MODEL-BASED RUN-TO-RUN OPTIMIZATION FOR PROCESS DEVELOPMENT

Research and development of new processes is a fundamental part of any innovative industry. For process engineers, finding optimal operating conditions for new processes from the early stages is a main issue, since it improves economic viability, helps others areas of R&D by avoiding product bottlenecks and shortens the time-to-market period. Model-based optimization strategies are helpful in doing so, but imperfect models with parametric or structural errors can lead to suboptimal operating conditions. In this work, a methodology that uses probabilistic tendency models that are constantly updated through experimental feedback is proposed in order to rapidly and efficiently find improved operating conditions. Characterization of the uncertainty is used to make safe predictions even with scarce data, which is typical in this early stage of process development. The methodology is tested with an example from the traditional innovative pharmaceutical industry.