two imperfect approaches. Corrective maintenance waits for a breakdown before taking action. It can lead to unplanned downtime, emergency repairs, production losses, rush orders for parts, and intense pressure on field crews. Systematic preventive maintenance improves the situation, but it relies on fixed intervals that do not always reflect the machine’s actual condition.

Condition-based maintenance, which relies on measured equipment conditions, offers a more robust approach. However, it requires reliable data that is collected regularly, interpreted correctly, and integrated into existing maintenance processes.

The challenges are very real.

  • Unplanned downtime is costly, but its impact depends heavily on the industry, the type of production line, the load factor, material costs, and the duration of the downtime.

  • Systematic preventive maintenance may result in the replacement of components that are still usable or, conversely, may be performed too late on equipment that has deteriorated rapidly.

  • Teams lack visibility into the actual condition of the machines between rounds or scheduled inspections.

  • Weak signals such as excessive vibration, overheating, current imbalance, cavitation, or pressure drift are not always detected without the proper instrumentation.

  • Technicians are called in on an emergency basis to address outages that sometimes showed warning signs several days or weeks in advance.

  • The CMMS is often updated after a service call, when it should also be receiving data from the field to help with prioritization and planning.

  • Spare parts inventories are kept on the conservative side due to a lack of visibility into the actual condition of critical equipment.

  • Process data, such as load, flow rate, pressure, or ambient temperature, are rarely correlated with signs of degradation.

  • Basic monitoring systems sometimes generate too many alerts, causing operators to ignore them.

  • Vibration analysis is valuable, but it is difficult to apply it to an entire fleet without decision-support tools and internal knowledge management.

The challenge, therefore, is not just to install sensors. We need to create a complete process: measurement, connectivity, analysis, alerts, validation, work orders, feedback, and continuous improvement.

Corrective, Preventive, Conditional, and Predictive

These terms are often confused. To structure a maintenance process, it is necessary to distinguish between maturity levels.

Maintenance TypeTriggerBenefitLimit
CorrectiveDetected failureEasy to understandUnplanned downtime, emergency, high cost
Systematic PreventiveSchedule or operating hoursPlanableMay be performed too early or too late
ConditionalMeasured equipment conditionMore targeted interventionsRequires reliable sensors and thresholds
PredictiveDrift model and probable failure timelineMore precise anticipationRequires historical data, data quality, and validation
PrescriptiveOptimized action recommendationsAdvanced decision supportRequires process maturity and CMMS integration

In many factories, the best place to start isn’t a complex predictive model. It’s robust condition-based maintenance, with reliable metrics, understandable trends, and actionable alerts.

Our Approach

Eziwan helps build an IoT-based condition-based maintenance chain by connecting critical equipment to a monitoring and analytics platform. Sensors can measure vibration, temperature, current, pressure, or other relevant parameters depending on the machine. The data is collected by an industrial gateway, transmitted securely, analyzed over time, and transformed into actionable alerts or reports.

The equipment typically involved includes:

  • Electric motors.

  • Centrifugal pumps.

  • Compressors.

  • Industrial fans.

  • Gearboxes.

  • Conveyors.

  • Shredders.

  • Refrigeration units.

  • Critical rotating machinery.

  • Equipment that is isolated or difficult to access.

The goal is to classify equipment by risk level, detect deviations, explain alerts, and provide the CMMS with operational recommendations.

Architecture of an IoT-Based Condition-Based Maintenance System

An effective architecture combines sensors, local data acquisition, connectivity, an analytics platform, alerts, and CMMS integration. Each component must be reliable; otherwise, the system as a whole becomes difficult to operate.

The local buffer is important on sites where connectivity may be unstable. It prevents the loss of critical data during a temporary outage.

Data to Be Collected Based on Equipment

Not all machines require the same sensors. A good strategy starts with the likely failure modes.

EquipmentUseful MeasurementsDetectable Defects
Electric motorCurrent, temperature, vibrationImbalance, overload, bearing failure
Centrifugal pumpVibration, pressure, flow rate, temperatureCavitation, misalignment, bearing failure, clogging
FanVibration, current, speedUnbalance, fouling, misalignment
GearboxVibration, housing temperature, oilGear wear, lubrication, bearing
CompressorVibration, temperature, pressure, currentLeakage, overheating, mechanical failure
ConveyorCurrent, vibration, speedFriction, overload, misalignment

The key lies in the combination of factors. An increase in vibration may be normal if the process load has increased. It becomes cause for concern if it increases at a constant load, accompanied by rising bearing temperature and unstable motor current.

3-Axis Vibration Monitoring

Vibration is one of the most useful indicators for rotating machinery. A 3-axis accelerometer measures vibrations in multiple directions, which helps identify issues such as imbalance, misalignment, mechanical play, bearing failure, or mounting problems.

Common indicators include:

  • RMS vibration level.

  • Vibration peak.

  • Crest factor.

  • Changes by axis.

  • Comparison with rotational speed.

  • Trend over several days or weeks.

