What Is Predictive Maintenance of Equipment (PdM)

Predictive maintenance (PdM) is a strategy in which repairs are scheduled according to the actual condition of a machine rather than the calendar. Condition is measured with vibration analysis, ultrasound, thermography, oil analysis and motor current analysis. The aim is to spot the onset of degradation and remove its cause before the component fails, and to plan the work without an unplanned shutdown.

Definition

The U.S. Department of Energy (DOE) Operations & Maintenance Best Practices Guide defines predictive maintenance as measurements that detect the onset of system degradation, allowing causal stressors to be eliminated or controlled before any significant deterioration in the component's physical state. The key difference from preventive maintenance is that the need for maintenance is based on the actual condition of the machine rather than a preset schedule.

The general procedures for setting up a machine condition monitoring programme are described in the international standard ISO 17359:2018 "Condition monitoring and diagnostics of machines — General guidelines". It applies to all machines and refers to the standards for individual techniques.

Four maintenance strategies compared

Strategy	When work is done	Advantages	Disadvantages
Reactive (run to failure)	After a breakdown	No maintenance cost while the machine runs	Unplanned downtime, overtime, secondary damage, large spare parts inventory
Planned preventive	On a calendar or run-hours schedule	Fewer failures, life closer to design; per DOE, 12–18% savings over reactive	Catastrophic failures still possible; some work is unneeded, and unnecessary disassembly can damage components
Condition-based	When a measured parameter exceeds an alarm limit	Work only where it is needed	Requires instruments, trained staff and justified alarm limits
Predictive (PdM)	Based on the condition trend, planned before failure	Per DOE, 8–12% savings over preventive; less downtime and lower parts costs	Investment in diagnostic equipment and training; savings not readily seen by management

In practice condition-based and predictive maintenance are closely related: both rely on measurements. The difference is one of emphasis. The first acts when an alarm limit is exceeded; the second follows the trend and plans work in advance while the fault is still developing. The DOE guide treats them together as maintenance based on actual condition.

What the DOE figures say

The DOE guide (Release 3.0, 2010) gives the following estimates:

Indicator	Value
Preventive savings over a reactive programme	12–18%
Predictive savings over a preventive-only programme	8–12%
Return on investment in a predictive programme (industrial average survey data)	10 times
Reduction in maintenance costs	25–30%
Elimination of breakdowns	70–75%
Reduction in downtime	35–45%
Increase in production	20–25%

The same guide describes the maintenance mix: at an average facility over 55% of work is reactive, 31% preventive and 12% predictive. At consistently top-performing facilities, reactive work is under 10%, preventive 25–35% and predictive 45–55%. These are U.S. figures; the result at a particular plant depends on the starting condition of the equipment and the share of emergency repairs.

The P-F curve: why catch faults early

The P-F curve shows how a component's condition deteriorates over time:

• point P (potential failure) — when a developing fault can first be detected;
• point F (functional failure) — the component can no longer perform its function;
• P-F interval — the time between them, the window for planning the repair.

Methods applied to a running machine that detect a fault earlier move point P to the left and widen the window. Closer to failure, the fault becomes evident through heat and noise. How this looks for a rolling bearing is covered in 4 stages of bearing failure.

Which technologies are used

Technology	What it detects	Where it is used
Vibration analysis	Unbalance, misalignment, mechanical looseness, rubs, resonance, bearing and gear defects	Pumps, electric motors, fans, gearboxes, turbine sets
Ultrasound (airborne and contact)	Early bearing defects and lack of lubrication, gas and vacuum leaks, failed steam traps, arcing and corona in electrical equipment	Bearings, compressed air and steam systems, valves, switchgear
Infrared thermography	Overheated contacts, bearings and belts, damaged insulation and refractory	Electrical panels, transformers, motors, boilers, heat exchangers
Oil and wear particle analysis	Oil degradation and contamination, water, wear debris	Gearboxes, hydraulic systems, engines, transformers
Motor current and electrical analysis (MCSA)	Turn-to-turn and phase-to-phase shorts, insulation deterioration, and some mechanical faults: misalignment, unbalance, gear wear	Electric motors and the machines they drive

According to NASA research cited by DOE, a 12–50x increase in the amplitude of a monitored ultrasonic frequency (28–32 kHz) can give an early indication of bearing deterioration. For electric motors, DOE notes that infrared and vibration do not provide all the answers on motor condition, so current analysis is added. More on the vibration method is in What is equipment vibration analysis.

How to start: a step-by-step plan

1. List your equipment and rank it by criticality. Consider the consequences of failure for production and safety and the cost of downtime. The DOE guide, citing NASA, offers a simple rule: small and non-critical items can stay on reactive maintenance, wear-prone and consumable items on preventive, and critical equipment and items with random failure patterns on predictive.
2. For each critical machine, identify the main failure modes and choose a method that detects them (table above). For a motor-driven pump this is usually vibration and ultrasound on the bearings plus motor current analysis.
3. Build a measurement route. Number the measurement points, fix their positions (markers or permanent mounting pads) and set the frequency. The interval between measurements must be clearly shorter than the P-F interval, otherwise a fault will slip through.
4. Take baseline readings. The first measurements on a healthy machine become the reference for the trend. To get started, DOE accepts generic vibration severity charts but stresses that these are not absolute limits: alarm limits are then refined from your own equipment's history.
5. Watch the trend and plan the work. A rise relative to the baseline is a reason for in-depth diagnostics and a planned repair at the next shutdown.
6. Train staff and assign responsibility. DOE stresses that equipment for an in-house programme should not be purchased without a serious commitment to implementation, operator training and instrument upkeep.

Other strategies that help reduce downtime are covered in 10 maintenance strategies that reduce downtime.

Which equipment to choose

• Ultrasound for bearings, lubrication, leaks and electrical equipment — SDT340.
• Portable vibration analyzer for route measurements — vb8.
• Online monitoring of electric motors and drives from current and voltage — Artesis e-MCM.
• On-site oil analysis — field oil analysis lab.

KEG TRK supplies predictive maintenance equipment in Kazakhstan. Price on request.

Sources

• U.S. Department of Energy, FEMP / PNNL. Operations & Maintenance Best Practices — A Guide to Achieving Operational Efficiency, Release 3.0 (PNNL-19634), 2010, chapters 5 and 6.
• ISO. ISO 17359:2018 Condition monitoring and diagnostics of machines — General guidelines.
• Fluke. P-F curve: stay ahead of equipment failure.