[Edaily Reporter Shin Yeong-bin ] Physical AI company MakinaRocks Co., Ltd. has conducted a pilot test of an AI-based anomaly detection solution HD KOREA SHIPBUILDING & OFFSHORE ENGINEERING(009540)specifically designed for welding robots used in shipyards. The company explained that it has confirmed the feasibility of this technology—which detects signs of robot malfunctions based on normal operation data—in manufacturing environments where failure data is scarce.

MakinaRocks Co., Ltd. announced on the 4th that it had completed the development of a physical AI-based anomaly detection model for welding robots—a core process in the shipbuilding industry—in collaboration with HD KOREA SHIPBUILDING & OFFSHORE ENGINEERING.

This project was carried out as a follow-up to the “Memorandum of Understanding on the Use of AI for Robot Process Efficiency” signed by the two companies last year. The goal is to rapidly apply AI technology to shipyard sites to enhance the stability and efficiency of robotic processes.

In the shipbuilding industry, welding is a critical process that determines a vessel’s structural stability. Even minor malfunctions in welding robots can lead to large-scale rework and losses in upstream and downstream processes, underscoring the growing importance of anomaly detection technology that proactively monitors robot status.

Given the limited availability of actual failure history data in industrial settings, this project employed an anomaly detection method based on normal operation data. MakinaRocks Co., Ltd. identified the primary causes and patterns of abnormal signs in welding robots through the analysis of data from experimental facilities.

Subsequently, the model was refined using operational data from 12 welding robots actually in use at a shipyard. The model was validated alongside a monitoring function that visualizes anomaly scores and key signals.

Based on the results of this pilot application, MakinaRocks Co., Ltd. plans to continue exploring the potential for integration with shipyard operational systems. The company will also examine ways to refine its Physical AI model using on-site robot data and physical signals and connect it to standard shipyard operational systems.

Yoon Seong-ho, CEO of MakinaRocks Co., Ltd., stated, “Physical AI—which overcomes the data limitations of various physical assets in industrial settings, including robots, and enables AI to function—is an unavoidable technological trend for global manufacturing companies.” He added, “We will set a technological milestone for Physical AI that can be applied in various ways within the shipbuilding industry.”