This direction closes the manufacturing domain with the finding that runs through all ten: the technology is bought and the capability to use it is not.
The alert nobody can act on
Condition monitoring and predictive analytics genuinely work. Vibration, temperature, oil analysis and current signature can identify developing faults well before failure.
What happens next is where the value is created or lost.
A system that tells you a bearing will fail in three weeks has given you three weeks. If your maintenance team is fully occupied responding to things that already failed, it has given you three weeks of knowing.
Four requirements for a prediction to be worth anything.
Capacity to act. People with time not already consumed by reactive work.
Planning capability. Someone who can schedule the intervention into a production window, order the part and arrange the resources.
Spare parts. A three-week warning is no use if the lead time is eight weeks and nothing is in stock.
Production agreement. A window to do the work, which requires production to accept planned downtime instead of hoping the machine lasts.
Organisations that implement monitoring without these four have bought better information about failures they will still experience.
What the direction covers
The scope: preventive and predictive maintenance, reliability-centred maintenance and the management of industrial assets.
Four areas.
Maintenance strategies. Choosing the right approach per asset, covered below.
Condition monitoring. Techniques for detecting developing faults and interpreting what they indicate.
Planning and scheduling. Turning identified work into completed work, which is the bottleneck in most organisations.
Asset management. Whole-life decisions about repair, replace, upgrade and retire.
The four maintenance strategies and when each is right
The framework that prevents the most common mistake, which is applying the most sophisticated strategy everywhere.
Run to failure. Let it break, then fix it. Correct when failure is cheap, safe, quick to fix and does not stop anything important. Applying condition monitoring to such an asset wastes effort that a critical asset needs. Deliberately choosing this is a decision; arriving at it by neglect is not.
Preventive, on a schedule. Replace or service at fixed intervals. Appropriate where failure is age-related and the interval is known. Its weakness is well documented: for many components failure is not age-related, so scheduled replacement introduces failures through the disturbance of the intervention itself while achieving nothing.
Condition-based. Monitor and act when the condition indicates. Appropriate where a measurable indicator exists and there is enough warning to act. This is where predictive analytics belongs.
Design out. Modify so the failure cannot occur. The most permanent fix and the one least often chosen, because it requires engineering effort rather than maintenance effort.
Reliability-centred maintenance is the structured method for deciding which applies to each asset, based on how it fails and what the consequences are. Done properly it typically reduces total maintenance work while improving availability, because it stops effort going where it does not matter.
Where this sits in the domain
Maintenance and asset management is the tenth and final direction in Astra Trainer's advanced manufacturing domain, connecting to industrial automation and control for equipment diagnosis, to industrial IoT and digital twins for condition data, and to quality engineering because equipment condition drives process variation.
It also connects to advanced materials for failure analysis and corrosion, and to robotics for robot maintenance as installed bases age. Partners implementing predictive maintenance usually need the planning and capacity layer before the analytics layer, which is the sequencing most often reversed. Lessons are five minutes, which suits shift-based maintenance teams. You can see the ten directions here.
Why the reactive trap is self-reinforcing
The dynamic that keeps organisations stuck, described plainly because escaping it requires recognising it.
A maintenance team spending most of its time on breakdowns has no time for planned work. Planned work not done produces more breakdowns. More breakdowns consume more time.
Four features that make it stable.
Firefighting is visible and rewarded. Fixing a broken machine quickly is recognised. Preventing a failure that would have happened is invisible, which is the same attribution problem as corrosion prevention and training measurement.
Planned work is always deferrable. There is always a reason to postpone it for production, and each postponement is individually reasonable.
Reactive work is inefficient. Unplanned jobs take longer, need parts that are not staged and are done under pressure, so the same work consumes more hours than it would planned.
Skilled people leave. Permanent firefighting is exhausting and unsatisfying, and the technicians with options go elsewhere.
Escaping it requires temporarily protecting planned work capacity, which means accepting short-term pain for a benefit that arrives later. That is a management decision rather than a maintenance one, and it is the reason most organisations do not escape.
The knowledge that leaves with people
The workforce risk specific to this direction, and it is the same pattern as the metallurgy and refractories articles.
