Development of an automated reactor selection and design procedure in the pharmaceutical industry
The drug market has become very competitive with the development of existing companies and the emergence of new companies. The emergence on the market of so-called generic drugs makes the speed of development of new drugs a prime target. The pharmaceutical industry has until now focused on speeding-up drug development by focusing on the chemistry aspects with very little attention given to potential engineering solutions. When looking at the engineering side of the pharmaceutical industry, and more particularly to the design side of equipment like chemical reactors, the pharmaceutical industry has restricted itself to using stirred tank reactors to analyse, develop and produce products. This type of reactor is perceived as flexible and easy to operate for batch operations. Chemical engineering research has opened up the route for new potential reactor designs, which can offer effective solutions to the problems of reaction scale-up. This work aims to develop an automated comparison of different types of chemical reactors in accordance with their ability to scale-up bulk pharmaceutical products. It looks at the engineering side of drug scale-up in order to provide a reliable and fast evaluation procedure for the selection of a reactor which facilitates reaction scale-up. In this piece of research, indicators of success in the scale-up and assessment of quality of process design is centred around yields (or selectivity), capital expenditures and costs and regulatory compliance. Due to the complexity of the reactions i.e. the number of components or the number of steps involved in the synthesis of a compound, side reactions occur undermining the production of a targeted compound. This problem can be solved in some cases by changing the chemistry of the synthesis of the product. This procedure tends to be very time consuming as each possible route has to be first developed at bench scale. Then each promising route is scaled-up over increasing sizes of stirred tank reactor. The different engineering factors involved within the reaction are commonly neglected, so is the possibility of using different types of reactor. In this study, the level of selectivity achieved, i.e. how much product is generated for the amount of reactant used, is simulated by mathematical models. Selectivity is then used to evaluate profit generated or cost (capital or running) endorsed and reflect on the efficiency of the process design. Regulatory compliance is also looked at through the impact that parameter uncertainty can have on the process design. The influence of mixing on product selectivity is shown in this work as being critical. A comparison of the mixing efficiency of various types of reactors is studied through the influence of mixing on selectivity. Mass transfer is also of prime importance in organic synthesis with the emergence of catalysts. This aspect has also been given particular attention in this work, and various types of reactor have been evaluated against their potential to scale-up a mass transfer sensitive reaction. The comparison of the different types of chemical reactor is carried out through the use of an optimisation procedure. The optimisation is done over the different variables which represent inherent characteristics of each reactor type e.g. energy dissipation rate, mass transfer achievable in a reactor. Uncertainty analysis is also used through the Fourier Amplitude Sensitivity Test (FAST), a global sensitivity analysis method, to evaluate the impact that parameter uncertainty or parameter tolerance has on the process design. The FAST method is a global sensitivity analysis method which offers the advantage that a limited number of sampling points are necessary to evaluate the parameter sensitivity analysis of a model. The selection is then based on a multi-objective optimisation procedures. This allows us to trade-off uncertainty and annualised costs. The advantages of this methodology are the speed of evaluation, the ease of use as well as the versatility and generic aspect. It permits to evaluate the impact that parameter uncertainty has on each type of reactor as well as the identify the equipment which generate the best profit. The mathematical model is able to incorporate dynamic behaviour with out having to resort to the level of detail required for computational fluid dynamics (CFD) and have been implemented in gPROMS.