Auxiliary Engine Performance - Applied AI Foundation

Executive summary

Auxiliary engines do not power the vessel — they power everything else on it. A single AE failure with no standby is a blackout. A blackout in port is a delay; a blackout at sea is a casualty. The job of a Technical Superintendent reviewing AEs is therefore not just “is the engine healthy” but “is the fleet of generators healthy enough to ride through any one of them failing”. The Auxiliary Engine Performance pipeline answers both questions on a rolling basis. It captures every monitored AE parameter from the ERP’s monthly performance reports, evaluates engine availability and load-sharing posture, scores blackout risk, and produces a senior review whenever a vessel’s redundancy degrades or an engine drifts toward overhaul.

Architecture

       ┌─────────────────────────────────────────┐
       │  ERP telemetry  ·  Engine register  ·  PMS  │
       └─────────────────────┬───────────────────┘
                             │ monthly reports + specs
                             ▼
       ┌─────────────────────────────────────────┐
       │   Stage 1 — Collection                   │
       │   Per-engine telemetry, hours, alarms   │
       └─────────────────────┬───────────────────┘
                             │ structured telemetry
                             ▼
       ┌─────────────────────────────────────────┐
       │   Stage 2 — Analysis                     │
       │   • Engine availability & redundancy    │
       │   • Load-sharing posture                 │
       │   • SFOC variance per engine             │
       │   • Six-month parameter rollup           │
       │   • Overhaul timing                      │
       │   • Blackout-risk score                  │
       └─────────────────────┬───────────────────┘
                             │ verdict + scores
                             ▼
       ┌─────────────────────────────────────────┐
       │   Stage 3 — Expert review                │
       │   Auto-escalate to TSI when triggered   │
       └─────────────────────────────────────────┘

Stage 1 — Data collection

Where the data comes from

Source	What it provides
Vessel ERP	AE register, running hours, status (running / standby / fault), alarms history
Forms (AE Performance)	Monthly per-engine data filed by Chief Engineer
PMS counter feed	Hours since overhaul per engine — drives overhaul timing

The AE collector pulls the latest performance entry per engine from the telemetry feed. The same six-month rolling window used for ME reports applies — every monthly performance report, parameter by parameter, per AE unit.

What gets captured

Domain	Data
Engine state	Status (running / standby / fault / maintenance), running hours, hours since overhaul
Load	Load percentage, power output, exporting / importing
Per-cylinder	Pmax, exhaust temperature, FW inlet/outlet temp
Aspiration	Boost pressure, exhaust back-pressure
Fuel	SFOC actual, fuel grade in use
Power management	Number of engines online, blackout-risk inputs
Alarms	Active alarms with severity at sample time

Output

A single timestamped record per vessel covering every AE plus the power-management context.

Stage 2 — Analysis

2.1 Engine availability and redundancy

Each AE is bucketed by status. The fleet-of-generators view is the availability tier:

Running	Standby	Tier	Risk
1	2+	Adequate redundancy	LOW
1	1	Minimum redundancy	MEDIUM
1	0	No redundancy	HIGH
2+	0	Load sharing, no backup	MEDIUM
0	any	Blackout — terminal	CRITICAL

Utilization is tracked too: if the same engine always runs, wear is uneven; if standby rotation is happening, the fleet stays balanced. Standby rotation cadence is a leading indicator of how the engineers actually operate the plant.

2.2 SFOC variance

Per-engine SFOC variance follows the same definition as the main engine:

\text{Variance \%} = \frac{\text{SFOC}_\text{actual} - \text{SFOC}_\text{design}(P)}{\text{SFOC}_\text{design}(P)} \times 100

Sustained variance above 8% is flagged. Above 10% triggers escalation. AEs typically run at narrower load bands than ME, so variance shows up faster.

2.3 Daily fuel-cost impact

Variance is converted to dollars per day so the review carries commercial weight:

\Delta C_\text{day} = \frac{(\text{SFOC}_\text{actual} - \text{SFOC}_\text{design}) \cdot P \cdot 24}{10^{6}} \quad \text{[MT/day]}

\text{Cost}_\text{day} = \Delta C_\text{day} \cdot \text{Price}_\text{fuel} \quad [\$]

For a 1.2 MW AE running at 75% load with 5% variance and a

700/MT MGO price, that's roughly **

120/day of waste**, or $44k/year per engine — usually enough to justify an injector replacement.

2.4 Six-month parameter rollup

The same dashboard format used for ME applies — every monitored parameter per engine per month. The grid logic, condensed:

# For each AE (AE1 / AE2 / …) build a six-month grid keyed by parameter
for ae_number in sorted(engines):
    for parameter_key in PARAMETERS:                  # PMAX, EXHAUSTTEMP, …
        main_row = [f"{ae_number} {parameter_key}"]

        for month_date, month_label in zip(six_months_dates, six_months_labels):
            performance = performance_data.get(month_date)

            if performance == "No Data Available":
                current = (month_date.month == datetime.utcnow().month and
                           month_date.year  == datetime.utcnow().year)
                main_row.append("Report yet to be submitted" if current
                                else "Missing Report")
                continue

            value = performance["data"].get(f"{parameter_key}AVERAGE")
            main_row.append(round(value, 2) if value is not None else "No Data Available")

Two states matter in the gaps: report yet to be submitted (current month, expected) versus missing report (past month, compliance issue). The analyzer surfaces both.

