Presentation — Calculation memo (1 to 4 years)

The PC200 fleet delivered up to the cutoff has 2.539 machines. To measure the Y1 failure rate we only consider machines with ≥ 12 months since technical delivery: 2.265. For Y2, ≥ 24 months: 2.106.

In hours, the fleet accumulated 9.413.700 h of observed exposure up to ~3.612 h SMR — basis for the per-bucket calculations below.

Y2/Y1 ratio of failure rate: 0.37×. Year 2 shows fewer claims than Year 1 — benign profile so far.

By this path, the Year-2 cost per machine is R$ 1.851.

The time-based window treats every machine equally, but models with highly variable usage have under/overestimated rates. For this model the median is 1.806 h/year; reconverting the history into 250 h buckets isolates the cost per operating hour at each usage band and projects Y2 over the band each machine actually reaches.

For each machine we sum the cost of the buckets overlapping each 12-month window: Y2 = [1.806h, 3.612h], Y3 = [3.612h, 5.418h] and Y4 = [5.418h, 7.224h]. Where the observed history ends (6.000h), the remainder is extrapolated by the fit weighted by reached machines:

R$/h ≈ -4.69e-4·h + 2.25

On the calendar-time basis, years 3 and 4 follow the Weibull bathtub cumulative hazard H(t) = (t/η)^β, with β fitted on the observed Y2/Y1 cost per machine:

r21 = 53,7% · β = log₂(1 + r21) = 0.620 · Y3 = Y2 × (3^β−2^β)/(2^β−1) = Y2 × 0.818 · Y4 = Y3 × (4^β−3^β)/(3^β−2^β) = Y3 × 0.879

Comparison with the time window: Y2 by time = R$ 1.851 · Y2 by hours = R$ 1.418 (Δ -23,4%). The divergence reflects machines that did not operate full 12 months in Year 2, or operated above the median. The same reading applies to years 3 and 4.

Fleet totals (hours basis): 2 years R$ 12.649.823 · 3 years R$ 12.850.207 · 4 years R$ 12.850.831