Maintenance scheduling
Unplanned downtime almost always costs more than planned maintenance. Time each asset’s service to its condition, its failure risk and your production plan.
- Free 30-minute call
- No commitment
- You own the code
In short: A maintenance scheduling optimisation plans when each machine or vehicle gets serviced, based on its condition, its failure risk and the production plan. It also fits the work to available technicians, spare parts and the times when production doesn’t need the machine.
The problem
Tuesday, two in the afternoon, and production line 3 stops. A bearing has failed on a pump that was due for service next month. The technician who knows that pump is at another site, and the spare part is out of stock at the supplier. Meanwhile, a perfectly healthy compressor on the same line was stripped down last week, because its six-month service date had come up.
Signs your maintenance runs on the calendar
Every asset is serviced at the interval in the manual, whatever its state. Breakdowns still happen between services. The same technician is in the plan for two jobs at the same time, and one of them doesn’t get done. Or technicians drive from site to site for jobs that could wait. A planned stop turns into a wait because the part isn’t there. Production and maintenance argue over every hour a machine may stop. After a few days neither plan works any more.
Why the fixed calendar fails
With a fixed calendar, every asset is serviced equally often, however hard it works. A pump on light duty gets stripped down as often as one running hot. Healthy machines are serviced too often. Machines that are wearing out are checked too rarely. The calendar is easy to run, but it doesn’t know what state a machine is in.
The calendar plans each service on its own. It doesn’t check which technicians are free. So two jobs end up needing the same technician at the same time, and one of them has to wait. The calendar doesn’t check that the spare parts arrive in time. And the calendar doesn’t know when production needs the machine. Say you have hundreds of machines and a few technicians who look after all of them. Then you can hardly pick every service date by hand. Service too late, and you pay for a breakdown. Service too early, and you pay for hours, parts and a stopped production line the asset didn’t need yet.
What changes
Today, the pump is serviced every six months because the manual says so, and it fails in month five anyway. Our model reads the pump’s condition data and pulls its service forward when it starts to run hot. The job lands in a shift when production line 3 isn’t running anyway, with a free technician and the bearing already on the shelf. The healthy compressor is left alone until its condition data show wear.
Maintenance scheduling optimisation: what you get
Each asset is serviced when its condition and failure risk call for it. Every job gets a technician with the right training, such as electrical or mechanical, at a time when that person is free. The spare parts a job can’t start without are ready before the stop. So a planned stop doesn’t turn into an unplanned wait. Maintenance and production are planned together, so a service no longer lands in the middle of a production run.
We build the model on your own condition, failure and production data. We test it on your past breakdowns and stops before you use the plan in daily work. When new condition data comes in, the model updates the plan. You stop paying for services that weren’t needed and for breakdowns that were visible in the data.
- A model written around your own maintenance scheduling rules.
- Results measured in the numbers you already track.
- It slots into how you already work, and you can check every answer it gives.
How it works
Your machines, technicians and spare parts go into one model. So do the times when each machine isn’t producing. So do your machines’ condition data: readings such as temperature or vibration. The model decides in one run when each machine is serviced, by whom and with which parts. When a machine starts to wear, its service moves forward. The service is planned for a time when the machine isn’t producing anyway. And the parts are there when the work starts. You check the plan and release it.
Condition-based maintenance scheduling
There are two ways to time maintenance: by the fixed calendar, or by condition and risk. Condition- and risk-based scheduling services each asset when its condition, failure risk and cost of failure say so. A healthy, well-monitored asset is left alone. An asset that starts to wear is serviced earlier, before a breakdown stops the production line.
Does a fixed interval prevent failures at all?
It depends on how a part fails, and that’s an old finding. In 1978 two United Airlines engineers, Stanley Nowlan and Howard Heap, published a study of how aircraft components fail. Its title, Reliability-Centered Maintenance, became the name of a whole method. For most parts they studied, the risk of failure did not rise with age.
For those parts, replacing or overhauling them at a fixed age does little to prevent failures. The report even notes that an overhaul can make failures more likely. A freshly rebuilt machine can fail early, just like a new one. Those were aircraft parts, and your pumps and presses may wear out more with age. The lesson still applies. A fixed interval pays only where an asset really wears out with age. Everywhere else, watch its condition.