2.5 Cylinder uniformity

Same logic as the main engine — cylinder deviation against the engine mean:

\Delta T_i = T_i - \bar{T}, \quad \bar{T} = \frac{1}{N}\sum_{i=1}^{N} T_i

Severity	Threshold
OK	$\lvert \Delta T_i \rvert \leq 30\,°\text{C}$
WARNING	$30 < \lvert \Delta T_i \rvert \leq 50\,°\text{C}$
CRITICAL	$\lvert \Delta T_i \rvert > 50\,°\text{C}$

2.6 Overhaul timing

H_\text{remaining} = H_\text{interval} - (H_\text{current} - H_\text{last\_overhaul})

Tiered identically to ME (see main engine page). The difference is that an AE overhaul typically degrades the redundancy tier above — booking maintenance forces a tier check on the rest of the fleet.

2.7 Blackout-risk score

The composite that drives the senior verdict:

R_\text{blackout} = w_1 N_\text{fault} + w_2 (1 - C_\text{margin}) + w_3 R_\text{redundancy} + w_4 N_\text{overdue}

where:

$N_\text{fault}$ — count of engines in FAULT
$C_\text{margin}$ — capacity margin = $\frac{\text{available kW} - \text{load kW}}{\text{available kW}}$
$R_\text{redundancy}$ — redundancy tier multiplier (LOW=0, MEDIUM=1, HIGH=2)
$N_\text{overdue}$ — overhauls overdue

The composite maps to LOW / MEDIUM / HIGH / CRITICAL. HIGH or CRITICAL escalates immediately.

Worked example

Vessel: MV POSUN (IMO 9388340), 3 × MAN 7L21/31 generators. After a refresh:

Engine	Status	Hours	SFOC variance	Notes
AE1	RUNNING	12,400	+6%	Cylinder 4 hot (38 °C above mean)
AE2	STANDBY	8,200	—	OK, available
AE3	FAULT	11,900	—	Tripped on low LO pressure 6 days ago

State: 1 running, 1 standby, 1 in fault → redundancy tier HIGH, capacity margin tight if AE1 fails. Composite blackout-risk: HIGH. 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 engine evidence and recommended actions

Recommended actions:

Investigate AE3 LO-pressure trip root cause — restore service before next port departure.
Replace AE1 cylinder 4 fuel injector at next port — variance plus cylinder hot signature.
Confirm AE2 standby readiness drill within 24h while AE3 is out.

Output deliverables

Executive summary — overall AE status, redundancy tier, blackout risk
Per-engine table — status, load, hours, SFOC variance, alarms
Cylinder uniformity — per-cylinder deviation per engine
Six-month rollup — every monitored parameter per engine per month
Power-management assessment — load profile, capacity margin, blackout-risk score
Overhaul plan — urgent / upcoming, port-call coordination
Cost impact — variance translated to daily and annual fuel waste
Recommendations — prioritised actions with timeline
Escalation decision — auto-routed to TSI when triggered

Escalation triggers

Trigger	Severity
Any engine in FAULT	CRITICAL
Blackout-risk score = HIGH or CRITICAL	CRITICAL
Cylinder deviation above 50 °C	CRITICAL
Overhaul overdue or under 500 hours away	CRITICAL
Multiple engines showing WARNING	HIGH
SFOC variance above 8% on any engine	HIGH
Load-sharing failure	HIGH
Exhaust temp above 420 °C	CRITICAL

Why script-driven

Redundancy tiers, SFOC variance, and blackout-risk weighting all live in deterministic Python. The reviewer interprets the result; the verdict is reproducible. A review where the redundancy tier shifts from MEDIUM to HIGH between two consecutive runs is a real event in the data, not a model decision.

References

Source templates

Performance analysis suite — AE monthly rollup and engine specifications & running hours.

Related: Main engine

Same analysis pattern applied to the main propulsion plant.

Related: Lube oil

AE LO trends often confirm cylinder findings.

Related: Fuel oil

Fuel quality is a common cause of variance and cylinder asymmetry.

Documentation Index

​Executive summary

​Architecture

​Stage 1 — Data collection

​Where the data comes from

​What gets captured

​Output

​Stage 2 — Analysis

​2.1 Engine availability and redundancy

​2.2 SFOC variance

​2.3 Daily fuel-cost impact

​2.4 Six-month parameter rollup

​2.5 Cylinder uniformity

​2.6 Overhaul timing

​2.7 Blackout-risk score

​Worked example

​Output deliverables

​Escalation triggers

​Why script-driven

​References