Aggregated windows

Latest cycle

Time to damage

Counted from the last pump start, assuming the most recent leakage rate holds and that 150,0 m³ can accumulate before the flooding damage level is reached. Counting from the start rather than the stop ignores the drawdown during that run, so the figure errs on the pessimistic side.

Pit geometry & pump delivery

Everything here is anchored on the pit itself. The pump starts at EL 654,6 and stops at EL 651,1 — 3,5 m over 7,04 m², or 24,6 m³ nominal. No pump rating enters the leakage figures at all, which is what stops the calculation feeding on its own assumptions.

Delivery is derived, not assumed: a complete cycle moves exactly one drawdown, so it is (cycles × drawdown) ÷ total run time, plus what leaks in while the pump runs.

The effective drawdown is fitted to a measured 2,42 L/s leakage rate taken on 2026-09-01, against the standing times of the same day. Nominal geometry alone put it 24,6 m³, which forces a delivery well above the plate rating — pipework and supports occupy pit volume, and the floats need not swing the full switch separation, so the water actually moved is less than plan area times setpoint difference. Re-fit it if the floats are reset or the pit is cleared.

Run time carries all of the timing error, so delivery is the least certain figure on this page. Leakage never divides by it, which is why its band collapses to almost nothing — but that band covers timing only. It does not carry any tolerance on the pit dimensions or the switch levels, and those now dominate: a few per cent out on either moves every volume on this page by the same few per cent. Re-measure them if the float settings are changed.

Derived flow

Volumes come from the pit drawdown and the cycle count, so they carry no pump-rate assumption. Pump time per day is measured directly from the log.

How this is worked out

The PLC rewrites DrainPumpCycle on every drain-pump start with the number of seconds the pump stood idle beforehand, so a change in that value marks a start and the value itself is an exact stoppage measurement.

Start times, though, are only resolved to the 2m 30s log interval. Run time is therefore taken as span less summed stoppage across a whole window rather than measured cycle by cycle, which spreads a fixed error of roughly one log interval over the full span instead of landing it on each short pump run. Until complete cycles cover 75% of a window that error swamps the result, so no figure is published.

Two back-to-back cycles that report identical idle times are indistinguishable from one long gap. Any interval lasting at least twice the idle time it reports could hide such a start, so it is dropped from both span and stoppage rather than guessed at — those are the "Gaps" column above.

The latest-cycle figure divides the calibrated pit drawdown by the exact standing time. It carries no sampling error of its own, so it responds cycle by cycle, but it is only as good as the drawdown volume the aggregated windows calibrate.