Data-driven Process Systems Engineering Lab
Data-driven Process Systems Engineering Lab
2012 AIChE Annual Conference in Pittsburgh
October 29, 2012
FDA harmonization guidelines  aim to promote building quality into design of pharmaceutical processes through mechanistic understanding of powder process behavior. Based on this paradigm, the industry is moving with a fast pace towards a model-based process and product design, which will undoubtedly lead to competitive manufacturing and better product quality . However, process modeling is highly dependent on experimental or computational data, either in the form of empirical correlations or during the stage of parameter estimation. At the same time, there has been a high interest in the population of a flowsheet model library for solids-based pharmaceutical processes. Flowsheet modeling has been an imperative process engineering tool for the design and optimization of liquid-based processes [3-4]. This work aims to summarize all of the identified challenges that originate from the incorporation of experimental and computational data into a dynamic flowsheet model library for pharmaceutical processes, in order to develop a generic model library which may be used for a wide range of raw materials and processes, integrated to form complete production lines. Many challenges arise due to the nature of pharmaceutical materials and processes compared to other industries, such as increased variability of powder material properties and operating conditions, uncharacterized materials, lack of well-established correlations of material properties and process operation and oftentimes lack of predictive models for certain processing steps. However, during the recent years where there has been a strong focus in enhancing process understanding, large amounts of data for the characterization of processes or materials have been collected. In certain cases, these data have provided insight about the governing underlying mechanisms and have lead to the development of first-principle models, while in other cases data have been used to build empirical response surface models, or multivariate latent variable models in cases of high dimensionality. In fact, even in the well-established liquid-based flowsheet models, certain physical phenomena which are insufficiently understood, such as chemical reaction kinetics or heat and mass transfer effects are represented through empirical correlations which are incorporated into first-principle models to create hybrid or semi-empirical modules . Moreover, the complex nature of powder based-processes has also lead to the development and use of high-fidelity spatially distributed simulation models that aim to capture the complex behavior of interacting particles in order to provide insight into detailed critical material attributes such as particle size distributions. This introduces another challenge towards the final goal of flowsheet model building, which is the incorporation of multiscale computational data, derived from very expensive computer models. Fortunately, however, through this work it will be shown that the handling of such computational data is very similar to that of experimental data. Finally it should be mentioned that the link between data and models is always a closed-loop process, since even after the development of the necessary models, the need of experimental data is once again imperative for the estimation of the model parameters. In this case the experimental design will be guided by the models, in order to ensure that they contain the necessary information needed for the parameter estimation. A great opportunity arises in order to efficiently incorporate the information obtained from large data-bases and build and/or refine the flowsheet model library following well-established techniques of the process systems engineering community. This work aims to discuss the state of the art in first-principle models used to simulate pharmaceutical manufacturing and their proposed combinations and modifications in order to maximize the captured information through incorporation of useful data-based correlations. Most importantly, the need to consider each process as part of a large integrated puzzle should be considered even from the early stages of experimental design, since due to the need for simulation of continuous integrated lines, the propagation of a common set of material properties of powder mixtures is essential. The challenges and proposed remedies will be shown through specific case studies for a variety of manufacturing scenarios of pharmaceutical tablets such as direct compaction, dry and wet granulation. References: 1. Q8 Pharmaceutical Development. 2006; Available from:http://www.fda.gov/downloads/RegulatoryInformation/Guidances/ucm128029.pdf. 2. Oksanen, C.A. and S. García Muñoz, Process modeling and control in drug development and manufacturing. Computers and Chemical Engineering, 2010. 34(7): p. 1007-1008. 3. Boukouvala, F., et al., An integrated approach for dynamic flowsheet modeling and sensitivity analysis of a continuous tablet manufacturing process. Computers and Chemical Engineering, In Press, 2012. 4. Werther, J., C. Reimers, and G. Gruhn, Design specifications in the flowsheet simulation of complex solids processes. Powder Technology, 2009. 191(3): p. 260-271. 5. Kahrs, O. and W. Marquardt, The validity domain of hybrid models and its application in process optimization. Chemical Engineering and Processing: Process Intensification, 2007. 46(11): p. 1054-1066.