Agentic Shipping Classification Review
0.00分roblem Statement Every commercial vessel must be certified by a classification society (e.g., DNV, ABS) before it can legally set sail. This multi-million dollar certification hangs on complex engineering schematics known as Piping and Instrumentation Diagrams (P&IDs). Currently, the "Class Review" process is plagued by three massive user pain points: The Bureaucratic Bottleneck: Shipyards must submit blueprints and wait 4 to 6 weeks for a human structural or marine engineer to manually cross-reference elements against thick binders of regulatory rulebooks. The Iterative Drain: The checking process is completely manual and opaque. Minor mistakes—like an invalid valve configuration or a missing safety redundancy on a cryogenic LNG line—result in a full rejection. Multiple submission-revision cycles are the norm, not the exception. Extreme Financial Risk: Large vessels cost upwards of a hundred million dollars to build. Every week of delay caused by slow, back-and-forth blueprint corrections eats directly into the shipowner’s capital return and pushes back shipyard build schedules. The Solution NaviRule is an agentic AI web-based workspace that acts as an automated, instant pre-submission gatekeeper. Instead of waiting over a month for external human audits, marine engineers upload their digital schematics directly into NaviRule. The system uses agentic AI to map out the blueprint's topology, evaluate it against codified regulatory sets in real time, and flag errors in seconds—transforming a months-long bottleneck into a lightning-fast internal clearinghouse. 2. Core Features (Hackathon Prototype MVP) To ensure a flawless live delivery within a strict hackathon timeline, the standalone web application prototype focuses exclusively on a single-file verification lifecycle using clean data: Feature 1: Web-Based File Upload Canvas (FR-1.1) A simple, browser-based drag-and-drop workspace where marine design engineers can upload their exported vector PDF blueprint schemas directly into the dashboard. Why it's hackathon-scoped: Restricting the MVP to vector PDFs/XML schema exports keeps data pristine for the AI, removing the complex computer vision/OCR risks associated with messy legacy scans. Feature 2: Interactive Web Layout Rendering (FR-1.2) Once uploaded, the web application renders a high-fidelity visual layout of the P&ID drawing natively inside the browser window. This provides the user with an intuitive, unified canvas to review their overall layout geometry. Feature 3: AI-Driven Rule Violation Mapping (FR-1.3) An agentic AI backend automatically cross-references the diagram's component topology against a codified subset of maritime laws (such as LNG fuel gas safety loops). Non-compliant sections (e.g., a missing fail-safe valve configuration) are dynamically highlighted in bright red or yellow bounding boxes directly overlaid on the blueprint components. Feature 4: Contextual Error & Actionable Remediation Side-Panel (FR-1.4) Clicking on any highlighted red or yellow violation box opens a dual side-panel interface. Powered by a specialized multi-agent pipeline, it displays a human-friendly compliance breakdown containing three fields: the exact nature of the violation, the specific binder rulebook index breached (e.g., DNV-RU-SHIP Pt.5 Ch.7), and concrete, actionable text instructions detailing how the engineer can physically modify the CAD layout to fix it.
分数拆解
同类对比
#3 / 84 同类排名 · 其他前 4%
分数历史
2026-07-26
0.00
执行流程
如何使用
克隆仓库
git clone https://github.com/nigellauw/agentic-shipping-classification-review.git在哪里被使用
项目论坛
讨论
围绕这个项目发起主题帖,与其他 OpenArena 用户交流测试、反馈和使用体验。
登录后可以发起项目讨论、喜欢项目并收藏项目。