New in the library Presented by Jaadu Digital ↗
Free Business Profile Free tools
Motors, Drives and Electrical

Predictive Maintenance Starter Guide for Plants

How to start predictive maintenance without big software: rank critical machines, pick vibration, thermal and oil checks, set baselines and run a 90-day plan.

Jignesh S Panchal 7 min read
Predictive Maintenance: A starter guide for plants (cover graphic)
Approx. 8 min read
On this page
  1. Three ways to maintain, and where each fits
  2. Step 1: Rank your assets by criticality
  3. Step 2: Match techniques to failure modes
  4. The P-F curve, and why inspection frequency matters
  5. Worked example
  6. Step 3: Establish baselines
  7. Step 4: Build routes and responsibilities
  8. Step 5: Close the loop
  9. Counting the benefit honestly
  10. What I see in plants
  11. A 90-day starter plan
  12. Worked example: ranking machines and costing the first route
  13. First route sheet
  14. Go deeper: the full guide series
  15. Frequently asked questions
  16. Do I need expensive software to start?
  17. Which machines should I monitor first?
  18. How is predictive different from preventive maintenance?
  19. How often should I take vibration readings?
  20. Can small plants do this?
  21. Sources and further reading
  22. More on Motors, Drives and Electrical

Short answer: predictive maintenance means measuring the health of a machine and repairing it when the measurements show it is starting to fail, rather than on a fixed calendar or after it breaks. The sensible way to start is small: rank your machines by how costly a failure is, choose one or two simple techniques such as vibration and temperature checks, and build a habit of recording readings before buying software.

When I talk to owners and maintenance heads, almost everyone wants to move away from breakdown repair. Few know where to begin, and many have been sold monitoring systems that nobody on the shift uses. The plants that get results usually start with a clipboard, a handheld meter and a list of ten critical machines.

Three ways to maintain, and where each fits

Approach What it means Best for
Run to failure Fix it when it breaks Cheap, duplicated, non-critical items
Preventive (time-based) Service at fixed intervals Parts with a known wear-out life, safety items
Predictive (condition-based) Act when the measurement says so Critical machines where failure is costly and gives warning

Predictive maintenance is not a replacement for the others. A good plant uses all three, chosen asset by asset.

Advertisement

Step 1: Rank your assets by criticality

List all major machines and score each on: effect on production if it stops, safety and environmental risk, repair cost and lead time, and whether a standby exists. Pick the top 10 to 20. Nobody can monitor everything on day one.

Step 2: Match techniques to failure modes

Technique Finds Used on
Vibration analysis Imbalance, misalignment, bearing and gear faults, looseness Motors, pumps, fans, gearboxes (see gearbox guide)
Infrared thermography Hot connections, overloaded circuits, failing bearings, insulation gaps Electrical panels, motors, steam lines
Oil analysis Wear particles, contamination, oil degradation Gearboxes, compressors, hydraulics
Ultrasound Air and steam leaks, early bearing wear, steam trap faults, electrical discharge Compressed air, steam, bearings
Motor current analysis Rotor bar faults, load changes, supply problems Large motors
Performance trends Falling efficiency, rising power or temperature Pumps, compressors, heat exchangers

The P-F curve, and why inspection frequency matters

A failure does not happen suddenly. There is a point P where a fault first becomes detectable, and a point F where the machine fails. The time between them is the P-F interval. Your inspection interval must be shorter than that interval, and a common rule is to inspect at half of it so you get at least one chance to catch the fault.

Worked example

For a motor bearing, the vibration signature shows early wear about 6 weeks before failure. Checking every 3 weeks guarantees at least one reading inside the warning window. Checking every 3 months misses it most of the time. Bearings with a shorter P-F interval, such as grease starvation, may need continuous monitoring or a different strategy like better lubrication practice.

Step 3: Establish baselines

A reading means little without a reference. When the machine is known to be healthy, record vibration, temperature, current and pressure at the same points and loads. Then trend. A rising trend tells you more than one high value. International standards, such as ISO 10816 and the newer ISO 20816 series, give vibration severity zones by machine class; use them as a guide alongside trends.

Step 4: Build routes and responsibilities

  • Define a route: which machines, which points, which tool.
  • Train operators on simple checks: sound, heat, smell, leakage. Operators see the machine every shift.
  • Name one person to review the data every week.
  • Set alert and action limits, and write what happens when a limit is crossed.

Step 5: Close the loop

A finding that does not lead to a work order is wasted. Track each finding to completion. After the repair, record what was found: was the diagnosis right? This is how the team learns and how you justify the programme.

Counting the benefit honestly

Track a small set of numbers: unplanned downtime hours, emergency repair cost, spare parts used, and overall equipment effectiveness (see the OEE calculator). Compare the year before and after. Do not promise large percentages before you have your own data. Reported savings vary widely between plants.

Advertisement

What I see in plants

  • The best programmes are run by a technician who gets curious, not by a system. Good people matter more than the tool.
  • Owners often ask for a dashboard before they have a single reading. Start with the reading.
  • Many breakdowns come from basic causes such as lubrication errors and loose bolts. Fix those first. Predictive tools are for what remains.
  • Spare parts planning goes with it. An early warning is only useful if the bearing is in stock.

A 90-day starter plan

  1. Weeks 1 to 2: criticality ranking and a list of failure modes.
  2. Weeks 3 to 4: choose techniques, buy a basic vibration meter and a thermal camera or temperature gun, set up a log.
  3. Weeks 5 to 8: take baselines on the top machines.
  4. Weeks 9 to 12: start routes, set limits, hold the first weekly review.

