Decision Tree — Overhaul, Monitor or Run to Failure (Risk-Averse)
A maintenance strategy for a critical pump, minimising 12-month cost with a risk tolerance of $400k: condition monitoring has the lowest expected cost ($81.0k) but its tail of unplanned trips gives it a certainty equivalent of $94.7k, so the risk-averse choice is the planned overhaul (CE $87.6k). Illustrative values.
Make it your own.
title "Boiler feed pump P-101B — overhaul, monitor or run to failure?"
subtitle "12-month cost per pump, US$ thousands — parts, labour and lost generation"
currency $
unit k
objective min
measure "E[cost]"
risk-tolerance 400
risk-profile on
branches orthogonal
note "Risk tolerance of US$400k is the plant's stated appetite for a single unplanned loss."
note "Illustrative failure probabilities for a fictional plant; calibrate them from your maintenance history."
decision "Maintenance strategy"
"Planned overhaul now" cost 85 -> chance "After overhaul"
"Runs clean" p rest -> 0
"Infant-mortality failure" p 0.04 -> 60
"Vibration monitoring" cost 18 -> chance "Condition over 12 months"
"No degradation" p 0.45 -> 0
"Fault caught early" p 0.40 -> 70
"Fault caught late" p 0.10 -> 140
"Missed: unplanned trip" p rest -> 420
"Run to failure" -> chance "Failure in 12 months?"
"No failure" p 0.45 -> 0
"Fails, spare on site" p 0.33 -> 160
"Fails, 6-week lead time" p rest -> 610