AI-Driven Spare Parts Optimisation for Smart Manufacturing: A Predictive Decision-Support Framework for Saudi Arabian Industries

Abstract

In asset-intensive manufacturing, spare-parts demand is irregular, equipment condition changes over time, and conventional reorder rules may not exploit enterprise and machine data. This article examines how artificial intelligence can connect demand forecasting, asset-health information and inventory decisions within a predictive decision-support framework for manufacturing in Saudi Arabia. A structured narrative review synthesises 29 journal publications from 2020 to 2025 covering spare-parts forecasting, predictive maintenance, digital twins, inventory optimisation, reinforcement learning and explainable artificial intelligence. The evidence is organised by decision problem, data foundation, analytical method, operational output and boundary of application; Saudi policy material provides context, while Project Management Institute guidance informs a separate lifecycle-governance adaptation. The review is not presented as an exhaustive systematic search or meta-analysis. The synthesis supports evaluating forecasts through the inventory decisions they inform and shows how demand estimates and equipment condition information can be connected to replenishment and maintenance decisions. Building on this evidence, the paper proposes a five-layer framework linking industrial data, context and feature engineering, predictive services, constrained inventory optimisation, and governed execution and learning. Proposed governance controls include review checkpoints, assigned decision rights, bounded automation and auditable feedback. The framework is conceptual and has not been validated in Saudi manufacturing plants. It provides a staged pilot pathway from data readiness and criticality segmentation to condition-informed forecasting and closed-loop optimisation, while requiring plant specific assessment of technical compatibility, organisational readiness and economic value.