(341c) Industrial Experience with Model-Based SMB System Design and Optimization: Discretization Scheme and Case Study

Performance of simulated moving bed (SMB) systems can be compromised by poor resin selection, improper design, or poorly chosen operating parameters. Understanding the ultimate performance at scale of an SMB system requires investigation of all three via good experimentation and modeling at the beginning of a process development effort. This report describes an industrial experience with an SMB-based process development project (sugar polyol separation for food applications) outside the pharmaceutical and fine chemical spaces where SMB systems are a more common unit operation.

The project leveraged a custom chromatography model framework suitable for both analyzing experimental data and estimating / optimizing SMB system performance. This framework follows a full discretization approach, using a modified Kurganov-Tadmor scheme for finite domains. This modified scheme retains the conservative properties of the KT scheme, and will be discussed in detail. The scheme also provides solutions to chromatography models that are relatively insensitive to mesh spacing. This allows parameter estimates calculated by regression against lab scale experiments to be transferred directly to SMB simulations with minimal or no correction for differences in grid spacing or time step sizes.

A model that relied on this scheme provided a foundation for both parameter estimation from lab data and simulation of SMB system performance. Simulation-based optimization of operating parameters for a pilot-scale SMB system reduced a key impurity level by 50% vs. the best heuristic-based performance. Simulation of a properly sized full-scale system indicated performance and costs sufficient to justify the business case for the new SMB-based process.