In recent years, the Fourth Industrial Revolution has accelerated the adoption of intelligent manufacturing technologies, with predictive maintenance emerging as a key approach for improving equipment reliability and operational efficiency. This study proposes and validates a data-driven predictive maintenance framework for centrifugal pumps using multisensor measurements, including vibration, power consumption, temperature, flow rate, and pressure. The framework integrates data preprocessing, exploratory analysis, and machine-learning based classification to identify abnormal operating patterns and predict maintenance requirements prior to critical failures. To demonstrate practical feasibility, a prototype implementation was developed to support real-time monitoring and decision support. The proposed model achieved strong predictive performance, with 99.97% accuracy, 99.4% F1-score, 99.3% precision, 99.5% recall, and an ROC–AUC of 0.9999, indicating robust failure discrimination. The results highlight the effectiveness of multisensor learning for early fault detection and provide a scalable foundation for intelligent maintenance systems that reduce unplanned downtime and maintenance costs. Overall, this work contributes to the advancement of smart and sustainable industrial operations in the context of Industry 4.0.