Global optimization techniques using powerful algorithms have led to a wide range of applications to increase the efficiency of chemical processes. Nevertheless, the performance for optimization of process models is limited by a certain complexity, especially accounting for existing processes (retrofit). Due to the great combinatorial diversity of possible alternatives a systematic approach is essential. The local integration of modifications in the overall process leads to changes in internal streams. Therefore new operating points have to be found and resulting effects on the plant performance have to be evaluated. An optimization framework for the purpose of retrofitting using a rigorous process simulation tool is proposed to fulfill this task. Here, flowsheet simulation software packages are offering a high performance for the prediction of new operation points for following units, and for units affected by recycle streams. An optimization approach using flowsheet simulation software and the stochastic optimization algorithm Molecular-Inspired Parallel Tempering (MIPT) implemented in the programming software Matlab (TM) is presented. Both of these programs are linked via OPC (OLE for process control), a standard communication platform. The toolbox provides a quick evaluation of the process by searching for the global optimum. The MIPT algorithm is suitable for large optimization problems and can handle constraints and infeasibilities. The usage of a rigorous process simulator is providing a high accuracy of the thermodynamic results which is necessary to evaluate the influence of the new process design. Furthermore, a simulation model of the industrial plant can directly be used for the optimization. A complex multicomponent separation process with recycle streams is used to demonstrate the advantages of the proposed toolbox. To simplify the user input a graphical user interface was programmed. The results of a sensitivity analysis and the optimization for different feed compositions are presented. (C) 2015 Elsevier Ltd. All rights reserved.