Reliability-Centered Maintenance (RCM) plays a crucial role in minimizing operational downtime and lifecycle costs in petrochemical industries. However, conventional RCM approaches often lack dynamic failure diagnosis and prioritization capabilities under uncertain operating conditions. This study proposes an enhanced framework integrating Failure Mode and Effects Analysis (FMEA) with data-driven linguistic rule extraction to improve maintenance decision-making for centrifugal pumps. The proposed methodology utilizes OREDA-based failure classification to identify critical failure modes and introduces a weighted severity–occurrence model to overcome limitations of traditional Risk Priority Number (RPN) ranking. The framework establishes relationships between failure causes and key operational parameters such as flow rate, discharge pressure, vibration, temperature, and efficiency using linguistic variables. A rule-based diagnostic system is developed to enable real-time fault identification and maintenance scheduling. The framework is validated through a case study of centrifugal pumps in a petrochemical aromatic plant. Results demonstrate improved fault detection accuracy, reduced maintenance time, and enhanced system reliability. The proposed approach provides a scalable and intelligent decision-support tool for predictive maintenance and industrial asset management.