Exploring large language models (LLMs)-based chatbots for pump maintenance in ships

초록

With the growing interest in the Fourth Industrial Revolution and autonomous ships, interest in maintenance is also growing. Difficult access to ships complicates maintenance work, and the decline in specialized engineers is leading to a shift to remote maintenance systems. These remote maintenance systems reduce costs and improve capabilities. However, existing systems require minimal network usage to collaborate with onshore experts for remote maintenance. However, remote maintenance is challenging in areas such as the ocean or submarines, where networks are unstable. Therefore, we conducted research on a chatbot based on a large-scale language model (LLM), which can guide workers through maintenance tasks without relying on networks. The LLM model used the BERT algorithm, and the chatbot was developed to recommend maintenance tasks by incorporating field information and work history. Finally, the model's results were evaluated qualitatively and quantitatively to verify its effectiveness.

키워드