Hybrid simulation-equation based synthesis of chemical processes

Abstract

A challenging problem in the synthesis and design of chemical processes consists of dealing with hybrid models involving process simulators and explicit constraints. Some unit operations in modular process simulators are slightly noisy or require large CPU times to converge. In this work, this problem is addressed by combining process simulators and surrogate models. We have replaced some unit operations, which cannot be used directly with a gradient-based optimization, by surrogate models based on Kriging interpolation. To increase the robustness of the resulting optimization model, we perform a degree of freedom analysis and aggregate (or disaggregate) parts of the model to reduce the number of independent variables of the Kriging surrogate models (KSMs). Thus, the final model is composed of KSMs, unit operations (maintained in the process simulator) and also explicit equations. The optimization of the well-known vinyl chloride monomer (VCM) production process is performed to test the proposed approach. The effect of the heat integration is also studied. In addition, the economic feasibility of the optimized process is calculated assuming uncertainty in raw material and product prices.