AI-Based Predictive Maintenance Framework for MEP Systems in Smart Buildings under Vision 2030

Smart buildings in Saudi Arabia are moving from conventional automation toward adaptive infrastructure that can protect asset reliability, occupant wellbeing and energy performance at the same time. This review develops an AI-based predictive maintenance framework for mechanical, electrical and plumbing (MEP) systems in smart buildings under Vision 2030. The review synthesises literature published between 2020 and 2025 on artificial intelligence, fault detection and diagnosis, digital twins, building management systems, machine learning for HVAC and water systems, and facility management governance. It responds to a practical gap: many buildings collect large volumes of operational data, yet maintenance remains calendar-based, alarm-driven or dependent on individual technician experience. The proposed framework connects sensor strategy, data architecture, AI analytics, risk scoring, work-order governance and continuous commissioning into one lifecycle model. It treats predictive maintenance as a decision system rather than a single algorithm. The contribution is a Saudi-relevant review framework that links AI maintenance decisions to Vision 2030 outcomes: lower energy intensity, fewer unplanned failures, improved indoor environmental quality, longer equipment life, resilience for high-occupancy assets, and measurable support for national sustainability targets. The paper also identifies implementation risks, including data quality, cybersecurity, vendor lock-in, model drift, false alarms, workforce readiness and the difficulty of translating predictions into accountable maintenance action. The review concludes that AI-based predictive maintenance can become a core operational capability for Saudi smart buildings when it is embedded within IBMS governance, verified by commissioning evidence and aligned with business-value metrics.