  • Threshold exceeded based on machine criticality.

Vibration monitoring must be interpreted in context. A universal threshold is not always reliable: a small pump, a large fan, and a grinder do not all behave the same way mechanically.

FFT Spectral Analysis

The FFT, or Fast Fourier Transform, converts a time-domain signal into a frequency-domain spectrum. It helps identify the characteristic frequencies of a fault.

Examples of indicators analyzed:

  • Rotation frequency.

  • Harmonics caused by misalignment.

  • Bearing failure frequencies, such as BPFI, BPFO, BSF, or FTF, when the bearing geometry is known.

  • Unbalanced signatures.

  • Broadband noise.

  • Appearance of side stripes.

FFT is powerful, but it requires high-quality data: an appropriate sampling rate, a properly positioned sensor, a known rotation speed, and sufficient historical data.

Spectral analysis must be combined with business validation. An alert is not an automatic conclusion: it indicates a probability of a defect that must be confirmed based on the context.

Motor Current Signature Analysis

Motor current signature analysis, often referred to as MCSA, uses electrical current to detect certain mechanical or electrical faults. It can be useful when it is difficult to install vibration sensors or when the goal is to supplement the diagnosis.

It can help identify:

  • Unbalanced diet.

  • Overload.

  • Abnormal load variation.

  • Rotor defect, as applicable.

  • A mechanical issue that affects fuel consumption.

  • Performance drift or increased friction.

The advantage is that it relies on an electrical measurement that is sometimes easier to install. The limitation is that the interpretation depends heavily on the type of motor, the load, the drive, and the operating speed.

Temperature, Pressure, and Process Context

Temperature alone is not always enough to predict a failure, but it becomes very useful when combined with other signals. A rise in the temperature of the bearings, the gearbox housing, or the motor windings may indicate a lubrication problem, overload, friction, or insufficient ventilation.

Process variables help reduce false positives.

SignalRaw InterpretationContext to be Verified
Increasing vibrationPossible mechanical faultLoad, speed, flow rate
Increasing bearing temperatureFriction or lubricationAmbient temperature, operating time
High motor currentPossible overloadSetpoint, product, pressure
Unstable pressurePossible hydraulic problemValve, flow rate, filter, cavitation
Reduced flow ratePossible clogging or wearProcess setpoint, line condition

Multisensor correlation is essential for distinguishing between actual degradation and normal variations caused by operating conditions.

Health Scoring of Equipment

A health score makes it easier to set priorities. It does not replace maintenance expertise, but it helps classify equipment based on its risk level.

A score can combine several dimensions:

  • Vibration drift.

  • Abnormal temperature.

  • Power instability.

  • Alert history.

  • Equipment criticality.

  • Time since the last service visit.

  • Operating conditions.

  • Confirmation from multiple sensors.

Example of scoring logic:

score_sante:
base: 100
penalites:
vibration_hausse_7j: 15
temperature_roulement_haute: 20
courant_instable: 10
alerte_recurrente: 15
equipement_critique: 10
niveaux:
vert: "80-100"
jaune: "60-79"
orange: "40-59"
rouge: "0-39"

Thresholds must be tailored to the plant and validated with the teams. A poorly calibrated score breeds mistrust; a score that is explained becomes a planning tool.

Bidirectional CMMS Integration

Operational value is realized when IoT alerts trigger a maintenance process—not just a notification. CMMS integration links field detection to the work order.

Useful integration may include:

  • Automatic creation of a service request.

  • Linking alerts to the CMMS.

  • Priority based on severity and health score.

  • Trend lines are attached.

  • Assignment to a team or a contract.

  • Feedback after the event.

  • Enriching the model with actual diagnostic data.

Feedback is often overlooked. However, it helps determine whether an alert was actually caused by a bearing defect, misalignment, process overload, or a false positive.

Automatic Maintenance Reports

Automated reports provide a regular overview of the fleet’s status and facilitate planning meetings. They must be understandable to maintenance, production, and industrial management.

A useful report includes:

  • Classification of equipment by risk level.

  • Changes in health scores.

  • Alerts generated and processed.

  • Planned or completed procedures.

  • Equipment drifting slowly.

  • Sites with no data or inactive sensors.

  • Inspection recommendations.

  • Summary of gains or savings, when the data supports such calculations.

Estimated savings without a clear basis should be avoided. Savings should be calculated based on actual historical data: avoided downtime, reduced downtime duration, consolidated maintenance, fewer emergencies, or improved availability.

Detection of Cavitation and Hydraulic Imbalance

Centrifugal pumps are natural candidates for condition-based maintenance. Cavitation, hydraulic imbalance, clogging, recirculation, or impeller wear can manifest as vibration, noise, pressure instability, flow rate fluctuations, and overheating.

Detection requires the cross-correlation of multiple signals.

  • High-frequency vibration.

  • Pressure variation.

  • Fluctuating flow rate.

  • Fluctuating motor current.

  • Temperature of the bearing or pump housing.

  • Operating conditions outside the nominal range.

