What is Predictive Maintenance?

Predictive maintenance is a maintenance strategy that uses data from sensors and monitoring systems to predict when equipment is likely to fail, enabling maintenance to be performed just before failure occurs. Unlike reactive maintenance (fix it when it breaks) or preventive maintenance (service it on a fixed schedule regardless of condition), predictive maintenance optimises maintenance timing based on the actual condition of the equipment.

The Evolution of Maintenance Strategies

Reactive Maintenance

The simplest approach: operate equipment until it fails, then repair or replace it. While this minimises maintenance expenditure in the short term, it leads to unplanned downtime, secondary damage (a failed bearing can destroy a motor), safety hazards, and higher overall costs. Studies consistently show that reactive maintenance costs 3-5 times more than planned maintenance over an asset's lifetime.

Preventive (Scheduled) Maintenance

Equipment is serviced on a fixed schedule based on time intervals or operating hours, regardless of its actual condition. This reduces unplanned failures but introduces inefficiency: equipment may be serviced too frequently (wasting labour and parts) or not frequently enough (if usage patterns differ from the assumed schedule). Approximately 30% of preventive maintenance activities are performed more frequently than necessary.

Predictive (Condition-Based) Maintenance

Maintenance is performed when monitoring data indicates that equipment condition has degraded to a point where failure is approaching. This approach delivers maintenance at the optimal time: late enough to extract maximum useful life from components, but early enough to avoid unplanned failures and secondary damage.

How Predictive Maintenance Works

The predictive maintenance process follows a systematic cycle:

1. Data collection: IoT sensors continuously measure equipment condition parameters: vibration, temperature, electrical characteristics, pressure, flow rate, oil quality, and others.
2. Baseline establishment: During normal operation, the monitoring system establishes baseline patterns for each parameter. What does a healthy pump's vibration signature look like? What is the normal relationship between motor current and load?
3. Condition monitoring: The system continuously compares current measurements against baselines, detecting deviations that indicate developing faults.
4. Diagnosis: When an anomaly is detected, the system (or a human analyst) diagnoses the probable cause. Different fault types produce characteristic signatures in the monitoring data.
5. Prognosis: Based on the diagnosed condition and its rate of change, the system estimates remaining useful life, the time until failure becomes likely.
6. Maintenance scheduling: Maintenance is scheduled before the estimated failure date, ideally during a planned production window that minimises operational impact.

Key Monitoring Technologies

Vibration Analysis

Vibration monitoring is the most established predictive maintenance technique for rotating machinery (motors, pumps, fans, compressors, turbines). Accelerometers measure vibration at specific points on the machine, and frequency analysis reveals the specific fault type:

• Bearing defects produce characteristic frequencies related to the bearing geometry
• Shaft imbalance appears at the rotational frequency
• Misalignment produces vibration at twice the rotational frequency
• Gear tooth wear appears at the gear mesh frequency

Electrical Signature Analysis

The electrical characteristics of motors and drives provide rich information about equipment condition without requiring additional sensors beyond power monitoring. This makes it particularly attractive because the same IoT energy monitoring system used for consumption analysis can also support predictive maintenance.

Key electrical indicators include:

• Current imbalance: Unequal current across the three phases of a motor can indicate winding faults, poor connections, or supply voltage imbalance.
• Power factor changes: A decrease in power factor may indicate bearing wear (increased friction), belt slippage, or changed load conditions.
• Harmonic content: Changes in the harmonic spectrum of motor current can reveal rotor bar defects, eccentricity faults, and other electrical issues.
• Starting current analysis: The current waveform during motor starting contains information about rotor condition that is not visible during steady-state operation.
• Power consumption trends: Gradual increases in power consumption for the same output often indicate mechanical degradation (bearing wear, fouling, impeller erosion).

Thermal Monitoring

Temperature is a universal indicator of equipment health. Elevated temperatures in motors, transformers, switchgear, and electrical connections can indicate overloading, poor ventilation, insulation degradation, or high-resistance connections. Continuous temperature monitoring via IoT sensors provides early warning of developing thermal issues.

Oil Analysis

For equipment with lubricating or hydraulic oil (gearboxes, transformers, hydraulic systems), oil condition monitoring detects contamination, degradation, and wear particles. IoT-enabled inline oil sensors provide continuous monitoring rather than relying on periodic laboratory samples.

Benefits and ROI

Organisations implementing predictive maintenance typically report:

• 25-30% reduction in maintenance costs: By eliminating unnecessary preventive maintenance and reducing emergency repairs.
• 70-75% reduction in unplanned downtime: By catching failures before they occur.
• 10-20% extension of equipment life: By maintaining equipment in optimal condition and avoiding secondary damage from undetected faults.
• Reduced spare parts inventory: With advance warning of maintenance needs, parts can be ordered just in time rather than stockpiled.
• Improved safety: Equipment failures can create safety hazards (electrical faults, mechanical failures, chemical releases). Early detection reduces these risks.

The return on investment for predictive maintenance programmes is typically 5-10x the cost of the monitoring equipment and analysis software, with payback periods of 12-24 months.

Implementation Challenges

Despite the clear benefits, predictive maintenance implementation involves challenges:

• Data infrastructure: Collecting, storing, and processing continuous sensor data from hundreds of assets requires robust IoT infrastructure.
• Expertise: Interpreting vibration spectra, electrical signatures, and other diagnostic data requires specialist knowledge. Machine learning can automate much of this, but initial setup and validation require human expertise.
• Asset criticality assessment: Not every piece of equipment justifies predictive maintenance. Focus monitoring resources on critical assets where unplanned failure has the highest operational and financial impact.
• Organisational change: Shifting from time-based maintenance schedules to condition-based maintenance requires changes to maintenance processes, work order systems, and staff training.

Getting Started

A practical approach to implementing predictive maintenance:

1. Identify critical assets: Start with the 10-20% of equipment that causes 80% of downtime and maintenance cost.
2. Deploy monitoring: Install IoT sensors (vibration, temperature, electrical) on critical assets. Start with a pilot of 5-10 machines.
3. Establish baselines: Collect 2-4 weeks of normal operation data to establish baseline patterns.
4. Set initial alerts: Configure threshold-based alerts for obvious anomalies (high temperature, excessive vibration).
5. Build diagnostic capability: Train maintenance staff on interpreting monitoring data, or implement ML-based automated diagnostics.
6. Expand: Based on pilot results, extend monitoring to additional assets and refine diagnostic and prognostic capabilities.

EpiSensor for Predictive Maintenance

EpiSensor's energy monitoring platform provides a foundation for electrical signature-based predictive maintenance. By monitoring the power consumption, current, voltage, and power factor of individual motors and equipment with the ZEM wireless electricity monitor, the system captures the electrical indicators that reveal developing faults. The continuous, high-frequency data collected by EpiSensor sensors, transmitted via the ZigBee mesh network and stored in Core, provides the historical baseline and trend data needed for effective condition monitoring.

For facilities already using EpiSensor for energy management, extending the system to support predictive maintenance leverages existing infrastructure to deliver additional value without significant incremental cost.