In-silico process design and evaluation tool for pharmaceutical manufacturing

A generic dynamic model library was developed and integrated with a data driven Radial Basis Functions (RBF) control library for the support and guidance for developing and studying the robustness of new upstream pharmaceutical processes. This library was developed by collecting and developing models within the state of the art Process System Engineering (PSE) field, combining multi-dimensional modelling, tools for the integration of different process units, methods to study the required Process Analytical Technology (PAT) systems to achieve pre-determined critical attributes as well as methods of performing upstream error propagation through the application of uncertainty and sensitivity analysis.

The collected and implemented tools allow the study and simulation of new or existent pharmaceutical processes from the beginning of the reaction chain (upstream) to the isolation of the desired Active Pharmaceutical Ingredient (API). The dynamic simulation and study of the integrated process units allows the identification and evaluation of the impacts of bottleneck processes and potential risks for the chosen critical attributes in early stages of process development.

With the implementation of uncertainty and sensitivity analysis, process parameters and physical attributes can be ranked accordingly to the degree of impact they pose on the current system, leading to future decisions of control or further laboratory investigation. Furthermore, with the application of the RBF control methodologies the risk quantification for the failure/success in achieving the critical attributes can be identified and newly developed processes can operate robustly against upstream uncertainties.

The application of the model library was focused on the production of Ibuprofen through the industrial Hoechst process and can be divided in three sections: i) Identification, collection and integration of the different upstream unit processes for the development of the Ibuprofen API and study the dynamics and production metrics of the upstream plantwide unit until the isolation step, crystallization: ii) identification of the uncertain physical attributes and process parameters and quantification of the impact they pose downstream their process unit: iii) development and application of the proposed RBF methodologies for soft sensor control and for predictive control under process uncertainty and with no known disturbances to the control system.

The application of the RBF methodologies has been further validated with experimental work performed in Ibuprofen crystallization with disturbances in the initial conditions of each cooling batch. The results of this study facilitate significantly the application of model-based and data-based engineering tools to develop robust and efficient pharmaceutical manufacturing processes.