This research develops an integrated technological system combining multicriteria decision-making (MCDM) methods and machine learning (ML) techniques to support industrial maintenance in Industry 4.0 and digital transformation. The system supports strategic asset maintenance decisions, and its application has been validated at the Brazilian Mint (Casa da Moeda do Brasil).
A structured case study was conducted using qualitative and quantitative data from interviews, focus groups, direct observations and document analysis. Guided by an Operations–Maintenance fit lens, MCDM methods (AHP-AIP, MOORA, MULTIMOORA and Borda Count) and forecasting techniques (ARIMA, ANN) were implemented in a Python-based decision support system (DSS) to prioritize assets for predictive maintenance and forecast corrective maintenance needs.
The integration of MCDM and ML improved decision-making in asset maintenance by addressing complex data challenges. The developed system has enhanced the accuracy of maintenance decisions and is currently operational at the Brazilian Mint. The fit analysis shows that the DSS delivers the highest value under contingencies, clarifying when and where the tool is transferable beyond this specific organization.
This study contributes to the Maintenance 4.0 literature by (1) extending Contingency Theory by formalizing how digital maintenance DSSs depend on the alignment among asset criticality, information-processing requirements and analytical capabilities; (2) introducing an operations–maintenance fit perspective that explains how machine learning forecasting and MCDM jointly support the reconciliation of production throughput objectives with reliability and availability goals and (3) presenting a transparent, transferable and benchmark-ready DSS architecture, validated within a real industrial environment, that integrates four MCDM methods and two forecasting techniques to assess key performance metrics, such as overall equipment effectiveness, for cross-firm comparison.
