CRC 2022 experiment — raw & validated measurements

On 2022-07-11 the o-t-b pilot rig ran two back-to-back experiments on the same three-pump loop: a conventional cascade sweep and a CRC (Carbon Reducing Controller) run that redistributes control across all machines toward their combined best-efficiency point. Both were logged at ~1 Hz. This page publishes the recovered measurement data behind the 2022 white paper^[1] — the raw per-machine time series and the validated control-axis series — clearly separated. Across the measured total-flow range the CRC run held an average efficiency advantage of +26.0 % over cascade (flow-per-watt proxy), peaking at +97.2 % at the 30 l/min bin and dipping to −7.2 % only at a low-flow range edge that is within measurement error.

+26.0 % average advantage for CRC over cascade across 14 total-flow bins (0–75 l/min).
0.859 vs 0.544 (l/min)/W — CRC vs cascade average per-machine efficiency (ratio 1.58×).
Limitation — no pressure was logged, so “efficiency” is a within-rig flow-per-watt proxy, not hydraulic efficiency.

Rig & method

Both runs used the 2022 pilot rig: 3× DAB CPAE-80-15-S-PWM circulation pumps on a shared 70 L basin, each metered by a Grundfos VFS-2-40 flow sensor (total loop by a VFS-5-100) and an Eastron SDM120 power meter, with a RevPi Core/AIO/DIO as logger and controller. The cascade run lasted ~19 min (~1104 samples/machine); the CRC run ~17 min (~1018 samples/machine); both were sampled at ~1 Hz on 2022-07-11.

limitation No differential pressure was logged into these series (the rig's pressure sensor was not recorded). Every “efficiency” figure on this page is therefore flow per electrical watt, (l/min)/W — a proxy that is only valid as a comparator between the two control strategies on this same rig and system curve, not as an absolute hydraulic efficiency. The two runs share the same pumps, basin and pressure regime, which is what makes the comparison fair.

Raw measurements time axis

Below are the two runs exactly as measured: per-machine flow (left axis, solid) and electrical power (right axis, dashed), plotted against seconds from the start of each run. All three machines are shown; a machine sitting at zero flow/power is switched off (state = 0) at that moment. Nothing here is smoothed or collapsed — this is the raw, as-measured telemetry.

Cascade run (2022-07-11), raw per-machine flow and power vs time. The cascade strategy ramps one machine at a time: m0 carries the low-demand range, then m1 and m2 switch in as demand rises, which is why m1/m2 sit at zero early. Flow in l/min (left), power in W (right, dashed). Raw, as measured; full series embedded (no downsampling).

CRC run (2022-07-11), raw per-machine flow and power vs time. CRC distributes the demand across all three machines simultaneously toward their combined best-efficiency point, so all three carry flow together rather than one-at-a-time. Flow in l/min (left), power in W (right, dashed). Raw, as measured; full series embedded (no downsampling).

Full raw series — row counts & provenance

Each raw sheet has the schema id; ctrl; flow; power; state; timestamp (flow in l/min, power in W, state 0/1, timestamp epoch-ms). The complete per-machine series are embedded inline in this page as JS arrays (view-source), not downsampled.

run	machine	samples	start epoch (ms)

Validated measurements control axis

“Validated” here means the raw time series were collapsed onto the accumulated control axis: for each control setpoint the rig reached, the steady-state flow, power and per-machine efficiency were taken, giving one point per accumulated-control % rather than one per second. This transformation is part of the original 2022 analysis (Pilot_pump_setup.xlsx), not a 2026 reinterpretation — the figures below are the recovered basis of white-paper Figs 10–12^[1]. The cascade series is shown both as measured and as the derived cascade_without_step variant with the machine-switch step removed. Efficiency is the flow-per-watt proxy defined above.

Validated total flow vs accumulated control %. Both strategies deliver essentially the same flow-vs-control relationship (as they must, on one rig) — the difference is in what it costs, shown next.

Validated total electrical power vs accumulated control %. For a comparable flow demand the CRC run generally draws less total power than cascade over the mid-range, which is the source of the efficiency advantage.

Validated total efficiency (flow-per-watt proxy, (l/min)/W) vs accumulated control % — the recovered basis of white-paper Figs 10–12. Three traces: cascade (as measured), cascade without the machine-switch step, and CRC. CRC holds a higher flow-per-watt across most of the working range.

Overview — CRC vs cascade by flow bin

The headline comparison bins average measured efficiency by total flow (0–75 l/min) and contrasts the two strategies. CRC leads in 11 of the 14 populated bins; the ratio (CRC/cascade − 1) is plotted on the right axis. The three negative bins sit at the low- and high-flow range edges and are within measurement error (see notes).

Average measured efficiency per total-flow bin: CRC vs cascade (bars, left axis, (l/min)/W) with the CRC/cascade advantage ratio (line, right axis). Best bin +97.2 % at 30 l/min; worst −7.2 % at 16 l/min. Source: Overview cascade vs crc sheet.

Recovered headline numbers, with values recomputed from the source sheets during this build vs the numbers stored in the 2022 Overview sheet. They agree.

metric	recomputed	sheet	source

Per-bin overview values

total flow (l/min)	CRC (l/min)/W	cascade (l/min)/W	CRC/cascade − 1

Honest notes on the data

note “1952” readable-time slip. The validated sheets carry a human-readable time column showing 1952 dates — an epoch-unit conversion mistake in the original 2022 Excel. It is cosmetic: the raw timestamp epoch-ms values are correct and decode to 2022-07-11. All time axes on this page are rebuilt from the raw epochs, so they are unaffected.

note No averaging window per measurement. The white paper's own caveat: each binned measurement is a single steady-state read, with no settling/averaging window per setpoint. This is why the few negative bins appear at the range edges — they are within measurement error and occur for both algorithms, not evidence that cascade wins there.

note No ΔP. As stated in the method, no differential pressure was logged, so no true hydraulic efficiency can be computed from these series — only the flow-per-watt proxy.

note Raw sheet column variance. The m0 raw sheets carry extra helper columns (a spare Column7 plus a stray header-echo cell), while the m1/m2 raw sheets have the clean 7-column layout. The five data columns (ctrl, flow, power, state, timestamp) are identical in meaning across all six; only the trailing helper columns differ, and they are ignored here.

Sources

CRC — “Carbon Reducing Controller”, R. de Ren, white paper v3 (2022, unpublished), archived on this track. archival copy ↗ archived 2026-07-17
Pilot_pump_setup.xlsx — original 2022 measurement & analysis workbook (raw cascade_m*/crc_m*, *_validated_data, cascade_without_step, Overview cascade vs crc). recovered from author's archive 2026-07-17

Data vintage experiment 2022-07-11; recovered from the author's archive 2026-07-17.

Method raw .xlsx sheets extracted to semicolon-CSV (python-docx / openpyxl); CSVs parsed and validated with a Node build script; all figures plotted from the inline series.

Verification the +26.0 % average, +97.2 %/−7.2 % extremes and the 0.859/0.544 per-machine averages were recomputed from the source sheets and match the stored 2022 values (see Table 1).

Tools Node (CSV parse) · Plotly (helix-plotly.js) · data embedded inline (view-source).

Contact R&D-lab · lab.wbd-rd.nl