With a measured gear pump, the small-excavator pump question is not close

Rerun with a bench-tested gear pump, a smaller fixed pump beats load-sensing on a small excavator's fuel in all 384 geometries and all 18 idle settings.

The heavy-hydraulics programme asks whether a small excavator burns less fuel with a stock load-sensing pump or with a smaller fixed gear pump on an open-centre valve. Earlier steps narrowed the answer to one number: the fixed gear pump's overall efficiency at about 150 bar. Above about 0.83 the smaller fixed pump wins in more than half of 384 readings of the machine's cylinder geometry. Below it, load-sensing wins.

The last step found a bench test of a pump of the right size, a Casappa KP30 of 44 cm³/r: 0.913 at 150 bar, against 0.764 from Casappa's own catalogue curves. It only bounded the excavator result, though. It said "at least 94.8% of geometries", taken from an older sweep, and it said nothing about the bench pump's drag when the machine sits idle.

This step reran the full calculation with the bench pump. The bench data became a pump family. Each catalogue size gets the measured torque at 50, 150 and 250 bar, scaled by displacement, and the measured volumetric efficiency. The same code that read the catalogues reads it. The fuel model is unchanged: the 7.5 t excavator's measured joint motion, 384 cylinder-geometry readings, and idle at 38% of engine hours.

fixed pumpefficiency at 150 barsmaller fixed pump wins inbreak-even fuel rate, central
Casappa catalogue curves0.7649.9% of geometries1.88 L/h
bench-tested KP300.913100%8.82 L/h

The break-even fuel rate is how much the rest of the machine would have to burn per hour of extra working time before load-sensing caught up. The tractor test standard puts a comparable engine's no-load rate at 3.99 to 5.06 L/h. With the bench pump the median break-even is 8.1 L/h, so the fixed pump wins with room to spare. It wins in every one of 18 idle settings too (three valve neutral curves, three standby flows, two idle speeds).

Which half matters. Swapping the two halves of the pump model shows what does the work. The bench pump's moving efficiency with the catalogue's idle drag still wins everywhere. The catalogue's moving efficiency with the bench's lower idle drag wins in only 13%. The answer is carried by how efficiently the pump works under load, not by what it costs at idle.

A side check. A finer scan of Casappa's catalogue curves, scaled up in small steps, crosses the 50% line at 0.8283. The earlier coarse grid gave 0.8281. So the threshold was not an interpolation artefact.

What the predictions say. Seven predictions were written before the sweep ran, and all seven held. That sounds better than it is: each bar was set below where the earlier bound already pointed. The useful information is the size of the margins.

Limits. This is still one pump unit, measured on its maker's bench and published in a CFD paper, not an independent ISO 4409 test report. Other sizes are scaled from it. That matters little, because the best fixed pump comes out at 85 to 108 L/min and the tested unit delivers about 92 L/min at the working speed. The excavator answer flips back only if typical units of this class run about 0.08 below this one. That is roughly the gap between this unit and its own catalogue, so the question left is how much pumps vary from unit to unit.

Code and results: excavator_gear_pump_bench.py, excavator_gear_pump_bench_results.txt, and the research log.

All writing · More in Research findings · All tools and research