Gains:
- Being able to distinguish where AI saves real time in the ship and marine engineering workflow and where the responsibility for life/property safety and class should remain with the engineer.
- Ability to apply a three-layered discipline that verifies each engineering output by order of magnitude, independent recalculation, and class rule/source control
- Ability to acquire the habit of anonymizing the context and creating prompts to benefit from AI without sharing commercial and technical confidential data of the ship, shipyard and shipowner.
Look at a marine engineer's day: hull form iterations, resistance and propulsion calculations, stability and strength checks, class rule scans, engine room sensor records, survey reports, shipyard correspondence and endless emails with the shipowner. The time devoted to the actual engineering judgment, that is, the questions "is this boat safe, does this construction meet the class code, does this voyage plan really reduce fuel consumption", is crushed under the repetitive calculation and document work. This is where artificial intelligence (AI for short; software that works on text, code and numbers with a large language model and machine learning) comes into play. AI doesn't make the decision for you; It prepares you for the decision, produces a draft, speeds up the calculation and puts processed information in front of you.
However, marine engineering is a field where mistakes are measured not by text but by lives, the environment and millions of dollars. A class rule clause made up by a language model, an incorrect unit conversion, or a “reasonable-looking” but physically impossible stability result may be a minor correction in other sectors, but here it can turn into a capsize, a structural collapse, or an oil spill. So throughout this module we will position AI not as an “automatic engineer” but as a safety-critical disciplined assistant whose output is verified every time.
In this first unit, we clarify three things: At what stages of the shipboard and marine engineering workflow does AI add real value; which decisions should remain strictly with the qualified person (engineer, class surveyor, master); and what is the verification, confidentiality and liability discipline you must adhere to when doing so. Without this roof installed correctly, techniques in subsequent units may become dangerous.
Concepts: Hallucination: AI's convincing fabrication of a number, coefficient, class rule clause, or source that does not actually exist. Context: The input you give to the AI (boat data, assumptions, question). Verification: Checking the output in an independent way (hand calculation, second software, rule text, model experiment). Classification society: The organization that inspects and certifies the design and construction of the ship according to the rule set (e.g. ABS, DNV, Lloyd's Register, Bureau Veritas, TL/Türk Loydu). These concepts are the backbone of the entire module.
In Which Businesses Is AI Accelerator, In Which Businesses Is It Risky?
Marine engineering tasks fall on a two-pronged spectrum in terms of their results. At one end are retrievable, low-risk prep work (a literature summary, a code skeleton, a presentation outline); At the other end, there are irreversible decisions (a safety coefficient, a stability criterion, a structural approval) that directly determine maritime safety, life/property safety and the environment. The value of AI varies depending on where you stand on this spectrum.
business type
AI contribution
Engineer's role
Rule/standard summary
Extracting essence from the long rule book
Comparing the clause with the original class text
Hand calculation / pre-sizing
Formula setting, first number
Unit, rank and assumption checking
Analysis code / Python script
Skeleton and logic generation
Validation with test input, edge case testing
Report/survey draft
Suggesting structure and narrative
Link each value to the measurement and source
Sensor/IoT data analysis
Pattern and anomaly candidate extraction
Confirmed by raw data and physical plausibility
Stability/strength approval
Analysis material preparation
Final assessment, class approval, signature
The rule is simple: the risk of an AI output equals the damage it will incur if that output makes an error. Misspelling the title of a slide is harmless; In a loading situation, it is disastrous to miscalculate the GM (metacentric height; a measure of the ship's righting ability) value and incorrectly determine the ship's rollover limit. So the first question to ask before using the output is: "What happens if this is wrong, who gets hurt, and who notices?"
Attention: AI produces fluent and confident text. Fluency is no guarantee of accuracy. A language model fills the gap with statistically "most likely" words, even if it doesn't have exact data; In maritime, this gap filling may appear as a made-up material yield strength or a non-existent SOLAS clause.
Security-Critical Verification Discipline: Three Layers
Maritime culture is already based on the principle of "trust but verify"; Survey, double checking and independent class certification are part of this culture. AI should be used to strengthen this principle, not weaken it. Pass each AI output through a three-layer filter.
First layer: order of magnitude control. See if the result is roughly within the expected power of 10 range. The displacement (weight of the submerged volume) of a Panamax bulk carrier is on the order of tens of thousands of tonnes; If the AI tells you 800 tons, you know there's a mistake without going into detail. The cruising speed of a container ship is in the range of 18-24 knots; 90 knots is physically impossible.
Second layer: unit and size consistency. In maritime, unit confusion is a classic source of error: knots versus m/s, tonne versus kip, kW versus horsepower (BHP), meter versus feet, bar versus kPa. The unit of each number in the AI output should be clearly stated and dimensional analysis should be performed. Is a resistance value given in kN or kgf? This directly changes the propulsion power and therefore the fuel budget.
Third layer: independent reproduction. Reproduce a critical result with a second method (a different manual account, separate software, approved loading computer, or a second engineer). Confidence increases if two independent paths give the same answer; If it doesn't, stop until you understand why it's different.
Tip: When you ask the AI for a result, also ask the same question in reverse. After saying "Calculate the propulsion power required for this boat", say "Recalculate the cruising speed with the power you provided and tell me if it is reasonable for a typical ship." Asking the model to cross-check its own output makes silent errors visible.
