Lesson 12 of 20 · Chapter 12
Planned Maintenance and the Predictive Paradigm
Planned maintenance has always run on the calendar: change it every month, inspect it every quarter. This chapter keeps what the calendar is good for and replaces what it is bad at. A bearing does not wear out by the month; it wears out by load, lubrication, dirt and alignment. So the plan moves, task by task, from “when the date comes” to “when the machine’s health says so,” read through three signals: vibration, temperature and current.
Choose how you want to learn
You can change this any time. Every level does the same hands-on steps. Press Listen on any part to hear it read aloud.
- 1 · Calendar or condition
- 2 · The right trigger
- 3 · The health trio
- 4 · The health card
- 5 · Build the programme
- 6 · Prescriptive maintenance
- 7 · Try it at work
- 8 · Your card
- 9 · Check yourself
Part 1
Time is a guess about wear
Time-based maintenance does a job on a schedule: every month, every 500 hours. It is simple and it creates a rhythm.
But it guesses that time tells you how worn something is. A filter does not clog because the calendar changed; it clogs because dirt builds up. If it builds slowly, the calendar changes it too early. If it builds fast, the calendar changes it too late.
Condition-based maintenance uses the machine’s health instead: act in the window where the fix is cheapest and most controlled.
Every unnecessary intervention is a chance for poor restoration: a guard refitted without its original fasteners, a restart that relies on feel. Well-meant PM programmes can create some of the very recurrence they were meant to prevent. And the calendar does not stop a fast failure: a bearing that starts spalling between monthly checks is missed.
Condition-based maintenance is not “do nothing until an alarm.” It fails when sensors are treated as certainty (a temperature can rise because the room warmed up), when alarm fatigue turns into inaction, or when predictions never become owned work with verified completion.
“But time-based maintenance has a quiet weakness that many plants tolerate for years: it assumes time is a reliable proxy for deterioration.”
“A filter does not plug because the calendar changed; it plugs because differential pressure climbed from restriction and particulates.”
“It is not “do nothing until an alarm.” It is “act in the window where intervention is cheapest and most controlled.””
Read the whole chapter free in the book reader.
One filter, two kinds of dirt load, two ways to decide when to change it. The traces are an illustration.
Part 2
Choose the trigger by the failure mode
This is not a fight between two camps. Both are tools. The question for each task is: what triggers it best?
Keep it on the calendar when the law or safety demands it (a pressure relief valve test, a lifting device inspection), or when a simple routine is the point (basic operator lubrication).
Move it to condition when time is a poor guess and there is an early warning to read: filter changes by pressure across the filter, major bearing replacements by vibration, temperature and current.
Some tasks stay time-based because the failure is age-related, the consequence is severe, or the signal is not reliable enough to justify waiting. Others move to condition because time is a poor proxy and early evidence exists that can protect planning.
Lubrication shows both at once: basic operator lubrication stays on a schedule to prevent dry running, while deeper lubrication health (oil analysis, temperature trends) becomes condition-based. The plan stops being a blanket schedule applied equally to everything and becomes a risk-based schedule grounded in evidence.
“Time-based and condition-based maintenance are both tools.”
“If a lifting device must be inspected every defined period, or a pressure relief valve must be tested on a calendar, time-based is not a debate.”
“The planned maintenance program stops being a blanket schedule applied equally to everything, and becomes a risk-based schedule grounded in evidence.”
Six maintenance tasks. Should each one run on time or on condition?
Part 3
Three signals, three languages
The most useful health signals are still three familiar ones. Vibration speaks about the machine’s mechanical soundness. Temperature speaks about friction, lubrication and cooling. Current speaks about load and drag.
Read them together. Vibration rising with steady load points to something mechanical (loose, worn, misaligned). Current rising points to something making the machine work harder.
And read them in the right state: a temperature that is fine during warm-up can be a warning if it is still there at steady running.
A common mistake is to treat vibration as one number with one threshold. More useful is a small set of vibration features linked to the failures actually seen, and an agreement on what “actionable drift” looks like on the assets where recurrence is expensive.
Temperature is one of the most honest indicators of friction and heat management, but only as a trend, with normal bands defined by mode: warm-up, steady state, and post-changeover. Current is often the earliest clue: binding, product build-up, belt tension, restriction or a process setting.
“Vibration speaks the language of mechanical integrity.”
“Temperature speaks the language of friction, lubrication, and heat rejection.”
“Current speaks the language of load, drag, and electrical stress.”
“A temperature that is acceptable during warm-up may be abnormal if it persists into steady-state.”
Three temperature readings from the same gearbox. Which are normal for their operating state? The bands are an illustration.
Part 4
A health card for each critical machine
A health card is a short, agreed standard for one machine. Not a thick document: just four things.
1. The operating states that matter and how to recognise them. 2. The normal bands for vibration, temperature and current in each state. 3. The drift rules: what change matters. 4. The response rules: what triggers an inspection, a planned job, or escalation, and how to check the work after.
When a reading crosses a line, it becomes a work item with a response window, not a comment in a report.
