In modern industry, uninterrupted production is essential for maintaining competitiveness and operational efficiency. To ensure continuity, industries rely on advanced diagnostic approaches supported by performance indicators such as reliability and risk assessment. With the emergence of Industry 4.0, smart devices and Internet of Things (IoT) technologies further enhance monitoring capabilities and enable intelligent optimization of industrial processes. In this context, this study addresses the challenge of detecting combined faults in three-phase induction machines using vibration signals. The main objective is to identify fault signatures at an early stage, even when multiple defects interact simultaneously. To achieve this, an intra-mode variational mode decomposition (IM-VMD) method incorporating a source separation strategy is applied to decompose vibration signals into intrinsic modes. This approach allows effective isolation of fault-related components and improves signal interpretability in both time and frequency domains. The results demonstrate accurate identification of fault signatures, with extracted frequency components closely matching theoretical predictions. A high correlation coefficient ranging from 0.95 to 1 confirms the robustness of the method, highlighting its potential for reliable early fault detection in industrial applications.