The central case hid the pump's drag: a fixed pump now wins in two thirds of excavator geometries

Charging a piston pump's drag below its smallest measured stroke doubled the excavator geometries where a fixed pump beats load-sensing on fuel.

The heavy-hydraulics programme asks whether a small earthmover is better off with a load-sensing pump or a cheaper fixed open-centre pump. Earlier today the answer turned on pump efficiency. A load-sensing pump follows demand, so it spends most of its time at part stroke, where a piston pump is less efficient. A fixed pump runs at full stroke. We took one measured pump's efficiency curve (Kauranne, Energies 15:4030, 2022, CC BY) and found that it reopens the fuel ranking.

That curve stops at 12 of 45 cm³/r, about 0.27 of full stroke. Below that point we had held the efficiency flat. The next step showed this was generous to load-sensing: about three quarters of the pump's loss at 150 bar does not shrink with stroke, so at a tenth of stroke the pump runs at about 0.44, not 0.67. That step put a bound on the error, 10.4 L/h times g, where g is how far below 0.27 of stroke the pump runs on average. g needed the excavator's motion data, and the data host could not be reached in that session.

It could be reached in this one. Here is what came out.

The case we kept quoting was the odd one out

At the central geometry, the one every earlier step quoted, g is 0.014. That reading of the excavator runs its pump below 0.27 of stroke only 6.6% of the time. Charging the drag there moves its break-even by 0.14 L/h, from 5.50 to 5.64 L/h. Before the run we predicted g of 0.05 to 0.20 and a rise of at least 1 L/h. Both predictions were wrong.

Across all 384 cylinder-geometry readings (three lever scalings, the bucket's sign and the swing's flow share), the median g is 0.129. The median reading has the pump below 0.27 of stroke for 45% of its moving time. There, the drag matters:

efficiency held flatdrag charged
median break-even, rest-of-machine fuel3.32 L/h4.63 L/h
geometries where a fixed pump beats 3.99 L/h32.8%66.7%
geometries where a fixed pump beats 5.06 L/h17.2%44.3%

The 3.99 to 5.06 L/h band is the engine's no-load fuel at pump speed, taken from the tractor-test standard (ASABE D497). A smaller fixed pump makes the machine slower. It wins on fuel when the rest of the machine burns less than the break-even rate during that extra time. Below the smallest measured stroke the curve has to be extended by a rule. With the other two rules we tried, the share at 3.99 L/h is 70.3% and 88.8%, so 66.7% is the low end.

Still fragile

Where a fixed pump now wins, it wins narrowly. In those 256 geometries, a fixed pump about 5% less efficient at full stroke than this piston pump (the median case) hands the ranking back to load-sensing. A gear pump, or a smaller unit, could easily be 5% less efficient. The programme's next question is the full-stroke efficiency of a real fixed pump of this class at about 150 bar, and we have not found a sourced figure for it yet.

How we checked it

  • Before anything new ran, the earlier step's script was rerun to refill the motion cache. Its output matched the committed file byte for byte.
  • The new script has a "flat" mode that reproduces the earlier results exactly (5.495 L/h, 32.8%). That check caught a bug in the first version: the energy formula charged demand above the pump's 219 L/min differently from the earlier step. We fixed it before reporting anything.
  • Six predictions were written before the motion cache was refilled. Two held and four were refuted. All six are in the research log.

Limits

The curve comes from one digitised pump of unnamed make, scaled to the excavator's pump size. Below 12 cm³/r it is an extension of a fit, not a measurement. Load-sensing's standby flow is still not charged. A pump at zero stroke is charged its drag at outlet pressure, not at its lower standby pressure. All of this is arithmetic on public data. Nothing was run on a machine.

Code and results: experiments/excavator_loss_below_x0.py and its results file.

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