Preventive maintenance interval optimisation
Where an asset does wear out with age, the question is how long the interval should be. The model picks the interval that minimises expected failure cost plus servicing cost. Expected failure cost is how likely a breakdown is, times what it would cost you. Say a breakdown costs 50,000 francs and has a 10% chance in the next month. Then the expected failure cost for that month is 5,000 francs. Shorten the interval, and you cut failure risk but pay more in labour and parts. Stretch it, and you save on servicing until a breakdown comes.
Each asset has its own best interval. A pump running around the clock needs a shorter interval than the same pump on one shift. And the older the pump gets, the shorter its best interval becomes. Preventive maintenance planning software should find that point for each asset separately, not once for a whole asset class.
Technicians, spares and production windows
A plan that says what to do but not who is free is of no use. So the schedule gives each job a technician with the right training, at a time when that person is free. It respects each technician’s shifts and the travel time between sites. A service only helps if the parts are there when the asset is opened. Otherwise you open the machine, find the bearing missing and pay for the downtime twice.
Most assets can’t be serviced while they produce. The work has to fit the times when the machine isn’t producing anyway: changeovers, planned shutdowns and shifts without orders. That’s why the model plans maintenance and production together. It also makes sure critical spare parts arrive before each window.
Fleet maintenance scheduling
With many assets, you have to plan maintenance for the whole fleet at once. A fleet can be trucks, wind turbines or a plant with 40 milling machines. It can mean thousands of service jobs. Each job concerns an asset with its own condition and failure risk, and it needs its own parts and has its own possible dates. Fleet maintenance scheduling allocates limited technicians, parts and downtime across all assets at once. You save through the order of the jobs: which asset is serviced when, so that the parts last and never too many assets stop at once?
Constraint programming and stochastic optimisation
The method is called constraint programming. The model first writes down every rule as a condition. Only then does it search for a schedule that keeps every one. Each machine gets a date for its next service and an interval for the ones after. Then the model assigns technicians and parts to those dates. So you can carry out the plan as it stands: every technician is free and every part is there.
Stochastic optimisation also plans for what is uncertain, which a fixed calendar ignores. Stochastic means the model takes chance into account. A machine may fail sooner or later than average. A job may take longer, and a part may arrive late. So the schedule still works when a part arrives late or a job runs long. You get the maintenance programme with the lowest expected cost for the whole fleet.
Common questions.
Something else? Ask us directly.
Many plants start from the manufacturer’s interval and adjust it for how hard the asset runs. That’s a fair start. Better is to ask how the part fails. If failures rise with age, set the interval where servicing cost plus expected failure cost is lowest. If failures are random, a fixed interval barely helps, and condition checks do more.
No. Condition-based means you service an asset when its condition calls for it, not on a fixed calendar. Start with the assets whose failure stops a production line or costs a lot. Focus on faults that show up early: through vibration, heat or wear. Rounds where a technician checks machines by hand count as condition data too. For cheap, non-critical items, running to failure is often the sensible choice. The model puts sensor-monitored assets, assets checked by hand and run-to-failure items into one plan.
When a failure is cheap, quick to fix, harmless to people and stops nothing that matters. A fan with a second fan as backup is a typical case. Servicing it on a schedule costs labour and can add early faults after reassembly. The model treats run-to-failure as one more option. It picks that option only where servicing would cost more than the failure.
Sometimes it can, and if that’s the case, we’ll tell you. Standard software is built for an average operation, though. Yours runs on its own rules: your contracts, your exceptions, the things you know by heart. A custom model writes those rules in, and that’s where the better plan comes from.
You don’t need either to start. We build the model from the data you’ve already got, even if it lives in three spreadsheets and one person’s head. You get a system you run yourself. We train your team until they no longer need us.
You see a first result with the prototype, before you commit to anything bigger. We start with a free 30-minute call about your operation and the model that would fit it. Then we build a working prototype and run it on your numbers.
Similar pages.
Terms on this page
Every abbreviation explained in plain words, in our article “Metrics and methods”.
Tell us what you’re planning.
In a free 30-minute call with a specialist, you find out whether optimisation can help.
- Free, no commitment
- You talk to a specialist directly
- A clear go or no-go