Optimizing set point control of the MCSGP process

We present a case study for iterative set point optimizing control, the control of a ternary chromatographic separation process for which structural and parametric model uncertainty must be taken into account. The process considered here is a continuous switched chromatographic multicolumn separation process with high purity and yield requirements. In each iteration, the optimal cyclic steady state of a system that is governed by partial differential equations is required. This is done efficiently by a new one-shot optimization method, that employs a two-grid Newton-Picard approach. We demonstrate the control algorithm by a simulation study with experimentally determined isotherm parameters under realistic plant-model mismatch conditions. Despite the large model uncertainty, the plant is driven to its true optimum, increasing the productivity of the plant to 99.8% of its maximum while decreasing the solvent consumption by about 37 %.