Модель визначення подібності між прецедентами для діагностики відмов обладнання суднової енергетичної установки

Modern ship power plants (SPPs) are high-tech systems that involve the continuous collection and analysis of large volumes of diagnostic data. Improving the efficiency of technical operation of such systems requires the development of intelligent diagnostic tools capable of accurate recognition and interpretation of typical failures. This paper addresses the problem of assessing similarity between cases within the framework of Case-Based Reasoning (CBR), which is a relevant approach under conditions of limited training data and high variability of failures. A comprehensive review of current similarity assessment methods is provided, including metric-based approaches, aggregation operators (e.g., Choquet integral), logic-based inference systems, and semantic graph models. The main limitations of these methods are identified, such as low adaptability to heterogeneous data, poor interpretability, and strong dependence on a priori knowledge. The authors propose a unified similarity assessment model that accounts for the heterogeneity of diagnostic features. The model integrates different similarity metrics – Euclidean, cosine, and Jaccard – depending on the type of data (numerical, categorical, or multi-valued). The diagnostic significance of features is represented by weight coefficients, which are optimized using numerical methods. The paper also presents a formal description of the case structure, a classification of typical failures, and an algorithm for matching based on the operational context. The proposed approach ensures interpretability, adaptability, and scalability of the model, confirming its applicability in intelligent decision support systems for failure diagnosis of SPP equipment. Practical implementation of the model can significantly enhance the reliability, safety, and efficiency of marine systems operation