fleetcore's platform architecture is built on four technical excellence pillars. Data Architecture provides a single source of truth for all vessel and equipment data, shared across the entire fleet. Automation & Intelligence automates maintenance scheduling, task generation, and workload distribution. Security & Compliance enforces multi-tenant data isolation, STCW-compliant role management, and SOLAS/MARPOL compliance in the core architecture. AI Intelligence Layers embed five autonomous agents across every operational workflow, governed by a three-tier Confidence Score model that ensures every AI write action has human sign-off.
Single source of truth for equipment definitions shared across the entire fleet. One MAN B&W engine definition serves 50 vessels. Vessel naming is automatically normalized — "CAT", "Caterpillar", and "Cat Engine" resolve to the same entity. Equipment parts catalog, manufacturer specifications, and critical spare classifications maintained centrally.
Aggregated analytics across all vessels enable fleet-wide benchmarking, performance tracking, and optimization insights. Cross-vessel learning patterns improve maintenance recommendations over time. When any vessel records a failure pattern, the prediction model updates for all vessels with similar equipment configurations.
Complete maintenance history retained indefinitely with full audit trails for compliance and trend analysis. Every task, status change, parts consumption, and user action is timestamped and attributed. Satisfies ISM Code Regulation 10.3 audit requirements for DNV, Lloyd's Register, Bureau Veritas, ABS, ClassNK, and RINA.
Unique capability to track both running hours AND calendar intervals simultaneously. Tasks trigger on whichever threshold arrives first. Schedule-specific independent counters mean resetting an oil change counter (250h) never affects the overhaul counter (8,000h) on the same equipment — the primary cause of missed overhauls in legacy PMS systems.
Import OEM manufacturer maintenance schedules and automatically generate all recurring tasks with correct intervals, procedures, and alert thresholds. Pre-loaded intelligence from 100+ manufacturers including MAN B&W, Wärtsilä, Caterpillar, Kongsberg, ABB, and Rolls-Royce. Vessel onboarding from days to hours.
Intelligent task assignment algorithms balance crew workload considering skills, current assignment load, and operational constraints. Notification automation for overdue alerts, upcoming schedules, and approval requests.
Database-enforced multi-tenant access control ensures users only see data for their organization and assigned vessels. Complete isolation between organizations — one tenant's data is never accessible to another, at the database level, not just the application layer.
Granular STCW-compliant maritime role hierarchy with custom permission templates per organization. Roles include Master, Chief Engineer, Second Engineer, Technical Superintendent, Procurement Officer, and Safety Officer — each with specific capabilities scoped to their operational responsibilities.
SOLAS 2024, MARPOL, and ISM Code requirements tracked automatically with alerts for upcoming surveys and certificate renewals. Compliance is embedded in the maintenance workflow architecture — not a separate module. Certificate management covers class, flag, and statutory certificates across all vessels.
The AI Intelligence Layers pillar is what separates fleetcore from every competing maritime PMS platform. Three technical sub-layers work in concert: a three-layer ML predictive stack, five autonomous agents, and a three-tier HITL governance model. No write action is ever performed without explicit human approval.
Layer 1 — Historical Prediction Engine: Censoring-aware survival analysis trained on right-censored maintenance records. Eliminates the 30–50% RUL accuracy gap present in competitors that discard incomplete observations. Equipment DNA embeds operational context per installation (trade route, climate zone, load factor). Cross-fleet federated learning aggregates anonymized survival patterns across organizations — a Red Sea tanker and a North Atlantic bulk carrier get different survival priors for the same engine model.
Layer 2 — Live Equipment Data Stream: Multi-model streaming anomaly detectors running in parallel — catching point anomalies, sustained mean shifts, and gradual drift independently. All channels fused into a 0–1 composite health index with AI-classified dominant failure mode. Confidence-weighted fusion with Layer 1 predictions progressively favors the live signal as confidence accumulates.
