Adaptive Enterprise Inventory Management System Based on Digital Twins and Predictive Analytics

Abstract

The article focuses on the development of an Adaptive Inventory Management System (AIMS) for replacement parts of complex technical systems at the operational stage. The objective of the study is to develop a model for managing inventory of replacement components that will improve reliability and reduce costs for maintenance and repair of production equipment. The research is driven by the need to enhance the reliability and cost-effectiveness of maintaining production means and objects amidst growing demands for their uninterrupted operation. The scientific novelty lies in the synthesis of a three-level architecture that integrates product digital twins, predictive analytics, and hybrid machine learning models (LSTM, Survival Analysis) for the dynamic calculation of optimal inventory levels. A key differentiator of the proposed methodology is the implementation of dynamic reservation based on supply disruption risk assessment and failure prediction with 81 % accuracy, utilizing the Apriori algorithm to identify repair kit composition. The proposed solution enables a transition from a reactive to a predictive management model. Practical testing shows significant improvement in key metrics: a 25 % reduction in average inventory levels, a 4-fold decrease in response time to failures (to 12 hours), and an 87 % reduction in downtime due to stockouts while maintaining a readiness factor of Kg ≥ 0.98. The implementation of the Adaptive Inventory Management System lays the foundation for increasing the competitiveness of high-tech enterprises by optimizing the resource cycle and minimizing production stoppage risks.