Water Treatment Surveillance - Applied AI Foundation

Steel and water are not a stable combination. The water on a vessel is a managed chemistry experiment that runs continuously: boiler water has to stay alkaline enough to prevent corrosion but not so alkaline it deposits scale; cooling water has to stay treated against bacteria; potable water has to stay safe to drink; the FWG output has to stay within salinity limits or the heat exchangers downstream will pit. The Water Treatment Surveillance pipeline reads the boiler and cooling water test forms, the chemical inventory, the noon-report water-test entries, and the FWG telemetry — and produces a parameter-by-parameter view of the four water systems on board, plus the chemical-stock view that tells a Technical Superintendent whether the vessel can continue treatment without re-supply.

Where the data comes from

Source	What it provides
Vessel-side test forms	Boiler / cooling / potable / FWG chemistry readings filed by Chief Engineer
Chemical inventory register (ERP)	Per-chemical ROB, supplier, daily dosing
Noon report (water tests)	FWG output salinity, additional spot readings
FWG telemetry	Continuous salinity / conductivity logging where instrumented

Chemicals are sourced from any of the major marine water-treatment suppliers:

	Supplier	Specialty
	Drew Marine	Boiler, cooling, fuel-treatment chemicals
	Wilhelmsen / Unitor	Boiler, cooling, fuel-treatment, cleaning chemicals
	Vecom	Boiler and cooling water treatment
	Chevron Marine	Boiler and cooling water treatment

The pipeline reads chemical-supplier acknowledgements (delivery + spec sheet) and reconciles them against the inventory.

The four water systems

System	Critical parameters	Failure mode
Boiler water	pH, conductivity, hardness, phosphate, chloride	Tube corrosion, scale, foaming
Cooling water	pH, conductivity, nitrite, chloride, microbial	General corrosion, biofilm, seawater ingress
Potable water	Chlorine, hardness, microbial, salinity	Health hazard, scale, taste
FWG output	Salinity, conductivity	Heat exchanger pitting downstream

Each system has its own spec range, its own dosing protocol, and its own consumption profile. The pipeline tracks each independently then integrates the findings into the senior review.

Parameter classification

For each parameter the analyzer compares the latest reading against the maker / equipment-specific spec. Three tiers:

Tier	Condition
OK	Within spec, trend stable
WARNING	Outside spec but within wider safety envelope
CRITICAL	Outside safety envelope, or step jump from previous reading

The wider safety envelope matters because spec ranges are usually conservative. A boiler at slightly elevated chloride is something to watch; a boiler at chloride three times the spec is a corrosion event in progress.

Equipment-damage risk mapping

Specific patterns map to specific risks:

Boiler chloride above limit

Saltwater contamination of boiler feedwater. Causes pitting corrosion of tubes; in severe cases leads to tube failure and boiler shutdown. Source is usually a feedwater contamination upstream — condenser leak, distillate carryover, or salt-water bunker fill error.

Cooling chloride above limit

Seawater ingress into a closed cooling-water circuit. Cooler shell-side leak is the typical source. Even small ingress drives general corrosion across the entire circuit.

Cooling pH falling

The closed circuit’s buffering capacity is being exhausted. Either the dosing is insufficient or there’s a leak admitting fresh water. Drives general corrosion across all wetted metals.

Microbial counts rising

Biofilm developing in the cooling circuit. Biofilms shield the metal under them from cathodic protection and corrosion inhibitor — pitting starts under the film. Treatment is biocide dosing plus circuit cleaning.

FWG salinity rising trend

Either the FWG demister is failing (allowing salt carryover) or the seawater feed is more saline than the unit can handle. Membrane FWGs are particularly sensitive — once salinity rises, the heat exchanger downstream starts pitting.

Trend assessment — direction matters

A three-sample trend per parameter classifies systems as Improving / Stable / Deteriorating. Direction matters as much as magnitude:

\text{slope} = \frac{x_n - x_{n-2}}{2 \cdot \Delta t}

For an upper-bounded parameter (chloride, conductivity, microbial count), positive slope is Deteriorating regardless of whether the current value is in spec — a vessel whose chloride has doubled from sample to sample is heading somewhere bad even if the latest reading is technically OK. For a lower-bounded parameter (pH, nitrite), positive slope is Improving. The pipeline encodes the bound direction per parameter so the verdict makes sense:

PARAMETER_BOUNDS = {
    "boiler_chloride":   "upper",
    "boiler_pH":         "lower",
    "boiler_phosphate":  "range",
    "cooling_chloride":  "upper",
    "cooling_pH":        "lower",
    "cooling_nitrite":   "lower",
    "cooling_microbial": "upper",
    "fwg_salinity":      "upper",
    # …
}

def trend_verdict(samples, bound_direction):
    if len(samples) < 3:
        return "Insufficient data"
    slope = (samples[-1] - samples[-3]) / (2 * SAMPLE_INTERVAL_DAYS)
    if abs(slope) < NOISE_THRESHOLD:
        return "Stable"
    if (slope > 0 and bound_direction == "upper") or \
       (slope < 0 and bound_direction == "lower"):
        return "Deteriorating"
    return "Improving"

