Centrifugal pump fault detection and isolation in an early stage is essential from an industry and consumer viewpoint for significantly reducing downtime and maintenance activities., resulting in cost savings. This paper presents a comprehensive framework for centrifugal pump fault diagnosis integrating wavelet coherence analysis and hybrid classification. The proposed framework consists of several modules aiming to attain accurate fault diagnosis. Firstly., the National Instruments-based data acquisition module acquires the vibration signals from the centrifugal pump for healthy baseline and fault-defected signals. A low-pass filter amplifies the fault-specific signal. We perform the Wavelet coherence analysis on the amplified healthy baseline and fault-defected amplified signals to assess the correlation and coherence among different frequency components., resulting in a Wavelet Coherence Visual (WCV) image. Gaussian filtering is applied directly to the WCV., reducing noise and enhancing data quality. Principal Component Analysis (PCA) reduces dimensionality., refining the representation of pump health and fault characteristics. In the classification stage., a tailored hybrid approach., combining t-Distributed Stochastic Neighbor Embedding (t-SNE) and K-Nearest Neighbors (KNN)., significantly enhances the precision of centrifugal pump fault diagnosis and is responsible for categorizing fault classes into specific categories such as healthy., mechanical seal scratch., mechanical seal hole., and impeller defect. Achieving an impressive accuracy of 95.23%., KNN ensures accurate categorization based on health-sensitive WCV features., while t-SNE aids in visualizing data clusters., improving fault interpretability and predictive maintenance reliability. This approach significantly contributes to developing intelligent maintenance strategies., ultimately enhancing the reliability and performance of industrial pump systems.