SCHIFFSTECHNIK | SHIP
SCHIFFSTECHNIK | SHIP TECHNOLOGY AI and class – a Chinese perspective When speaking of AI, it is natural to focus on topics like machine learning. However, we should not neglect the contribution of other factors, China Classification Society says, thinking about the big data set, the computing power and the application softwares When speaking of AI, it is natural to focus on the AI algorithms such as machine learning, whether it being supervised, unsupervised, reinforcement, or even go deep into the generative AI or the pre-training large scale language model. However, we should not neglect the contribution of other factors, the big data set, the computing power and the application softwares, they all together make up the AI technology ecosystem, determining the overall performance of AI. Different from the current mechanism-driven softwares like CFD or FEM, by mimicking human behavior in perception, learning, reasoning and planning, AI would be able to provide a variety of solutions, from raising the design level, improving operational efficiency to making navigation decisions. As a result, AI is regarded as a driving force in preparing for the future of shipping. The application of AI Even though we may image the bright prospect of AI, frankly speaking, at present, the AI applications in the maritime industry are still at the start-up stage, but in a rapid development. Many of AIrelated cases have come into service and got attention. Decarbonization, of course, has been given the first priority. In the ship designing area: the knowledge-based optimization method has been applied to the design of hull form and propeller. In the ship navigational area: by analyzing historical data of voyage & weather, the most efficient route will be recommended; Predictive maintenance for mechanical and engine systems gradually turn into reality; Some energy efficient propulsion modes have been equipped on vessels: the propulsion with sail assistance, vessels equipped with the hybrid power or the battery power; The maritime emission monitoring platform based on big data has been constructed; In China, the first shipping large language model was published this year. Two intelligent research vessels were also constructed and put into service last year. The »world’s first« intelligent unmanned research mother ship »Zhu Hai Yun« AI for class and CCS As a class society, why should we pay attention to AI? The development of AI will become an irresistible trend, into which the shipping industry will also be brought: AI will re-engineer the way in which ships be designed, constructed, and operated in the future. Many new things turn up with functionality that has never been covered by traditional rules and regulations. Development of rules will be shifting from experience-based to mechanisms-based, supported by simulation tools. The sensors networks and IoT, by monitoring the real-time conditions, make it possible to make predictive maintenance, which will change traditional time-based survey. All of such related applications will have an impact on the traditional business scope as well as the survey and certification regimes. To support digitalization of the maritime sector and also to meet the class’s development of its own, both with an common ultimate goal that is to increase The world’s largest capacity battery-powered cruise ship »Yangtze Three Gorges 1« © CCS 50 HANSA – International Maritime Journal 10 | 2024
SCHIFFSTECHNIK | SHIP TECHNOLOGY Functional overview of digitialized survey platform for battery-powered ships safety, to reduce cost and to improve efficiency, a series of pioneer researching program on AI has been conducted, mainly distributed in following aspects: • The first thing is to apply the risk assessment for the adoption of alternative energy. Previous risk assessment, either in quantitative form or in qualitative form, just be confined to the separate evaluation target without a systematic summary, extraction and analysis on the underlying knowledge. The Knowledge engineering method, by integrating AI, database technology and mathematical logic, has been utilized to conduct Risk Assessment in a rapid way. The rule and procedure are reformulated to form a rapid risk assesment ability on new energy ships. • When it comes to the knowledge acquisition, Both the Large Language Model (LLM) and Knowledge Graph (KG) can serve as knowledge base, they are often compared with each other. In the Proof of Concept (PoC) stage, we have long been disturbed by which way to choose. The pros & cons of both methods are so obvious: the LLM is parameterized knowledge with probability-based generative model, while the KG is symbolized knowledge with stable generative model. LLM be confronted with problems such as hallucinations, black-box operations, uncertainty, lack of domain-specific knowledge, and lack of real-time updates, which are exactly what the KG is good for – highly accurate, interpretable, rich in domain expertise, and up-to-date. The characteristics of KG ensure the accuracy and domain pertinence of its content. On the contrary, the strength of LLM lies in their wide range of applications, excellent language processing capabilities, and strong generalization and generality, which are the limitations of KG. KG, especially for specific domain, relies heavily on manual intervention by experts, making them difficult to apply automatically on a large scale. For us, a technical route by considering both LLM and KG in a mutual promotion way is next step consideration. • The GHG emission service system has also been established and put into operation. The system has integrated many functions, supporting data collection, verification, compliance calculation, reporting, and decisionmaking, and can provide self-assessment, decision-making and certification services for stakeholders, including Shipyards, ship owners and designing sector. • For batter-power ships, the digitialized survey platform for batterypowered ships is another service platform: supporting real-time monitoring, algorithm analysis, safety monitoring and evaluation of batterypowered systems; The system can generate auxiliary inspection report both for the and battery and the ship operation conditions. • For the AI auxiliary decarbonization, one point should be emphasized is that AI is not the goal, but a the process, tool and method to improve efficiency and safety in maritime operations. Firstly, what we should focus on is how to achieve emission-cutting goal, including design improvements, navigation optimization, the adoption of alternative fuels and power systems, or other energy-saving measures. Secondly, we will see how AI will empower these application scenarios, and the most important is how AI can ensure the safety in the process. We need to have a deep understanding of the application scenarios of the technology and obtain effective data support. Despite the fact that AI offers broader benefits in terms of efficiency and efficacy, a broad recognition has also showed that the evolution of AI technology deployed on ships presents some challenges, requiring proportionate concerns: • Firstly, the AI application largely depends on the so called digitialized data. The class, however, usually acting as the »third party« in the industry, doses no produce data directly. The data exchange and employment need to be supported by different stakeholders while ensuring the copyright and cyber security in the process. • Secondly, the AI application, while providing insight into actual and predicted condition of a ship and its systems, has transferred the approved object from real »entity« to virtual »algorithm model«. Provided that the alternative be equivalent to traditional survey, so the reliability and precision of AI should be amply verified. • Thirdly, the AI application may pose challenge to the traditional timebased survey and certificate regimes, which means that the class rules, as well as the qualification of surveyors should be evolved synchronously. As a key link in the shipping industry, class should proactively embrace the innovative technology, bearing in mind that challenges can be addressed by thorough and considerate engineering, nothing that is insurmountable. The potential benefit of AI must surmount the challenges and assure class of a sustainable future. Author: Zhao Bingqian Digital Transformation Center China Classification Society (CCS) This article is a written version of the presentation given by the author at the recent Maritime Future Summit at SMM in Hamburg HANSA – International Maritime Journal 10 | 2024 51
