Intervention policy

Remaining-useful-life alert threshold (days): 15

This case study models a fleet of 40 machines, each with a multi-sensor (vibration, temperature, current) estimate of remaining useful life that carries noise. Scheduling maintenance as soon as the estimate drops below the threshold avoids most failures, but a threshold set too high triggers many unnecessary early service visits.

90-day fleet outcome

Unplanned failures—

Scheduled interventions—

Net maintenance cost—

Machines by estimated remaining useful life (cyan = scheduled, red ring = failed unplanned)