Home  >  Article  >  Technology peripherals  >  Generative AI in ocean engineering: Insufficient proprietary data sets are limiting its practical applications

Generative AI in ocean engineering: Insufficient proprietary data sets are limiting its practical applications

WBOY
WBOYforward
2023-10-30 14:21:121454browse

Generative AI in ocean engineering: Insufficient proprietary data sets are limiting its practical applications

Modern computing is significantly improving the design and construction processes of shipbuilding and offshore engineering, but limited data sets are hampering further integration.

The discipline name Naval Architecture and Marine Engineering (NAME) may only be a few hundred years old, but its origins can be traced back to ancient civilizations thousands of years ago, when humans were already building ships to explore the world. , carry out commercial activities. Many people, including Archimedes, Bouguer and Chapman, had refined the concepts of buoyancy, stability and ship design into scientific methods and theories.

Naval Architecture and Marine Engineering is a professional engineering discipline involved in the design, construction, testing, surveying, maintenance and operation of marine vessels and structures. I graduated from the U.S. Coast Guard Academy with a bachelor's degree in marine construction and ocean engineering and received a master's degree from the University of California, Berkeley. For the past 22 years I have been working as a shipbuilder for a private marine consultancy, working on the design of passenger ships, marine research vessels, barges and other vessels

Naval architecture is an all-encompassing discipline that encompasses Design floating docks, complex 1,200-foot cruise ships, and even an aircraft carrier that resembles a sea town. Related professionals are also responsible for the design of offshore wind power platforms, submarines, container ships, autonomous vessels and virtually any vessel that operates underwater or on top of the water

Contemporary naval architects use pencils all the time in their college courses , drawing lines and small models to learn professional knowledge, including myself. However, the real-life design process has fully incorporated advanced computer applications based on machine learning.

Line drawing in the classroom is used to outline the shape of the hull, thereby evaluating the design and construction process of the ship. When using pencils and drawing lines, you often need to draw and erase repeatedly, and use visual inspection to determine whether the curve is smooth enough. But now, many software can help us quickly draw hull line diagrams, and use algorithms to verify the current plan using collected historical data.

The enhancement of modern computing power allows engineers to test variables and evaluate their effectiveness in seconds, a task that may have taken hours or even days in the past. The reason why this efficiency improvement can be achieved is inseparable from people's unremitting exploration of scientific principles and calculation formulas over the past hundreds of years. In addition, the software contains minimum security standards and compliance requirements that must be met.

As the saying goes, “the marine industry has always been slow to embrace new technologies.” However, if we walk into a modern shipyard or engineering design company, we will see the routine application of technologies such as 3D modeling, computational fluid dynamics, finite element analysis and robotic manufacturing. This shows that this understanding is often incorrect. Now, the shipbuilding industry should also consider introducing more advanced tools, such as machine learning and artificial intelligence

Nowadays, shipbuilding and marine engineers use advanced software packages to greatly improve the efficiency and accuracy of design and development, and Some of these tools are particularly suitable for integration with machine learning and AI. For example, computational fluid dynamics (CFD) primarily uses the Navier-Stokes equations to model the fluids in which ships travel.

Thanks to improvements in modern computing power, we can evaluate physical phenomena through calculations. But even with the most powerful computers available, such simulations often require a lot of time and considerable investment. Therefore, tow tanks are still widely used to evaluate and measure hull shape and performance. But with the help of computational fluid dynamics, we can now also create a "virtual" tow tank session to discuss with the AI ​​how to optimize the hull shape, and then conduct further testing and verification through actual scenarios. In an ideal world, the optimal design would not even need to be tested with a physical tow tank. Machine learning can reduce the running time of such simulations, saving customers money while reducing the potential for errors.

