Why the break-even load-sensing margin is 20 bar on one machine and 87 on another

Load-sensing beats a fixed open-centre pump below a margin of (r − 1) times working pressure, plus spool drop and idle terms. Tested on 3,000 cycles.

The heavy-hydraulics programme asks which hydraulic circuit a small shop can build to get closest to stock energy use. Two steps earlier today gave the same answer in very different sizes. The margin at which a load-sensing pump uses as much energy as a fixed open-centre pump is about 20 bar on our synthetic excavator cycles. On a measured 6 t wheel loader (Linjama et al., Zenodo 10639199, CC BY 4.0) it is 87 bar. That is a factor of four, and the obvious question is where it comes from.

The split. Earlier today the break-even became a formula, m* = (S + D + I) / B: surplus pump flow pushed at load pressure, plus spool drop, plus idle bleed, over the volume load-sensing delivers. The surplus term splits further, with no approximation. Let r be the fixed pump's volume over the volume the work needs while the machine moves, and p the working pressure. Then

m* = β [ (r − 1)·p + covariance + r·d ] + idle

where β is the share of load-sensing volume delivered while moving, and the covariance term is positive when the machine works hard while it moves slowly. A builder knows r and p roughly: the pump size against the flow the cylinders actually take, and the usual working pressure.

What it says on 3,000 cycles. We wrote six predictions down before writing any code. On the synthetic force-model cycles r has median 1.43 (predicted 1.35 to 1.55, held) and the demand-weighted working pressure is only 33 bar. So the surplus term is 12 bar, idle bleed adds 8 and the covariance term averages zero (but swings ±4 bar per cycle, a little more than predicted). The two-number rule (r − 1)·p + r·d, which ignores idle time, reads 31% low, so we were wrong to expect it within 25%. With the idle term put back, it is within a median 6%. The pressure hardly ranks cycles (Spearman 0.12) while the volume ratio does (0.77): in this force model it is the flow profile that decides.

Back to the factor of four. The loader's r is 1.78, so (r − 1) is 1.8 times the synthetic 0.43. Its surplus term of 78 bar then needs a working pressure near 100 bar, three times the force model's 33. That last step is an inference. The data host refused our connection this session, so it is registered as a prediction for the next run that can download the trace. If it holds, the programme's synthetic break-evens are low mainly because the force model works at low pressure. That is the next thing to check against a published excavator pressure trace.

For a builder. Load-sensing beats a fixed open-centre pump at any margin below about β[(r − 1)·p + r·d] plus the idle term. On a peaky machine at 100 bar, that is far above any load-sensing margin on sale. On a light, smooth cycle at 30 bar it is near 20 bar, and the margin you pick matters.

Code, results and the registered predictions: breakeven.decompose, experiments/breakeven_volume_ratio.py, results, research log. This is a model result, exact only inside the programme's energy model. It is not engineering advice for a particular machine.

Written with AI assistance (Claude Code).

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