A cavitation alert must be interpreted in conjunction with the process data. A pump operating far from its optimal efficiency point can exhibit symptoms similar to those of mechanical failure.

Reducing False Positives

A system that issues too many alerts becomes useless. Operators eventually start ignoring the notifications, even when a real outage is about to occur. Reducing false positives is therefore a key objective.

Best Practices:

  • Set thresholds by equipment type.

  • Use dynamic thresholds based on load or speed.

  • Cross-reference data from multiple sensors before issuing a critical alert.

  • Distinguish between informational alerts, alerts requiring monitoring, and urgent alerts.

  • Add a delay to prevent isolated spikes.

  • Compare with the machine’s history, not just with a generic threshold.

  • Incorporate feedback from technicians.

  • Review the thresholds after the first few weeks of operation.

An alert must be actionable: it must explain what has gone wrong, how long it has been going on, the level of risk involved, and what action is recommended.

Phased deployment across an industrial complex

An IoT predictive maintenance project should start with critical equipment, not the entire fleet. A phased rollout allows for the validation of sensors, connectivity, thresholds, and CMMS integration.

StepObjectiveExpected Result
MappingIdentify critical equipmentPrioritized list
Failure Mode AnalysisSelect appropriate measuresSuitable sensors
PilotInstall sensors on a few machinesField validation
Initial ThresholdsDetect simple deviationsActionable alerts
CMMS IntegrationLink alerts to interventionsOperational process
ExpansionDeploy by equipment familyMonitored fleet
OptimizationReduce false positivesMore reliable model

This approach avoids generating large amounts of data without a clear maintenance purpose.

Maintenance Data Security and Connectivity

Maintenance data may contain sensitive information, such as machine availability, production rates, recipes, workload, recurring incidents, or equipment criticality. Therefore, the collection of this data must be secure.

Best Practices:

  • Use a secure industrial gateway.

  • Quantify the data flows between the site and the platform.

  • Segment sensors, PLCs, and IT systems.

  • Limit access rights to data.

  • Log access and changes.

  • Ensure local operation in the event of a network outage.

  • Integrate monitoring with OT cybersecurity policies.

This security is particularly important when maintenance involves third-party service providers, OEMs, or connected service contracts.

Checklist Before Starting a Project

ControlQuestionPriority
CriticalityAre priority machines identified?High
Failure ModesAre the expected faults known?High
SensorsDo the selected measurements correspond to the faults?High
ConnectivityCan data be reliably transmitted?High
HistoryIs there data on failures or maintenance interventions?Medium
CMMSIs the equipment properly cataloged?High
AlertsAre the thresholds actionable?High
Field FeedbackCan technicians assess the alerts?High
SecurityAre data flows encrypted and segmented?High
ManagementHas a deployment manager been designated?High

Success depends as much on maintenance organization as it does on technology.

Key Performance Indicators to Monitor

To measure the effectiveness of IoT-based condition-based maintenance, it is necessary to track metrics over time.

MetricObjective
Number of unplanned outagesMeasure the reduction in outages
MTBFTrack the mean time between failures
MTTRMeasure the mean time to repair
False positive rateImprove alert quality
Confirmed alert rateValidate the model’s accuracy
Time between alert and interventionAssess responsiveness
Equipment in the red zonePrioritize maintenance
Planned interventions based on alertsMeasure actual usage
Machine availabilityLink maintenance and production
Maintenance cost per piece of equipmentTrack the economic impact

Earnings must be calculated based on these metrics; they should not be announced without a specific reference to the site.

How Eziwan Helps You Make the Switch to Condition-Based Maintenance

Eziwan provides the building blocks needed to connect industrial equipment, collect useful data, detect anomalies, and link alerts to maintenance processes.

NeedEziwan’s ResponseBenefit
Collect measurementsSensors and industrial gatewayReliable field data
Connect sitesSecure telemetryMulti-site monitoring
Detect deviationsTrend analysisAlerts before probable failure
PrioritizeHealth scoreClearer planning
Reduce false positivesMulti-sensor correlationMore relevant alerts
Take actionCMMS integrationActionable work orders
MonitorAutomatic reportsMaintenance management
SecureTunnels and access controlOT data protection

This approach can leverage the Eziwan gateway, monitoring via the Eziwan cloud, and industrial connectivity architectures.

Conclusion

IoT predictive maintenance doesn’t start with a magic algorithm. It starts with reliable condition-based maintenance: carefully selected sensors, consistent data, process context, understandable thresholds, actionable alerts, CMMS integration, and feedback from the field. It is this complete chain that enables a gradual shift from emergency repairs to planned maintenance.

Eziwan supports this transition by collecting early warning signs from critical machinery, analyzing trends, classifying equipment by risk level, and triggering the appropriate actions in maintenance tools. For manufacturers, the goal is clear: to reduce unplanned downtime, better plan maintenance interventions, and leverage field data to sustainably improve the reliability of their equipment fleet.

Further Reading

Frequently Asked Questions

You might also like