Worked example: ranking machines and costing the first route

Steps 1 and 4 can be put into numbers in an afternoon. Here is an illustration with a simple additive score: production effect, safety and environment, and repair cost and lead time, each scored 1 to 5, plus a standby score of 5 for no standby down to 1 for a full standby.

  • Boiler ID fan: 5 + 4 + 4 + 5 = 18.
  • Main process pump, with installed standby: 5 + 2 + 3 + 1 = 11.
  • Air compressor, two of three needed: 4 + 1 + 3 + 2 = 10.
  • Cooling tower fan: 3 + 1 + 2 + 3 = 9.
  • Workshop grinder: 1 + 2 + 1 + 1 = 5. Run to failure.
  • The route: say the top 12 machines, 4 measuring points each, about 3 minutes per point including walking: 12 × 4 × 3 = 144 minutes, 2.4 hours a round. Every 3 weeks is 52 ÷ 3 = about 17 rounds a year, or about 41.6 hours.
  • Cost: at an assumed ₹400 an hour of technician time, 41.6 × 400 = about ₹16,600 a year, plus an assumed ₹60,000 for a basic vibration meter. First-year total about ₹76,600.
  • Against one avoided failure: if a bearing failure on the ID fan would mean 8 hours of unplanned downtime at an assumed ₹50,000 an hour, that alone is ₹4 lakh, before the emergency repair premium. One catch every five years covers the programme.
Machine (illustrative) Production Safety Repair Standby Score
Boiler ID fan 5 4 4 5 18
Main process pump 5 2 3 1 11
Air compressor 4 1 3 2 10
Cooling tower fan 3 1 2 3 9
Workshop grinder 1 2 1 1 5

The scoring scheme matters less than using the same one for every machine and agreeing it with production and safety. Revisit the list when a standby is added or removed, or when a machine becomes a bottleneck.

First route sheet

  • Machine tag and the measuring points, marked on the machine
  • Instrument used and its settings
  • Reading, unit and direction at each point
  • Bearing temperature and any unusual sound or leak
  • Load or speed at the time of reading
  • Initials, date, and any finding raised as a work order

Go deeper: the full guide series

Each guide below takes one part of a maintenance programme, from condition monitoring to KPIs, and works it through with numbers.

  • Vibration Analysis Basics: Velocity, Acceleration and Spectra
  • Bearing Lubrication: How Much Grease and How Often
  • Oil Analysis for Gearboxes and Hydraulics
  • Infrared Thermography for Electrical and Mechanical Checks
  • Root Cause Analysis: 5 Whys and Fishbone With Plant Examples
  • FMEA for Plant Equipment: A Worked Example
  • Spare Parts Classification and Stocking Strategy
  • Choosing a CMMS Without Overbuying
  • Gearbox Failures: Causes and Early Signs
  • Coupling Selection Guide
  • Belt Drive Maintenance: Tension, Alignment and Replacement
  • Ultrasonic Testing for Leaks and Bearings
  • Shutdown Planning: A Practical Timeline
  • Preventive Maintenance Schedules: Templates by Equipment
  • MTBF and MTTR: How to Calculate and Use Them
  • Improving OEE Through Maintenance
  • Maintenance KPIs That Matter
  • Setting Machinery Alarm Limits
  • Fracture Surfaces: Reading Fatigue and Overload Failures

Free tools for this topic: Maintenance checklist generator · Equipment fault finder · all free engineering tools

Frequently asked questions

Do I need expensive software to start?

No. A spreadsheet and a handheld instrument are enough for the first year in most plants.

Which machines should I monitor first?

The ones whose failure stops production, causes a safety incident, or has long repair lead times.

How is predictive different from preventive maintenance?

Preventive acts on a schedule. Predictive acts on measured condition.

Advertisement

How often should I take vibration readings?

At less than half the P-F interval for the failure you want to catch. Monthly is a common starting point for general rotating equipment.

Can small plants do this?

Yes. A small plant with ten critical machines often gets the quickest benefit because each stoppage hurts more.

General guidance. Follow lock-out and tag-out and the plant’s safety rules when taking measurements around running machinery.

Sources and further reading

For standards, data and independent guidance on this topic, see Reliabilityweb (maintenance and reliability articles), Society of Tribologists and Lubrication Engineers (lubrication and tribology resources), NETA (electrical testing and thermography standards) and NIST (measurement and calibration standards). Always check the current edition and your site conditions before applying any figure.

Free tool for this topic

Electric Motor FLA, Starter & Overload Sizing CalculatorCalculate full-load current (FLA), starting peak current by starter type (DOL, Star-Delta, Soft-Starter, VFD), thermal overload setting and minimum contactor rating.
Open calculator

Continue learning

More on Motors, Drives and Electrical

View all →
Sponsored
Jignesh S Panchal
Technical Editorial Author Verified Editor

Jignesh S Panchal

Founder — Jaadu Digital & Musafir Infotech · 9+ Years in Industrial B2B

Jignesh S Panchal is the founder of Jaadu Digital and Musafir Infotech, and the publisher behind EngiTech. With over 9 years of hands-on experience in industrial marketing and B2B growth, he works closely with machinery OEMs, process equipment manufacturers, and industrial service providers across thermal processing, fluid handling, automation, packaging, and plant reliability. Combining deep practical knowledge of industrial manufacturing with how technical buyers actually evaluate equipment, he built EngiTech to give plant engineers and procurement teams reliable sizing tools, fault diagnostics, and verified B2B supplier discovery.

Your saved guides

Saved only in this browser. No account required.

Advertisement