Health data is not special. It enters the same readable pipeline as any other work: triaged, planned, waiting parts, ready, scheduled, executed, verified, closed with learning captured. If current rises at steady state, the card triggers a check for drag and process settings, ideally with the operator, who often knows what changed first.
The card also defines done. In a time-based world, “PM complete” means boxes ticked. In the predictive paradigm it means the trio is back in its normal band in the same operating state, and stays there.
“A practical method is to build a simple “health card” for each critical asset.”
“In a time-based world, “PM complete” often means boxes checked.”
“In the predictive paradigm, “complete” means baseline restored and verified in the language of condition.”
Six lines for a filler gearbox’s health card. Which of the four parts does each one belong to?
Part 5
A narrow programme that closes loops
A predictive programme is not a set of sensors, a dashboard, or a model that sends emails. It is work that gets done and proved.
Start where closure is strongest, not where the technology is easiest: the bottleneck, the machine with chronic small stops, or the one whose failures hurt quality or safety most.
A narrow programme that closes loops beats a broad one that floods people with warnings nobody owns. The bench below lets you read the trio, set a trigger and define done.
If the plant cannot consistently triage, plan, schedule, execute, verify and capture learning for its predictions, it should not widen the programme. Accurate predictions are celebrated in the language of what did not happen: no emergency overtime, no collateral damage, no rushed restart defects, no emergency freight.
If a trend improves briefly after a repair and then drifts again, the issue may not be the prediction but the restoration: alignment not verified, mounting surfaces not cleaned, lubrication not restored to standard. Over time, as focused improvement removes chronic causes, predicted work becomes more selective and the factory calmer.
“A predictive maintenance program is not a set of sensors, not a dashboard, and not a model that sends emails.”
“The most common reason predictive efforts stall is that they start where technology is easiest rather than where closure is strongest.”
“A narrow program that closes loops will beat a broad program that floods the organization with unowned risk.”
The Health Trio Bench. Set the three signals and the operating state, read them, choose a trigger, and define done.
Part 6
One step beyond: prescriptive maintenance
A predictive system tells you a bearing is wearing out and roughly how long it has left. A prescriptive system goes one step further: it recommends what to do. It weighs the parts on hand, the production schedule and the cost of each choice.
A prescription might say: wait until Saturday’s planned window, order the bearing today, and run the machine a little slower until then. Some systems can even carry out parts of that by themselves.
The book’s advice: do not jump straight there. Get the predictive loop in this lesson working first.
A prescription is only as good as what sits under it: the data, the model of the machine’s operating states, and the discipline of closing work and proving it on the trend. A plant that cannot yet turn a forecast into planned work will not be rescued by a system that writes the plan for it.
There is also a human limit. Researchers call it algorithm aversion: people tend to stop trusting an automated recommendation after seeing it make one mistake, even when it does better than they do overall. The answer is the one TPM uses everywhere: start narrow, on a few critical assets; show people the reasoning behind each recommendation; verify the outcome; and let trust be earned by evidence.
“A prescriptive system goes further and recommends what to do about it, weighing parts on hand, the production schedule, and the cost of each option.”
“A prescription is only as good as the data, the state model, and the closure discipline beneath it, and a plant that cannot yet turn a forecast into planned work will not be rescued by a system that writes the plan for it.”
“Researchers have documented what they call algorithm aversion: people tend to lose trust in an automated recommendation after watching it make a single mistake, even when it outperforms them overall.”
Read the whole chapter free in the book reader.
Is the plant ready, and how is the system rolled out? Set both, then see what happens after its first wrong recommendation.
Part 7
Try it where you work
Do these on a real machine: at work if you work with equipment, or at home on a car, washing machine, furnace or lawn mower if you do not. Tick each one when it is done. Safety first: look and clean only what you are allowed to touch, never reach into moving equipment, and follow your site’s lockout rules.
Part 8
Your “My Health Card” card
A short health standard for one critical machine. Nothing you type here leaves this device.
My Health Card
Global Sovereign University · Modern TPM, lesson 12
Time or condition?
- Law or safety demands a date: time.
- A stable routine is the point: time.
- Time is a poor guess and a warning exists: condition.
- Read vibration, temperature and current together, by state.
- Done means back in band, not boxes ticked.
A bearing does not wear by the month. Act in the window where the fix is cheapest and most controlled.
Part 9
Check yourself
Eight real situations. Pick what you would do. You will see why after each one.
Lesson 12 complete.
You can now choose time or condition for each task, read the health trio together in its operating state, write a health card, define done in the machine’s own language, and say when a plant is ready for prescriptive maintenance. Next: Lesson 13, focused improvement and data-driven kaizen.
Modern Total Productive Maintenance (TPM) by Dr. Gene A Constant. This lesson comes from chapter twelve. Read the whole book free in the reader, or keep a copy.
Buy a copy — it funds a free seat
GSU is a 501(c)(3) nonprofit and book sales are its only income. A purchase is not a tax-deductible donation. This lesson is free and always will be.