Protect the water
that reaches the tank.

A specific difference to investigate.

A sensor reading is taken at one place and one time. Water continues to move through tubing, valves and holding volumes. A useful controller has to consider what will actually arrive in the accepted tank.

The candidate adds a transport and inventory rule. A fixed delay, a larger tank or an automatic dump valve is not the claimed advance.

Start with what already exists.

Early wetting detection and automatic diversion already exist. Conductive-spacer sensing, salinometers, threshold alarms and persistence rules are meaningful comparisons.

The candidate must beat a tuned early-warning controller with sensible hysteresis, persistence, fault handling and the same sensors, tank, valves and permitted actions.

Conductive-spacer study · In-situ wetting-detection patent · Established freshwater-generator controls

Local imaging is useful supporting evidence but cannot substitute for a fast, independent delivered-water reference. The source audit also retains the microampere-versus-milliampere discrepancy between the patent PDF and searchable HTML.

Make the comparison decisive.

First measure sensor response, the full residence-time distribution, mixing, valve delay and tank carryover. Freeze the transport rule before using later feed, membrane lots and held-out challenge trajectories.

Compare three policies on matched hardware: conductivity-triggered routing, tuned early warning plus conductivity, and the proposed transport-aware policy. Give both conventional policies the same development access.

The plan contains 12 commissioning runs and 108 confirmation runs across gradual wetting, abrupt failure, benign variation, sensor faults, oil fouling and combined oil/surfactant exposure. Commissioning capacity may need revision if the delay model cannot be identified.

Read the full Study F protocol and measurement requirements.

Know when to stop.

• The independent reference cannot resolve the transient delivered-water endpoint.
• The competent baseline already performs equivalently.
• Uncertainty forces so much diversion that useful water service is worse.
• The rule depends on a favorable trajectory that fails on held-out challenges.

If the baseline already accepts effectively no bad water, a lower failure count has no meaningful room to improve. A water-retention comparison needs its own prespecified quality margin.

The next useful step.

Find an equipped membrane-distillation host with a real water-protection problem and a qualified independent outlet reference. This study can come first when that host is available before a suitable diagnostic-study partner.

The apparatus and controller are proposed. No field protection, drinking-water certification or operational deployment is established.