2026-08-182026-08-180026-07-24PIANO, Felipe Silveira. Análise de atributos de sinais de vibração para o diagnóstico de falhas em rolamentos utilizando PCA e clustering hierárquico. Orientador: Rafael Suzuki Bayma. 2026. 93 f. Trabalho de Curso (Bacharelado em Engenharia Elétrica) – Faculdade de Engenharia Elétrica, Campus Universitário de Tucuruí, Universidade Federal do Pará, Tucuruí, 2026. Disponível em: https://bdm.ufpa.br/handle/prefix/9961. Acesso em:.https://bdm.ufpa.br/handle/prefix/9961In industrial settings, achieving high equipment availability and reliability is essential to ensure operational continuity. Bearings are among the most common components in the rotating machinery of industrial plants. Consequently, monitoring the condition of these components has gained significant importance in recent decades, with predictive maintenance increasingly employing sophisticated techniques for bearing fault detection and diagnosis. Modern approaches involve analyzing vibration signals and classifying them using machine learning tools. Research in this field has largely focused on supervised learning methods. Therefore, this study aims to address research gaps regarding unsupervised learning approaches for analyzing these faults by employing Principal Component Analysis (PCA) and Hierarchical Clustering. The vibration signals analyzed are drawn from a publicly available dataset provided by Case Western Reserve University (CWRU). Initially, the signals are segmented into windows, and the following time-domain features are extracted for each window: Overall RMS Level, Crest Factor, K-Factor, Kurtosis, and Skewness. Boxplots, histograms, pairplots, and correlation matrices were generated based on these features. Subsequently, the features were standardized, and PCA was applied to minimize ambiguity and reduce data dimensionality. The first two principal components served as new features for Hierarchical Clustering. Finally, a dendrogram and a table tracking the classes present in each identified cluster were produced. The results demonstrate the feasibility of using unsupervised machine learning techniques to analyze features for bearing fault diagnosis.Acesso AbertoDiagnóstico de falhasRolamentosPrincipal component analysisClustering hierárquicoFault diagnosisBearingsHierarchical clusteringCNPQ::ENGENHARIAS::ENGENHARIA ELETRICA::SISTEMAS ELETRICOS DE POTENCIA::MAQUINAS ELETRICAS E DISPOSITIVOS DE POTENCIACNPQ::CIENCIAS EXATAS E DA TERRA::CIENCIA DA COMPUTACAO::METODOLOGIA E TECNICAS DA COMPUTACAOTrabalho de Curso - Graduação - MonografiaAttribution-NonCommercial-NoDerivs 3.0 Brazil