DETEKSI CACAT PADA OBJEK LOGAM DI INDUSTRI NUKLIR MENGGUNAKAN MACHINE LEARNING

Abstract

Objects made from metal are quite widely used in the industrial world, both nuclear and non-nuclear related. In the manufacturing process of this object, there is possibility of production defects occurring. Apart from that, metal objects can also experience defects during the operation of the object, for example due to impact, fracture, or due to corrosion or other chemical reactions. In the nuclear industry, metal objects include reactor tank, control rod, pump, pipes, and nuclear reactor fuel cladding. In order to ensure the continued optimal functionality of equipment, machinery, and/or installations constructed from metal, it is essential to implement a comprehensive maintenance programme. Such a programme must include regular inspections to identify any potential damage that may affect the integrity of the metal components. One method of detecting defects in technical objects that is currently being developed is using artificial intelligence or machine learning, either combined with visual inspection and analysis methods, X-ray, gamma ray, electromagnetic or ultrasonic results. Machine learning itself has been applied to several applications. The research carried out in this article used bibliographic analysis and was visualized using VOSviewer. This paper presents methodologies for the non-destructive analysis of metal objects, with a specific focus on visual analysis and machine learning. The integration of test data and visual analysis with machine learning enables the formulation of informed conclusions regarding the condition of metal objects, thereby facilitating decision-making processes.