Overhaul every so often, and replace after so many overhauls
Before you start: A repairable model of the unit with minimal repair (Crow-AMSAA, Duane, HPP or Cox-Lewis), saved or typed in, and what a repair, an overhaul and a replacement cost.
A gearbox, an engine or a pump is repaired when it fails and carries on, a little worse each time: its failures come more often as it ages. An overhaul takes off part of that age but doesn't make it new. A replacement does. So the question is two numbers: overhaul every X, and replace after N overhauls. Overhaul too rarely and you pay for repairs; too often and you pay for overhauls. Replace too soon and you throw away a working unit; too late and the age the overhauls leave behind piles up.
Reliafy finds the X and N with the lowest long-run cost, and puts the plan beside the simpler ones, so you can see what each choice costs.
Choose the model and enter the costs
Open Decisions → Strategy → Overhaul and replace. Under Repairable model, choose Saved model for a model from Models › Repairable, or Parameters to type in a Crow-AMSAA or Duane model's scale alpha and shape beta (with its unit). The model must be a minimal-repair one: an imperfect-repair model can't say how often an overhauled unit fails, so it's refused with that reason. Times stay in the model's unit, so a model in flight hours gives an answer in flight hours.
Then enter:
- Repair cost, per failure: fixing one failure.
- Overhaul cost: more than a repair, or you'd overhaul at every failure.
- Replacement cost: a new unit fitted.
- An overhaul takes off (% of its age): 100% makes it as good as new; 50% takes off half its age. There's no default: ask whoever does the overhaul.
- Most overhauls to search (19 if left blank).
Downtime (optional) takes the hours a repair, an overhaul and a replacement keep the unit down, and the cost of an hour down. Each event's cost then includes its downtime. The model's clock stops while the unit is down, so downtime adds cost, not time.
Press Compute.
Read the answer
The answer is the plan in one sentence, with four figures beside it: how often to overhaul, after how many overhauls to replace, the cost per unit of time and the repairs expected between replacements. For a gearbox modelled as a power law with alpha 500 and beta 2.2 flight hours, a repair at 2,000, an overhaul at 15,000 that takes off 60% of its age, and a replacement at 60,000, it says: overhaul every 805 flight hours and replace after 19 overhauls.
The chart shows the cost rate against the overhaul interval for no overhauls (replace only) and for a few numbers of overhauls, with the best plan marked.
Against the other plans puts the best plan beside three simpler ones, each at its own best:
- Run to failure: repair each failure and replace the unit at the n-th failure (the best n), with no planned work.
- Overhaul only, never replace.
- Replace only, no overhauls.
Read this table before acting. The cheapest plan of another kind can beat the best overhaul-and-replace plan, and when it does, the answer says so. In the gearbox example, overhauling every 737 flight hours and never replacing costs about 5% less again, because each overhaul takes off so much age that the unit never gets old enough for a replacement to pay. A best plan at the most overhauls searched points the same way: search more and compare.
The best X for each number of overhauls lists every N with its own best interval and cost rate, and How it's worked out gives the method and its assumptions.
Save it, or work from a position
Press Save analysis to keep the answer. It's worked out again on the server from the inputs, so it can be evidence for an RCM decision.
On a repairable model's own page the same tool is under Plan ahead → Overhaul and replace. In the Explorer's Modelling tab, a model's ⋯ menu has Add a strategy…: the strategy tools for that kind of model open beside the tree, the position's class costs filled in where they're set. Once you save one there, the model has a Strategy row under it.
What it assumes
- Each failure is repaired back to how it was just before it failed (minimal repair), so the failure rate follows the unit's age.
- An overhaul takes off a fixed share of the unit's age; a replacement makes it new.
- Costs are long-run averages over many replacement cycles.
- It's exact, worked out from the model's failure intensity, so it's free.