Autonomous driving paper index
Auditable Clean-in-Place Decision Support from Routine SWRO SCADA: Selecting Differential-Pressure Recovery and Falsifying a Per-CIP-Reset Trigger
One-line summary
We present an auditable workflow that converts these thresholds into plant-calibrated decision support.
Engineering notes
Key topics: autonomous driving. See the paper for implementation details and experimental results.
Chinese explanation / 中文解读
中文解读待补充:本站会优先为端到端自动驾驶、BEV感知、3D目标检测、轨迹预测、路径规划、LiDAR感知等高价值论文补充中文说明。
Original abstract
Reverse osmosis (RO) desalination operators time membrane clean-in-place (CIP) by non-site-calibrated vendor thresholds. We present an auditable workflow that converts these thresholds into plant-calibrated decision support. It pre-specifies the recovery target and analysis unit, compares candidate signals with a five-test label-free battery, and reports composite weights only when identifiable. Using ≈two years of routine 10 min SCADA from one three-train island seawater RO plant (≈600 m3 d−1), the battery selects normalised feed channel differential pressure (DP_norm) for site-specific cleaning-recovery review. On the false discovery rate (FDR)-effective unit (n = 14 campaigns), recovery is marginal and not FDR-significant (mid-p Benjamini–Hochberg q ≈ 0.141); DP_norm is therefore an operator-review signal, not an autonomous or FDR-confirmed trigger. The same battery invalidates a per-CIP-reset net driving pressure trigger as a clip-floor regression-to-the-mean artefact; it collapses under the pre-specified matched placebo and is reproduced by a fake-date null, a finding that is bounded to this plant and estimand, and not causal proof of a unique mechanism. The DP_norm of ≥1.20 review point is an exploratory, in-sample heuristic selected partly against the endogenous operator log, and its indexed economic comparison inherits that limitation. An identifiability-gated Bayesian power prior leaves the four-weight composite non-identifiable on this single-regime plant, pinning only the near-zero salt passage weight. The contribution is a bounded, estimand-based workflow that supports audit without replacing operators or the safety envelope.
Links and sources
Need this topic turned into a technical roadmap?
Full Self Driving can prepare a custom autonomous driving literature review, code map, dataset map, and B2B technology assessment.
Request B2B research
Comments