This paper introduces the idea of implementing digital twin for predictive maintenance under open system architecture. Predictive maintenance (PdM) is critical to machines operating under complex working conditions to prevent major and unexpected machine failures and production downtime. A cost and reliability optimized predictive maintenance framework for industry 4.0 machines key parts based on qualitative and quantitative analysis of monitoring data is proposed. Employing machine learning and advanced analytics for data fusion for PdM promises for accurate failure diagnostics and prognostics in addition to optimized maintenance decisions. Furthermore, a cost effective maintenance framework can be implemented under reliability centered maintenance strategy. The qualitative and quantitative analysis will help the decision-making process that leads to accurate predictive maintenance strategies. The proposed method is expected to provide cost-effective maintenance and improved intelligence of the predictive process and the accuracy of predictive results. © 2022 IEEE.