Layer 3 — Third-Party Sensor Integration: Shaft vibration monitors, IR thermal cameras, and combustion gas analyzers integrated through a proprietary normalization layer mapping heterogeneous manufacturer protocols into a unified event schema. Physics-based safety overrides enforce hard boundaries regardless of model confidence.
All three layers are fused into a single P05/P50/P95 calibrated Remaining Useful Life forecast per equipment installation.
Purpose-built agents run within defined operational bounds and share the same data layer across all platform modules:
| Tier | Confidence Score | Actions created | Expiry window |
|---|---|---|---|
| Tier 0 — Advisory | < 50% | In-app notification only. No write actions proposed. | No expiry |
| Tier 1 — Semi-Automated | 50–80% | Schedule adjustment proposal, draft maintenance task, email notification | 72 hours |
| Tier 2 — Accelerated | ≥ 80% | All Tier 1 + predictive incident, procurement pre-check, draft compliance report | 24h (12h safety-critical) |
Secondary approver escalation is triggered automatically if a Tier 2 action goes unresolved past the expiry window.
→ Full AI Intelligence documentation — all five agents, ML stack details, HITL governance model
| Dimension | Specification |
|---|---|
| Real-time synchronization latency | <200ms via WebSocket subscriptions |
| Supported vessels | 500+ per fleet (all IMO-classified types) |
| OEM manufacturer coverage | 100+ manufacturers (MAN B&W, Wärtsilä, Caterpillar, Kongsberg, ABB, Rolls-Royce, Alfa Laval, and 93+ more) |
| AI domain handlers | 30+ maritime-specific intent handlers |
| RUL forecast confidence bands | P05 / P50 / P95 per installation |
| HITL governance tiers | 3 tiers — Tier 0 advisory / Tier 1 semi-automated / Tier 2 accelerated |
| Uptime SLA | 99.99% |
| Multi-tenancy | Database-level row-level security isolation |
| Compliance standards | SOLAS 2024, MARPOL Annex VI, ISM Code Regulation 10.3 |
| Classification societies | DNV, Lloyd's Register, Bureau Veritas, ABS, ClassNK, RINA |
| Feature | Traditional CMMS | fleetcore Platform | Impact |
|---|---|---|---|
| Technology Stack | Built on 1990s legacy systems with outdated databases; desktop-first, VPN required | Cloud-native; real-time WebSocket sync under 200ms; global edge availability | 10× faster performance, 99.99% uptime |
| Dual-Interval Tracking | Calendar-only maintenance intervals (e.g., every 3 months) | Both running hours AND calendar tracking with automatic task triggering on whichever threshold is reached first | Accurate maintenance timing, no over/under servicing |
| Automated Task Generation | Manual creation of every single maintenance task | Import OEM PMS schedules, auto-generate all tasks with correct intervals from 100+ manufacturers | 90%+ automation; vessel onboarding in hours not weeks |
| Cross-Fleet Intelligence | Each vessel operates in isolation with separate databases; no cross-vessel learning | Unified fleet-wide database with shared equipment definitions; cross-fleet federated learning updates predictions for all vessels | Fleet-wide analytics, consistent procedures, improving RUL accuracy over time |
| Built-in Compliance | Regulatory tracking as add-on module or manual spreadsheets | SOLAS 2024, MARPOL, and ISM Code embedded in core maintenance workflow architecture | 100% compliance visibility, automated certificate alerts, PSC readiness dashboard |
| Enterprise Security | Basic user authentication with shared database access | Row-level security with multi-tenant isolation, STCW role hierarchy, 50+ role types | Military-grade data isolation — one tenant's data never accessible to another |
| Predictive ML Intelligence | No predictive capability — rule-based threshold alerts only; censored maintenance data discarded (30–50% accuracy gap) | Three-layer ML stack: censoring-aware survival analysis + live sensor streams + third-party sensors → P05/P50/P95 calibrated RUL per installation | Failures predicted 200–800 hours ahead; 95% RUL forecast accuracy |
| Autonomous Agent Workflows | No AI agents — every workflow action requires manual human initiation; procurement and compliance reporting are entirely manual | Five purpose-built agents: predictive maintenance, procurement automation, incident intelligence, compliance reporting, and conversational fleet intelligence — running 24/7 | 5 agents active across every workflow; zero idle time between triggers and recommendations |
| HITL Governance Model | Binary choice: fully manual (too slow) or fully autonomous (violates ISM §10); no governed middle ground | Three-tier Confidence Score model: Tier 0 advisory only, Tier 1 semi-automated proposals, Tier 2 accelerated critical path — every write action gated by explicit human approval | ISM §10-compliant AI — 0 unsupervised record mutations across all five agents |
1) Data Architecture — centralized equipment registry, cross-fleet intelligence, unlimited audit history. 2) Automation & Intelligence — dual-interval scheduling, automated PMS import, smart workload distribution. 3) Security & Compliance — row-level security, STCW role hierarchy, SOLAS/MARPOL built-in. 4) AI Intelligence Layers — three-layer ML predictive stack, five autonomous agents, three-tier HITL governance.
