Lube Oil Surveillance - Applied AI Foundation

Executive summary

Lube oil is the cheapest insurance policy on a vessel. A single sample tells a Technical Superintendent whether a bearing is wearing, whether seawater is leaking into the cooler, whether fuel is diluting the crankcase, whether the FW generator membrane is failing. Lube oil also moves through tanks: ROB management, bunker planning, and the cost of running out at sea — all driven by the same numbers. The Lube Oil Surveillance pipeline tracks four parallel streams:

Shore analysis results — wear metals, additives, contaminants, TBN, viscosity from the lab
Sampling compliance — per-machinery sample dates, frequency, next-due, overdue flags
Tank ROB & bunker planning — distribution by tank, capacity, recommended intake to reach 85%
Daily consumption — AE crankcase oil from the consumption form, compared against maker limits

Together, the four streams answer the only three questions a TSI cares about: Is the equipment wearing? Are samples on schedule? And do we have enough oil onboard to reach the next bunker?

Architecture

       ┌────────────────────────────────────────────┐
       │  Lab feed · vessel forms · ROB inventory   │
       └─────────────────────┬─────────────────────┘
                             │ samples + tank data
                             ▼
       ┌────────────────────────────────────────────┐
       │   Stage 1 — Collection                       │
       │   Per-vessel snapshot of lab batch +         │
       │   shore schedule + tank ROB + consumption    │
       └─────────────────────┬─────────────────────┘
                             │ structured evidence
                             ▼
       ┌────────────────────────────────────────────┐
       │   Stage 2 — Analysis                          │
       │   • Wear-metal trends                          │
       │   • Contamination flags                        │
       │   • TBN reserve & viscosity drift              │
       │   • Sampling compliance vs schedule            │
       │   • Tank distribution & bunker planning        │
       │   • Daily consumption vs maker limits          │
       └─────────────────────┬─────────────────────┘
                             │ verdict + scores
                             ▼
       ┌────────────────────────────────────────────┐
       │   Stage 3 — Expert review                    │
       │   Auto-escalate to TSI when triggered       │
       └────────────────────────────────────────────┘

Stage 1 — Data collection

Where the data comes from

The lube-oil pipeline pulls data from three independent surfaces and reconciles them:

Source	What it provides
Lab analysis (shore samples)	Wear metals, additives, contaminants, TBN, viscosity per equipment
Vessel-side LO forms	Daily AE crankcase oil consumption + monthly lube-oil ROB
ERP (LO inventory)	Tank distribution, capacities, supplier records

Lab analysis is sourced from any of the major marine lube-oil laboratories the vessel uses:

	Laboratory	Specialty
	Castrol	Lab-mail integration
	Chevron	Lab-mail integration
	Mobil Serv (ExxonMobil)	Lab-mail + API
	Shell	Lab-mail integration
	TotalEnergies	Lab-mail integration
	Gulf	Lab-mail integration
	Maritec	Lab-mail integration
	Marlab	Lab-mail integration
	VPS — Veritas Petroleum Services	Lab-mail integration
	Tribocare	Lab-mail + API
	Viswa Lab	Lab-mail + API

Most labs send results by email; the pipeline parses the lab report (PDF or structured email body) and normalises into the same shape regardless of source. For each lubricated equipment (main engine system oil, cylinder oil, AE crankcase, hydraulics, gearboxes), the collector captures:

Lab analysis

Domain	Parameters
Wear metals	Fe, Cu, Pb, Sn, Cr, Al
Additive metals	Zn, Ca, Mg, P
Contaminants	Water, fuel dilution, soot, silicon
Oil condition	Viscosity at 40 °C and 100 °C, TBN, TAN, oxidation
Sample meta	Sample date, equipment running hours at sample, lab

Sampling schedule

Per machinery: sample frequency, last sample date, next-due date, due-status (Overdue / Due / In Order), test lab, report links, line items per past sample.

Tank ROB

Per tank: grade, tank name, current ROB (litres), 100% capacity, 85% capacity, ullage, calculated bunker intake to reach 85%.

Daily consumption

Per AE per day from the consumption form: crankcase oil consumption (L/day), running hours, comparison vs maker limit.

Stage 2 — Analysis

2.1 Wear-metal trends

For each metal the analyzer computes period-over-period change:

\dot{m} = \frac{m_n - m_{n-1}}{\Delta t}

A step jump where

m_n > 2 \cdot m_{n-1}

is treated as a discrete event, not noise — bearing wear, ring failure, contamination ingress are all step events; gradual rise across three samples is a different signature.

