Predictive Analytics for Spare Parts Planning in Semiconductor Manufacturing: A Data Engineering Approach to Supply Chain Reliability

Abstract

This review examined how predictive analytics, supported by modern data engineering, can strengthen spare parts planning and supply chain reliability in semiconductor manufacturing. The study adopted a structured narrative review approach, synthesising evidence on intermittent-demand forecasting, predictive maintenance, equipment-failure modelling, feature engineering, inventory optimisation, data architectures, systems integration, governance, and organisational readiness. Particular attention was given to the operational realities of capital-intensive fabrication environments, where proprietary components, uncertain failure patterns, long replenishment lead times, equipment obsolescence, and production bottlenecks create substantial reliability risks. The findings show that effective planning depends on combining sensor streams, maintenance histories, inventory transactions, procurement records, supplier performance, and production priorities within scalable and governed data pipelines. Machine-learning models, survival analysis, anomaly detection, remaining-useful-life estimation, and intermittent-demand methods can provide earlier and more accurate indications of component requirements. However, analytical accuracy alone is insufficient unless predictions are embedded within maintenance, inventory, procurement, and supplier-management systems. The review further identifies poor data quality, weak asset-to-part mapping, model drift, cybersecurity exposure, skills shortages, and fragmented decision ownership as major barriers to implementation. The study concludes that predictive spare parts planning should be treated as an integrated reliability capability rather than a stand-alone analytical initiative. It recommends phased deployment beginning with bottleneck equipment and high-criticality components, standardised master data, confidence-based decision rules, continuous model validation, cross-functional governance, and supplier collaboration. Future progress should prioritise digital twins, uncertainty-aware forecasting, interoperable data platforms, and workforce development to reduce downtime, improve inventory productivity, strengthen resilience, and support more dependable semiconductor operations. These priorities provide a practical foundation for responsive planning across distributed facilities, suppliers, maintenance networks, and markets.