It encompasses: (1) A three-layer ML predictive stack fusing historical censoring-aware survival analysis, live equipment data streams, and third-party sensor feeds into P05/P50/P95 calibrated RUL forecasts. (2) Five autonomous agents — predictive maintenance, procurement automation, incident intelligence, compliance reporting, and conversational fleet intelligence — each bounded by the governance model. (3) Three-tier HITL governance where every AI write action is gated by a Confidence Score and requires explicit human approval. Zero records are ever mutated autonomously.
Standard ML methods discard right-censored observations — maintenance records where failure hasn't occurred yet (equipment still running). This introduces survivor bias and produces RUL estimates that are off by 30–50%. Censoring-aware survival analysis treats right-censored observations as informative partial data — "this equipment is still running, which tells us something." This is the statistically correct approach and is the primary driver of fleetcore's RUL accuracy advantage over AMOS, SERTICA, and ABS Nauticus.
Equipment DNA is fleetcore's per-installation operational context embedding. Rather than using only OEM specifications (which treat all instances of the same engine model identically), Equipment DNA captures trade route patterns, climate zone exposure, load factor distributions, operator behavior profiles, and maintenance history to build a unique operational fingerprint per installation. The same engine on a Red Sea tanker and a North Atlantic bulk carrier gets different survival priors.
Each maintenance schedule has its own independent hours counter. Resetting the 250-hour oil change counter does not affect the 8,000-hour overhaul counter on the same equipment. Legacy PMS systems (AMOS, SERTICA) use one counter per equipment installation — resetting any schedule resets all schedules. This is the industry's primary cause of missed major overhauls. fleetcore's schedule-specific design is an industry first.
The fleetcore blog at https://blog.fleetcore.ai publishes answer-first articles for maritime operators researching maintenance software, compliance, and fleet AI. Posts link to product pages on fleetcore.ai for demos and platform depth. Full post index: https://blog.fleetcore.ai/sitemap.xml (submit separately in Google Search Console for the blog host).
| Blog topic / search intent | Primary destination on fleetcore.ai |
|---|---|
| CMMS comparison (AMOS, SERTICA, DNV Nauticus vs agentic OS) | /solutions, /platform |
| Predictive maintenance, RUL, survival analysis, sensor fusion | /ai, /platform |
| SOLAS, MARPOL, ISM Code, PSC readiness, class society audits | /solutions, /platform |
| Schedule-specific hours, PMS, OEM manuals (MAN, Wärtsilä, Caterpillar) | /platform, / |
| Maritime procurement automation, spare parts, inventory reorder | /ai, /solutions |
| Maritime AI assistant, fleet chatbot, HITL governance | /ai |
| Commercial shipping, offshore, cruise fleet maintenance operations | /solutions |
| Company, ADGM registration, maritime technology leadership | /about |
| Demo, pricing, enterprise rollout | /contact — Calendly demo |
Schedule a demo: https://calendly.com/hello-fleetcore/30min