Recognized globally within the realm of maritime and ocean research and development under the context of Industry 4.0, there is a widespread understanding of the need for enhanced maritime safety and digitalization to overcome current and future challenges.  Consequently, the development of cutting-edge and demanding technologies rooted in Artificial Intelligence (AI) has become imperative to enhance the perception and cognitive abilities of marine systems. It is truly essential to develop systems with intelligence to tackle the futuristic challenges to increase the autonomy of ships for a sustainable future.

The research will focus mainly on the identification and application of appropriate mathematical (Machine learning) methods/techniques to advance the perception capabilities of new/existing marine systems. Specifically speaking, the research will aim at developing trustworthy digital tools/algorithms to optimise the energy consumption and provide health management solutions for marine systems. The focus will be given to verification and validation of the developed algorithms and frameworks ensuring reliability and safety of the operations in collaboration with Lloyd’s register. 

Research gaps

• adaptability to dynamic environment
  
  • The operating state depends on sea conditions, weather conditions, variable operating loads and long working durations. The data collection reliability also depends on the sensor condition, uncertainty, and health of the machine. Therefore, self-learning/correcting machine learning models corresponding to the changing conditions need active focus.

• multimodal integration
  
  • Various surrogate machine learning models predicting performance have been developed in literature considering time series data. Different sensors collect data from multiple modes containing important information that helps understand the operating state of the machine and need to be captured. Advanced fusion models extracting data from these multi-modal sensors are very essential to gain information about the surrounding and future state of the machines.

• Human-machine collaboration
  
  • The development of autonomous machines using AI models also requires integration of humans to provide informed decisions to crew members and personals. Instead of creating black box models giving results, these models/algorithms should contain a link that could explain crew the perception process and involve years of experience. The integration of human understanding to enhance capability and reliability is necessary to carry out efficient operations.

• robustness and trustworthiness
  
  • The trustworthiness and robustness of the data-driven models is very important for safety and accurate decision making. The literature review shows a lack of studies for validation and verification in diverse conditions.

Objectives

• Isolation of fault and failure conditions at the replaceable parts level. A health understanding of the system should be based on fundamental knowledge of the material condition of replaceable parts in the components and equipment.
• Maximizing the information extracted from a minimum number of sensors to convey useful information to aid in the awareness of the current state of the system and decision making to reduce the cost and complexity of the system and improve the overall operational efficiency. Identification of coverage (i.e. parts that can be monitored) and insight sophistication (fault detection, diagnostics or prognostics) given the number and configuration of sensors in the system.
• Minimum computational costs targeting onboard deployment of developed algorithms and models, ensuring the real-time data processing and decision-making, to reduce reliance on external data canters or cloud services, and lower computing costs can reduce energy consumption, which is critical for sustainable maritime operations.
• Explainable artificial intelligence methods instead of black box models so as to help the crew better understand the algorithms while keeping them efficient, in order to more easily correct biases and make better decisions. Definition of verification methods and tools against the machine learning development and operationalisation ML DevOps framework, to allow run-time AI explainability.

Further information

The student can get the opportunity to spend 1 to 2 years in Singapore for research stay at A* Agency for Science Technology and Research under A*STAR Research Attachment Programme (ARAP).