Pattern	Likely cause
Iron rising, gradual	Liner / ring wear
Iron + chromium step jump	Ring failure
Copper rising	Bearing wear
Lead + tin rising	Big-end bearing wear
Aluminium present at all	Piston / piston-pin damage
Silicon above 25 ppm	Dirt / air-filter ingress

2.2 Contamination flags

Each contaminant maps to a specific equipment risk:

Parameter	Threshold	Risk
Water	Above engine spec	Cooler leak, condensation
Fuel dilution	$> 5\%$	Injector leak, fuel-pump seal
Soot	Above engine spec	Combustion incomplete, ring blow-by
Silicon	$> 25$ ppm	Dirt ingress, air-filter failure

2.3 TBN reserve and viscosity drift

TBN reserve is tracked against the fresh-oil baseline:

\text{TBN reserve \%} = \frac{\text{TBN}_\text{current}}{\text{TBN}_\text{new}} \times 100

Tier:

$\geq 75\%$ — Healthy
$50\% \leq \text{TBN reserve} < 75\%$ — Trend down, monitor
$25\% \leq \text{TBN reserve} < 50\%$ — Oil change discussion
$< 25\%$ — Overdue

Viscosity drift outside maker spec at either 40 °C or 100 °C indicates fuel dilution (drop) or oxidation (rise).

2.4 Sampling compliance

Per machinery, the analyzer enforces the sample frequency:

D_\text{next} = D_\text{last} + F \quad \text{where } F \text{ is the configured frequency}

\Delta_\text{days} = D_\text{next} - D_\text{today}

Status mapping:

Verdict	Condition
Overdue	$\Delta_\text{days} \leq 0$
Due Soon	$0 < \Delta_\text{days} \leq 14$
In Order	$\Delta_\text{days} > 14$

The per-machinery view assembles the same shape across every lubricated component. Condensed:

formatted_row = [
    row["machineryName"],
    {"status": row["reportStatus"],
     "color": status_to_color_mappping.get(row["reportStatus"], "grey")},
    row["sampleDate"].strftime("%Y-%m-%d"),
    row["nextDue"].strftime("%Y-%m-%d"),
    row["frequency"],
    {"status": row["dueStatus"],
     "color": status_to_color_mappping.get(row["dueStatus"], "grey")},
    row["testLab"],
    report_links,                     # downloadable lab report
    {"lineitem": lineitem[index]},    # historical sample list
]

A vessel that is consistently overdue on samples will eventually show the wear it failed to detect — sampling gaps are the leading indicator of analytical blindness, not just a paperwork issue.

2.5 Tank distribution & bunker planning

For each tank, the analyzer computes how much to bunker to reach 85% capacity (the safe operating headroom):

\text{Bunker intake} = (0.85 \cdot V_\text{capacity}) - V_\text{current}

where

V_\text{capacity}

is the 100% tank capacity and

V_\text{current}

is the current ROB. Negative intake means the tank is already above 85% — flag for ullage check. The aggregate view collapses intake across all tanks per grade, then surfaces the recommended order quantity per grade for the next bunker call.

2.6 Daily consumption vs maker limit

For AE crankcase oil from the consumption form:

C_\text{daily} = \frac{V_\text{consumed (L)}}{H_\text{running}} \quad \text{[L/h]}

Compared against the maker’s specification limit:

\text{Variance \%} = \frac{C_\text{daily} - C_\text{spec}}{C_\text{spec}} \times 100

Sustained variance above 20% signals piston-ring blow-by or excessive cylinder feed-rate — both are top-end overhaul candidates.

2.7 Equipment-pattern correlation

The analyzer cross-references LO trends with equipment running hours and recent overhauls — wear after a recent overhaul is break-in; sustained wear well into a service interval is accelerated degradation. Cylinder cylinder uniformity findings on the main engine often confirm or rule out a wear-metal trend on the same engine.

