Contemporary manufacturing requires personalized production and fast delivery, so manufacturers must be able to quickly adapt to production line status and stakeholder requirements. To meet these requirements, a manufacturer must be able to quickly analyze data and to act on analysis results. In this paper, we introduce the architecture design and prototype of a smart factory architecture to help workers improve productivity, namely by MAPE/BD: 1) Monitoring data, 2) Analyzing them, 3) formulating Plans based on analysis results, and 4) Executing the plan with the support of Big data technology and Digital Twin. In the first part of this paper, we identify problems in current manufacturing environment, and derive corresponding design considerations. In the second part, we propose architecture development based on design considerations, then develop a prototype of a TO-BE model for comparison with AS IS model in an example operation scenario. We found that TO-BE model contributes to fast response and reducing work proficiency difference. (C) 2018 The Authors. Published by Elsevier B.V.