Confidentiality and Trade Secret: Maritime's Private Frontier
The hull form, hydrodynamic performance curves, shipowner-shipyard contract data, and the ship's operational telemetry are highly sensitive in terms of trade secret and intellectual property. Entering a boat's true resistance curve or a shipowner's true fuel consumption data into an uncontrolled external AI service can be both a breach of contract, a loss of intellectual property, and a competitive disadvantage. In addition, information about the OT (operational technology; control software that manages machinery and navigation systems) architecture of modern ships should not be shared for cyber security reasons.
Rule of thumb: never enter real, confidential or contractually protected data into an unapproved external tool. Instead, anonymize the context and replace real numbers with representative values. For example, instead of sharing the form data of a real ship, ask "explain the resistance approach with an example on the order of the main dimensions of a typical Handysize bulk carrier."
Weak prompt / Strong prompt
Weak prompt:
Actual form data of Shipowner
Powerful prompt:
Role: You are an experienced consultant on ship hydrodynamics. Context: I am doing a general preliminary analysis for a bulk carrier (values are representative, not actual project data): Lpp ~180 m, beam ~32 m, draft ~12.5 m, block coefficient ~0.84, cruising speed ~14 knots. Task: Explain through which steps the preliminary propulsion estimation is made (resistance components, propulsion efficiency, sea allowance) and what data I need to measure/verify list.Constraint: Write each formula and assumption explicitly; If you give numerical example, specify the unit (kN, kW, knots) and add how to verify.
The powerful prompt protects the actual project data, clarifies the role and constraint, and requests a verification path.
General Purpose Validation Templates
You will use the following templates repeatedly throughout the module.
General template for adding a verification request:
Answer the engineering question below. In your answer:1) State the unit of each number.2) Compare the order of magnitude of the result to a typical ship.3) Write the source (method name) for each coefficient/formula you use; If there is no source, mark it as "unclear". 4) Suggest a second way of calculation so that I can check the result independently. Question: [question here]
Hallucination hunt template:
Make a table of all the numerical values, coefficients, and rule/standard references in your answer below. For each row:- Value, unit- Basis (calculation / known coefficient / assumption / uncertain)- How do I verify this independentlyAlso highlight each row whose basis is "uncertain".
Mini Cases
Case 1 — Fabricated class clause. “Which class code clause requires this minimum sheet thickness” to support a structural share, a junior engineer asks the AI. AI cites a "Chapter 6, Section 3.4.7" clause that does not actually exist. When the engineer compares it with the current PDF of the class rule, he sees that the clause is fake and finds the correct clause. Lesson: Every rule reference issued by AI is verified from the original, current classification text.
Case 2 — Unit trap. A team gets help from AI in a dispatch power calculation. AI gives the power as "8500", but its unit remains unclear whether it is "kW or BHP". There is a difference of approximately 1.34 times between 8500 kW and 8500 BHP; This difference completely changes the choice of a host and the annual fuel budget. If the crew unit advances without explicitly asking, it will head for the wrong machine class. Lesson: any number without a unit is not accepted.
Case 3 — Rank control saves lives. A team of students had the AI calculate the sailing resistance of a 15,000 DWT ship and found the total resistance to be 12 kN. For a ship of this size, the navigational resistance is in the order of hundreds of kN and the propulsion power is in the order of several thousand kW. A difference of about 100 times immediately betrays an editing error; students find an exponent error in the friction coefficient formula. Lesson: always compare the result with a known reference range.
Common mistakes
- Mistaking fluency for accuracy. A well-written explanation may be numerically incorrect. The quality of the text and the accuracy of the result are two different things.
- Accepting a unitless number. In maritime, knots/m·s, ton/kip, kW/BHP, m/ft confusions are career-ending mistakes.
- Entering confidential/contracted data into the external tool. Trade secret and intellectual property infringement; Once leaked, form and performance data cannot be retrieved.
- Trusting a single source. Using a critical result without producing it in a second way is to violate the most fundamental rule of the security-critical discipline.
- Failure to verify rule attribution. Class/SOLAS item numbers provided by AI may be fictitious; Each reference is verified from the original, current text.
In summary
AI dramatically speeds up the repetitive work of calculations, code, documents and data in marine and marine engineering; But decisions determining maritime safety, class compliance and environmental safety remain with qualified people. Each output must be filtered through order of magnitude, unit consistency, and independent reproduction. For commercial confidentiality and cyber security reasons, actual boat/performance data should be anonymized and only approved tools should be used. This discipline is a prerequisite for all subsequent units.
Application task
Choose an engineering question from your field (form, stability, propulsion, maintenance, navigation). First type a "weak prompt" and ask the AI; Then write a "power prompt" that includes the role, anonymized context, constraint, and verification request and ask the same question again. Compare the two answers in terms of order of magnitude and unit consistency, apply the “hallucination hunt template,” and confirm at least one result by an independent hand calculation. Summarize your findings in half a page.
checklist
- [ ] I evaluated the risk of the output with the question "who gets hurt if it's wrong?"
- [ ] I compared the result to a known ship in terms of size.
- [ ] I have clearly confirmed the unit of all numbers (knot, kW, ton, m).
- [ ] I reproduced the critical result by a second independent method.
- [ ] I used anonymized proxy value instead of real/hidden boat data.
- [ ] I have verified every class/SOLAS reference given by AI from the original, current text.