(5l) A Systematic Property Based Approach for Process and Molecular Design

The selection of product/product mixtures that give the optimum performance of a process is a critical issue for a design engineer. The process performance is usually understood in terms of physical properties and on many occasions, the physical properties of the product rather than their chemical structure determine the suitability of a specific product as the input to the process. For example, in the design of a blanket wash solvent, the primary focus of designer are the solubility parameter, flammability, vapor pressure etc. of the solvent. Molecular design algorithms generally require target properties to design the molecules. Therefore, for obtaining the optimal solution for this type of problem, it is necessary to have a methodology to represent the product performance in terms of measurable physical properties and identify the molecule/mixture that gives the property targets corresponding to the optimum process performance.

In spite of the relationship between the process design and product design problems, they have been traditionally considered as two separate problems because the product design part is generally considered as outside the scope of chemical engineering design. Therefore, the product identified by chemists may not provide the best one corresponding to the optimum process performance and this makes the process of identifying the suitable molecule/mixture an iterative process. Therefore, on most occasions, the chemists try to design the product based more on expert knowledge, trial and error, heuristics and experimentation based on intuition rather than any specific scientific reasoning. The attempts to develop a procedure to pass the flow of information between the process and product designer brought about the development of reverse problem formulations

In my PhD work, systematic algorithms have been developed for the design of molecules corresponding to the optimum performance of a process. The concept of property clustering has been extended to the formation of molecular clusters based on second and third order group contribution methods. An algebraic approach has been developed utilizing higher order groups built from first order groups subject to the constraints of overlapping. The advantage of using an algebraic approach is that it can handle any number of molecular groups or properties and can generate all possible compounds within the required range of properties. The most significant aspect of the aforementioned method is that both the application range and reliability of the molecular property clustering technique are considerably increased by incorporating second and third order estimation. A methodology has been developed for incorporating the property contribution predicted using combined group contribution and connectivity indices into the design framework in case the property contributions of any of the molecular groups of interest are not available in literature. For the design of simple monofunctional molecules, a modified visual approach has been used where as for the design of more complicated structures and/or for treating more than three properties at a time, an algebraic method has been used.

The developed algorithm can be used as a tool for the solution of integrated process and product design problems. Since an algebraic approach for the solution of a process design problem is already available, it is now possible to solve the integrated problem irrespective of the number of properties involved. The process design problem can be solved in terms of constitutive variables without committing to any component beforehand. The property targets estimated in the process design step along with the molecular groups will form the input to the molecular design algorithm. The algebraic equations formed using the molecular operators are solved simultaneously to generate the possible molecular structures that match the optimum process performance defined by the property targets.

However, apart from the targets set by the process design, the molecule has to satisfy a number of other environmental and safety constraints in order to be used in an actual process. In addition, there are many properties which can not be estimated through group contribution methods because it is not always possible to find a correlation between the molecular groups and properties. Similarly, not all possible atomic arrangements are represented in group contribution methods. So, there is a need of an efficient methodology for the design of molecules with a more diverse property targets. In my thesis, I have developed an algorithm that uses the concept of molecular signature descriptors for molecular design. The signature is a systematic coding system of atom types and the signature of a molecule can be obtained as a linear combination of its atomic signatures. It has been proved that any of the topological indices of molecules can be represented in terms of molecular signatures and it is possible to correlate the topological indices to the actual properties and biological activities. Here, the new algorithm utilizes molecular property operators based on signatures for solving the reverse problem of obtaining the molecular structures that satisfy the property targets estimated in the process design step. A new set of equations will be employed to ensure that the molecule meets the safety and environmental constraints as well. The principles in graph theory are incorporated to avoid the generation of infeasible structures. Since the molecular operators are formed based on molecular signatures, the property models based on different topological indices can be represented on the same property platform. Since different properties are described using different topological indices, the height of molecular signature required to describe those topological indices will be different. However, techniques have been developed to describe all topological indices with a single signature height. The final solution will be obtained as a set of the biggest signature height used to describe the topological indices. Therefore, the accuracy of prediction will not be compromised because off this transformation. The property models required to describe the target properties may be based on QSPR/QSAR (quantitative structure-property/activity relationship) models or in the form of group contribution methods. In the developed algorithm, the group contribution models can also be represented using the signatures which allow for the simultaneous consideration of all property targets irrespective of whether the targets are based on process design or other constraints. The accuracy of this method depends only on two factors- how well the actual property-topological index relationships are estimated and the height of atomic signatures used to describe the topological indices. Since, many topological indices can be used to describe each property, this algorithm generally provide reliable results.

The aforementioned algorithms provide excellent tools for the design of molecules corresponding to the process requirements. The current research in this area is to extend these algorithms into the design of process flow sheets. The recently developed computer aided flowsheet design techniques will be utilized along with the molecular design techniques to obtain the required process performance.