Maritime safety and technical management are often challenged by data reliability, complexity, and time urgency. While digitization is becoming a mainstream topic in the industry, the awareness of artificial intelligence is growing to achieve operational goals such as safety and efﬁciency.
Current safety management relies a lot on human judgment which takes advantage of human agility and has to follow this logic: Gather and process information (learn) → Manipulate information (reason) → Consider the outcomes of these manipulated data (understand). Though the logic has been followed, research shows that 75% to 96% of marine accidents involve human error. Would artificial intelligence be the solution for it?
Artificial intelligence is the simulation of human intelligence in machines.
Unlike operating a machine, artificial intelligence is expected to achieve much more than processing a specific set of commands. This means that artificial intelligence needs to learn independently, identify patterns, deduce what actions need to be taken, and implement self-optimization. Fundamentally speaking, artificial intelligence is about processing data more quickly, accurately, and extensively than humans can.
The aim of AI research has always been about understanding the function of our brain and being able to artificially recreate it. These applications like behavioral recognition and prediction that are often shown in science fiction, normally refer to strong artificial intelligence, also known as general intelligence (AGI). Though researchers in both academia and private sectors are heavily invested in artificial general intelligence (AGI), a lot of applications remain to be fantasies.
On the other hand, the majority of applications we see in real life are reactive. They are based on weak AI and rely on computational power and smart algorithms to recreate each decision every time. For instance, autonomous driving, conditional recommendations, etc. Weak AI relies on human interference to define the parameters of its learning algorithms and to provide the relevant training data to ensure accuracy.
Theoretically speaking, the maritime industry is ideal for machine learning algorithms given that huge volumes of data are generated every day in the form of shipping inspection documents and operating metrics. However, studies on the synthesis of big data applications in maritime are rare, which has created a gap in the academic literature and AI in maritime operations. This draws a picture of a promised land of artificial intelligence but is hard for companies to achieve in a short time due to lack of practice.
With those limitations experienced, many maritime safety-related projects are starting “small”. Instead of diving deep with full automation, focusing on partial assistance and niche solutions could give a much better return on investment in terms of capital and time.
COVID-19 has made flying HSEQ teams to the vessel for in-person inspection even less convenient. Though routine inspections are conducted, the data collected lacks reliability and standards. Installing sensors, on the other hand, has a upfront high cost. At a time when internet connectivity is limited, leveraging sensors to collect data is far from ideal for many shipping companies. With no reliable data, artificial intelligence is useless. The crucial point here then is less about which algorithms to choose or what sensors to use, but about how to collect reliable and standardized inspection data. In this way, a digitized tool that can assist crews to collect standardized data is far more realistic to be achieved than sensors given technical and capital challenges.
In Kaiko Systems’ case, while the app assists crews to collect standardized data, the algorithm filters data reliability and then standardizes data under different structures. With the input of highly reliable and accurate data, the safety analysis algorithms then can produce reliable and actionable vessel health management insights. This not only prevents ships from incidents but also decreases time spent on administrative overhead like inputting data in excel and identifying data manually.
AI can also support predictive maintenance in shipping, which helps to provide a reliable and precise forecast of when a machine or specific component actually needs to be maintained or replaced. When purely leveraging sensors, data on temperature, usage, tear, weather conditions, and other factors are all needed to achieve such precise forecasts.
Predictive maintenance is expected to reduce risks of downtime and gains in operational efficiency. With the current predictive maintenance development, it still needs more technical research. The impact of each component on the entire system must be further ascertained. In addition to this, a sophisticated algorithm must be able to draw reliable conclusions from the operational status of a machine or a specific component.
Given that maritime insurance claims revolve primarily around the analysis and processing of information, a shift towards artificial intelligence should deliver a significant disruption.
Artificial intelligence can help insurance companies to monitor vessels and build a high-performing, well-diversified insurance portfolio. AI-enabled insurance underwriting tool can calculate the full repair costs by identifying which parts of the vessels have been affected and how, and provide a detailed estimate, including recommended repair, as well as costs and labor hours.
Though the maritime industry has become a data-rich sector, and most marine underwriters are fully aware of the need for better data analytics, marine underwriters struggle to find sufficient and accurate data to use when selecting risks and setting prices.
On the other hand, insurtech investors have identified marine as the focal point of the technology revolution affecting insurance because of the high value of the commercial marine insurance market, the emergence of a wealth of data within the industry, and the fact that it sits at the heart of global supply chain logistics.
In order to prepare for increased levels of automation within the shipping industry, certain development matters have to be addressed:
1. Start “small”: Strong artificial intelligence requires signiﬁcant data. Comprehensive shipping data needs big upfront investment and is difﬁcult to structure, so starting from an application where data is available generates better results faster.
2. Technology standardization and regulation: AI and automation are being promoted by various researchers and technology providers. Since they are based on the development of algorithms, monitoring of both standardization and regulation is required.