Integrated design under uncertainty for pharmaceutical processes
in pharmaceutical process development there is frequently a large element of process uncertainty since knowledge of the mechanisms of production and separation is often limited. The overall objective of this thesis is the development of a general methodology which combines process modelling with uncertainty techniques to support the process development of complete integrated sequences. In a structured approach the uncertainty can be managed and improved process performance may be obtained. The major concept of this work is the integration of stochastic methods into a general framework for batch and continuous process models, consisting of two main parts. The first combines systematic modelling procedures with Hammersley sampling based Uncertainty Analysis and a range of sample-based Sensitivity Analysis techniques, used to quantify predicted performance uncertainty and identify key uncertainty contributions. In the second, a stochastic optimisation approach is employed to solve different problems under uncertainty. The methodology was implemented on two case studies. The first study investigated a batch reactor process. Some undesirable performance characteristics were observed when the published nominal optimal isothermal operating policy was implemented in the uncertain system. It was found that a robust operating policy significantly improved the total process time characteristic but not the impurity content and an alternative non-isothermal policy strategy would be a better option. The second study investigated a complete process sequence. As models developed with incoming data, uncertainty in the reaction and crystallisation parameters were critical to the endpoint quality criteria. Expected performance was improved by considering the propagation of uncertainty in the complete process. Four different flowsheets were compared, considering profitability and control tolerance criteria under uncertainty. The case study results indicate the importance in considering uncertainty systematically and quantitatively when conventional modelling techniques are employed. The methodology showed the opportunity to improve process performance potential and provide more realistic information to support pharmaceutical process development.