An experienced maintenance technician knows which machine runs hot in summer, which alarm is genuine and which is spurious, which fault on line three is actually caused by something on line two, and what a particular noise means.
None of that is in the manual, and it is what makes the difference between an hour of downtime and a shift of it.
Three implications.
Overlap is the only reliable transfer. Months of working alongside, on real faults, not a handover document.
Capture what can be captured. Fault history with actual causes, not just closure codes. Most computerised maintenance systems record what was done and not why it happened, which destroys the learning.
Structured learning accelerates the rest. A technician who understands the systems extracts far more from time alongside an expert than one starting from pattern memorisation.
This connects directly to the mechatronics article in the robotics domain: maintenance technicians on modern equipment are already reasoning across mechanical, electrical and control layers, and giving them the theory turns pattern recognition into transferable capability.
The roles, named
Maintenance technicians. Multi-skilled, and the shortage everyone has.
Maintenance planners. The role that converts identified work into completed work, and the most under-filled position in most maintenance organisations.
Reliability engineers. Failure analysis, strategy and eliminating recurring problems.
Condition monitoring technicians. Vibration analysis, thermography, oil analysis, with recognised certification routes.
Asset managers. Whole-life and capital replacement decisions.
CMMS administrators. Making the maintenance system produce usable information rather than compliance records.
Lubrication specialists. Unglamorous, and a substantial share of bearing failures traces to lubrication.
Shutdown and turnaround planners. Where a year of work is compressed into days.
Who can be trained into it
Maintenance technicians. Into reliability engineering, condition monitoring and planning. They hold the equipment knowledge and frequently lack the analytical framework, which is the teachable half.
Machine operators. Into basic care and early fault detection. Operators notice changes long before any sensor, and involving them is the cheapest condition monitoring available.
Production planners. Into maintenance planning, where scheduling against capacity is the same skill.
Electricians and instrumentation technicians. Into multi-skilled maintenance on automated equipment.
Engineers. Into reliability and asset management, needing the practical failure knowledge technicians hold.
Data analysts. Into condition analytics, needing the equipment context that distinguishes a real developing fault from a sensor artefact.
Most serious incidents happen during maintenance. Maintenance work involves energy isolation, work at height, confined spaces, machinery hazards and hazardous substances, and is governed by regulation with duties around safe systems of work, permits, isolation and competence. Serious injuries occur disproportionately during maintenance, intervention and fault clearance rather than during normal operation, because protective measures are frequently reduced. Training builds technical understanding and hazard awareness. It does not confer isolation authorisation, permit authority, task-specific competence or any permission to work on equipment.
What to take from this
Predictive maintenance produces warnings, and a warning is worth nothing without capacity, planning, parts and a production window.
Not every asset deserves the same strategy, and deliberately choosing run to failure for some equipment frees effort for the assets that matter.
Scheduled replacement introduces failures where failure is not age-related, which is more common than the practice assumes.
The reactive trap is stable because firefighting is visible, planned work is always deferrable, reactive work is inefficient and good technicians leave.
And the diagnostic knowledge that separates an hour of downtime from a shift of it lives in people, not manuals, and transfers only through overlap.
Why do predictive maintenance programmes disappoint?
Because a prediction requires capacity to act, planning capability, available spare parts and a production window. Without those four, monitoring produces better information about failures that still happen.
Should every asset be monitored?
No. Run to failure is correct for equipment where failure is cheap, safe, quick to fix and stops nothing important. Applying sophisticated strategies everywhere consumes effort the critical assets need.
Is scheduled preventive maintenance always good?
No. Where failure is not age-related, scheduled replacement achieves nothing while introducing failures through the disturbance of the intervention itself.
Why is the reactive trap hard to escape?
Because firefighting is visible and rewarded while prevention is invisible, planned work is always deferrable for production, reactive work consumes more hours than planned work, and skilled technicians leave.
Who converts into these roles?
Maintenance technicians into reliability, condition monitoring and planning; operators into basic care and early detection; and production planners into maintenance planning, which is the same scheduling skill.