The current main obstacle to the further application of AI beyond basic machine learning algorithms is that the proprietary data sets available for AI are too limited. For AI to be successful, AI systems typically require large amounts of data to extract and construct effective query responses. Without powerful data sets, AI applications in shipbuilding and offshore engineering will not be effective. Our company has realized that collecting and applying real-world data over a multi-year cycle has become an important prerequisite for reducing errors and improving design, so we have begun taking steps to actively organize and use this data in future designs

Certain areas of the shipbuilding and marine engineering disciplines have indeed accumulated large data sets that are being used to improve engineering design, including wave and current measurements, ship maneuvering records, and marine equipment performance logs. Although there is a large amount of data available for such activities, they are often maintained by government agencies, operators and equipment manufacturers, and there is currently no global data set that is easily accessible to all stakeholders.

The practical application of ship design data is still quite limited for a number of reasons. These reasons include privately or government-funded designs (data are for owner use only), specialist ship designs (only available for specific types of ships, limited scope), and lack of necessary funding to collect, validate and deliver actually usable data

Australian ship design manufacturer Austal has revealed some details of its DeepMorpher tool. This tool uses AI and machine learning technology to optimize the hull shape. The company said in the news that it uses high-performance computing to reduce the execution time of computational fluid dynamics, while also working with its 3D hull design data set. Austal has a large library of hull shapes and, after testing and validation, found that DeepMorpher's performance was orders of magnitude higher than several other existing models.

If data anonymity can be guaranteed throughout and open data access can indeed bring quantifiable benefits to owners, then the current strict restrictions on proprietary data should be eased in the future. In addition, certain design decisions, such as master plans designed for specific use cases or customer needs, can remain proprietary and closed to the public.

However, some other design decisions are made to meet the mathematical certainty or regulatory requirements required for the overall design. If owners, including government agencies, can realize how data and artificial intelligence systems can improve the efficiency of design work, they may take the initiative to provide data sets

There is no good solution for various special ship types. The application space may be common between different ships in some aspects, so you can try to mine relevant data. But even so, the process of validating the data and checking whether it can be transferred requires a lot of time and money, so it is necessary to find ways to "clean" the data to control costs and improve its actual scope of application

Shipbuilding and Marine engineering is an interesting field of engineering that requires good design in terms of hull shape, internal structure, power generation, distribution, interior design, living and working arrangements, ergonomics, etc., to form a floating offshore platform that can sustain life. and assist the crew in completing work on the platform.

All space and system design takes a considerable amount of time, and must meet the needs of multiple parties at the same time, and various potential design errors must be discovered and resolved in a timely manner. AI can be used to reduce errors, ensure compliance with compliance requirements, meet a variety of ergonomic needs, and provide a "just works" electrical system and equipment layout design.

Today's modern ships have extensive electrical and control systems, and the entire ship is controlled by computers, and new all-electric/hybrid ferries need to be equipped with advanced control and monitoring systems. AI can help us discover and resolve conflicts between these systems during the design phase faster than human experts, and hopefully reduce errors in judgment.

Water transport is the main method of trading goods, accounting for about 90%. Millions of people travel around the world on ferries and other passenger ships every year. The demand for talents in the marine industry is very high. Due to the limited number of qualified crew members, many scheduled flights had to be postponed or canceled. Only a few universities in the United States offer degrees in naval architecture and ocean engineering, making it difficult to promote research and application of machine learning and artificial intelligence technologies in this field. Human resources problems can be solved through powerful AI technology, but This falls into an endless cycle of which came first, the chicken or the egg. Considering that government projects have signed multiple ship contracts, can retain some design openness (as long as it does not affect ship safety), and have generous R&D funding reserves, it seems that the shipbuilding industry can cooperate with the U.S. Coast Guard or the U.S. Navy to jointly explore The application prospects of AI technology

The U.S. Navy has a large fleet of ships, so it is particularly suitable as a starting point for the exploration of AI and machine learning. Task Force Hopper, launched in 2021, hopes to accelerate the AI ​​capabilities of the entire Navy surface fleet, but it will mainly focus on the combat level. However, with the support of a large amount of available data and a series of ship designs, it is believed that AI is likely to be applied to actual ship design in the future.

The above is the detailed content of Generative AI in ocean engineering: Insufficient proprietary data sets are limiting its practical applications. For more information, please follow other related articles on the PHP Chinese website!

Statement:
This article is reproduced at:51cto.com. If there is any infringement, please contact admin@php.cn delete