Worked example

Vessel: MV POSUN, MAN 6S60ME-C main engine. After a refresh:

Equipment	Sample	Findings
ME system oil	2026-04-12	Iron 42 → 68 ppm (+62% step), silicon 8 ppm — bearing wear suspect
ME cylinder oil	2026-04-08	TBN reserve 48% — trend down; flag for change
AE1 crankcase	2026-04-15	Fuel dilution 3.2% — within range
AE3 crankcase	sample overdue 38 days	analytical blind spot
Hydraulics	2026-03-20	OK

LO tank ROB:

Grade	ROB (L)	85% capacity (L)	Bunker intake (L)
ME system	8,400	12,750	4,350
Cyl oil	18,200	21,250	3,050
AE crankcase	1,250	1,275	25

Verdict: HIGH — ME system oil iron step-jump combined with sample overdue on AE3 means the analyzer is missing data on one engine while another shows accelerating wear. The pipeline:

Tags the run with escalation_required: true
Updates the case to awaiting-tsi-review, priority CRITICAL
Sends an A2A message to the TSI inbox with the wear-metal evidence and the bunker-planning view

Recommended actions:

Investigate ME system oil iron step-jump — pull bearing inspection at next port.
Schedule AE3 LO sample within 48h — restore sampling cadence.
Plan ME cylinder oil change at next port — TBN reserve below 50%.
Bunker request: ME system 4,350 L, cylinder 3,050 L; AE crankcase already at 85%.

Output deliverables

Executive summary — overall LO posture, headline flags
Per-equipment lab table — wear metals, additives, contaminants, TBN, viscosity
Wear-metal trend chart — period-on-period rate of change per metal
Contamination findings — per equipment with likely cause
TBN & viscosity status — reserve %, drift direction
Sampling compliance — per machinery, next due, overdue list, link to lab report
Tank ROB & bunker plan — per grade, recommended intake to reach 85%
Daily consumption — AE crankcase oil vs maker limit
Equipment correlation — recent overhauls, running hours, ME / AE cross-reference
Recommendations — prioritised by impact, with timeline
Escalation decision — auto-routed to TSI when triggered

Escalation triggers

Trigger	Severity
Wear metal above CRITICAL threshold	CRITICAL
Step jump in any wear metal	HIGH
Water above engine spec	HIGH
Fuel dilution above 5%	CRITICAL
TBN reserve below 25%	HIGH
Sample overdue more than 30 days	HIGH
Daily consumption variance above 20% sustained	HIGH
Tank ROB below safety reserve with no stem	CRITICAL

Why script-driven

Wear-metal rate maths, TBN-reserve calculation, sampling-frequency arithmetic, and bunker-intake formula all live in deterministic Python. The reviewer interprets the result; verdicts are reproducible. A wear-metal trend that flips from “monitor” to “escalate” between two samples is a real change in the data, not a model decision.

Sample compliance and lab analysis matter together. A vessel sampling on time but ignoring the trend is not better than a vessel skipping samples — both end with the same engine overhaul. The pipeline reports both streams in the same review so the TSI sees them together.

References

Templates: lube-oil-analysis

Lube-oil-analysis suite — shore-analysis status and tank distribution & bunker planning.

Templates: performance-analysis

Performance-analysis suite — AE LO consumption, shore-analysis data, and shore-analysis scheduling.

Related: ME performance

Wear metals confirm or rule out cylinder findings on the main engine.

Related: AE performance

Crankcase oil consumption is the top-end overhaul leading indicator for AEs.

Documentation Index

​Executive summary

​Architecture

​Stage 1 — Data collection

​Where the data comes from

​Lab analysis

​Sampling schedule

​Tank ROB

​Daily consumption

​Stage 2 — Analysis

​2.1 Wear-metal trends

​2.2 Contamination flags

​2.3 TBN reserve and viscosity drift

​2.4 Sampling compliance

​2.5 Tank distribution & bunker planning

​2.6 Daily consumption vs maker limit

​2.7 Equipment-pattern correlation

​Worked example

​Output deliverables

​Escalation triggers

​Why script-driven

​References

Templates: lube-oil-analysis

Templates: performance-analysis

Related: ME performance

Related: AE performance

Executive summary

Architecture

Stage 1 — Data collection

Where the data comes from

Lab analysis

Sampling schedule

Tank ROB

Daily consumption

Stage 2 — Analysis

2.1 Wear-metal trends

2.2 Contamination flags

2.3 TBN reserve and viscosity drift

2.4 Sampling compliance

2.5 Tank distribution & bunker planning

2.6 Daily consumption vs maker limit

2.7 Equipment-pattern correlation

Worked example

Output deliverables

Escalation triggers

Why script-driven

References