Chemical inventory and consumption

The chemical-inventory view focuses on the supply side: per-chemical ROB, supplier, and the daily consumption rate computed from period-on-period stock changes:

\dot{C}_\text{daily} = \frac{ROB_{n-1} - ROB_n}{\Delta t}

Days remaining at current burn:

\text{Days remaining} = \frac{ROB_n}{\dot{C}_\text{daily}}

Compared against the next bunker call, the analyzer flags chemicals running short before resupply is feasible:

Days remaining	Tier
$> 60$	OK
$30$ to $60$	Plan re-order
$< 30$	Re-order urgent
$< 7$	Critical — operational risk

A vessel that runs out of boiler treatment chemical at sea is forced to either secure the boiler or run untreated water — neither is acceptable as a planned outcome.

A sample sweep on `MV POSUN`

End-of-April water-treatment review:

System	Parameter	Latest	Spec	Trend	Verdict
Boiler	pH	10.8	10.5–11.5	Stable	OK
Boiler	Conductivity	1,650 µS	$\leq 2,500$	Rising	Watch
Boiler	Chloride	42 ppm	$\leq 25$	Step jump from 18 ppm	CRITICAL
Boiler	Phosphate	28 ppm	20–40	Stable	OK
Cooling	pH	8.6	8.5–9.5	Falling	WARNING
Cooling	Nitrite	720 ppm	$\geq 500$	Falling	Watch
Cooling	Chloride	45 ppm	$\leq 60$	Stable	OK
Cooling	Microbial	320 cfu/mL	$\leq 1,000$	Rising	Watch
FWG	Salinity	3.2 ppm	$\leq 5$	Rising	Watch

Chemical inventory:

Chemical	ROB (kg)	Daily use	Days remaining
Boiler treatment (Drew Liquitreat)	28	1.2	23
Cooling inhibitor	84	1.8	47
Biocide	12	0.4	30
FWG cleaner	45	1.1	41

Verdict: HIGH.

Boiler chloride step jump (18 → 42 ppm) is a feedwater contamination event — most likely a condenser tube leak. Investigation required immediately.
Cooling pH falling + nitrite falling: dosing rate insufficient or fresh-water leak. Cross-check the dosing pump and the FW expansion-tank level.
Boiler treatment ROB 23 days: re-order urgent; next bunker is in 14 days but supplier lead-time is 21.

The pipeline:

Flags the boiler chloride event to the Technical Superintendent — condenser inspection at next port.
Generates the cooling-water dosing review for the chief engineer.
Routes the re-order to procurement with the chemical-supply chain lead-time visible.

What the senior review contains

Per-system parameter table — current value, spec, trend, verdict, with damage-risk note for any out-of-range parameter.
Out-of-range findings — the focused list with associated equipment risk.
Trend chart — multi-period view per system, with direction-aware classification.
Chemical inventory — per chemical, ROB, days remaining, re-order status.
Dosing review — dosing pump performance, dosage-rate vs target.
Recommendations — dosing changes, equipment inspection, sample re-runs.
Escalation decision — to whom, and why.

Escalation triggers

Trigger	Severity
Boiler chloride above critical limit	CRITICAL
Boiler chloride step jump from previous sample	CRITICAL
Cooling chloride above critical limit	CRITICAL
FWG salinity rising trend with current value $> 7$ ppm	HIGH
Microbial count above safety limit	HIGH
Critical chemical with no re-supply in lead-time window	HIGH

Why it’s mostly trend-driven

Spot readings on a water system are noisy — pH meters drift, sample bottles get contaminated, the engineer takes the sample at the wrong time. Trends are robust: a pH that’s been falling for three samples in a row is real even if any single reading might be wrong. The pipeline weights the trend more heavily than the spot value for exactly that reason.

The single most useful improvement in water-treatment data quality is calibrating the test kits monthly. A vessel with un-calibrated kits produces readings that look like rapid deterioration when actually nothing is happening — and the pipeline’s trend logic dutifully escalates the noise.

References

Templates: bwt-management

BWT-management template — boiler and cooling water chemistry from vessel-side test forms, with chemical ROB tracking across suppliers (Drew Marine, Unitor, and others).

Related: Lube oil

Lube oil chemistry has its own pipeline, but cooler leaks show up in both.

Related: Forms

Boiler and cooling water test forms are the primary source — late submissions break this pipeline.

Related: Procurement

Chemical re-supply runs through procurement; lead-time gaps surface in both pipelines.

Documentation Index

​Where the data comes from

​The four water systems

​Parameter classification

​Equipment-damage risk mapping

​Trend assessment — direction matters

​Chemical inventory and consumption

​A sample sweep on MV POSUN

​What the senior review contains

​Escalation triggers

​Why it’s mostly trend-driven

​References

Templates: bwt-management

Related: Lube oil

Related: Forms

Related: Procurement

Where the data comes from

The four water systems

Parameter classification

Equipment-damage risk mapping

Trend assessment — direction matters

Chemical inventory and consumption

A sample sweep on `MV POSUN`

What the senior review contains

Escalation triggers

Why it’s mostly trend-driven

References