Integration of Enterprise Resource Planning (ERP) with Artificial Intelligence- and Data Analytics-Based Predictive Maintenance: A Systematic Literature Review Using the PRISMA Method
TL;DR
A Systematic Literature Review using the PRISMA protocol to map the current state of ERP integration with AI-based predictive maintenance shows that the dominant AI methods are Long Short-Term Memory and ensemble learning, often combined with Digital Twin technology.
Abstract
The integration of Enterprise Resource Planning (ERP) systems with Artificial Intelligence (AI) and data analytics-based predictive maintenance is a pillar of industrial digital transformation in the industry 4.0 and 5.0 era. However, a comprehensive understanding of integration patterns, AI methods, benefits, and challenges remains scattered across the literature. This study conducts a Systematic Literature Review (SLR) using the PRISMA protocol to map the current state of ERP integration with AI-based predictive maintenance. A search was performed on the Scopus database, yielding 59 initial articles that were reduced to 10 final open-access articles published between 2024 and 2026. The results show that the dominant AI methods are Long Short-Term Memory (LSTM) and ensemble learning (XGBoost, Random Forest), often combined with Digital Twin technology. ERP integration patterns range from decision support to full bidirectional integration via OPC-UA and REST API protocols. Key benefits include up to 40% improvement in mean time to failure (MTTF), a 30% reduction in maintenance costs, and a return on investment (ROI) of 42.5%. The main challenges include data quality, legacy system interoperability, cybersecurity, and limited multi-site validation.