This Is AuburnElectronic Theses and Dissertations

Data Driven Methods for Chemical Process and Product Synthesis and Design




Davis, Sarah E

Type of Degree

PhD Dissertation


Chemical Engineering


Data driven methods for chemical process and product synthesis have become integrated in all aspects of design. The responsibly of the academic community should be to provide users with guidance when managing the ever-increasing amount of data and possible data analytics methods with a goal of utilizing these new design tools to ensure that their applications provides meaningful results. Progressive model improvement will lead us to improve characterization techniques to better describe molecules, more advanced modeling methods provide more correct results, and uncertainty management will ensure that the results are more accurate. The methods presented in this work illustrate applications of data driven methods for chemical process and product synthesis and design with a focus on two specific tools computer aided molecular design and surrogate modeling. Computer Aided Molecular Design is a framework that allows us to utilize data to design molecules specific to a process. This is important because it eliminates the need to alter the design to match the available inputs, rather the inputs are modified to match the design. Once issue with this method is that it is reliant on characteristic data for each molecule or building block. The work presented in this dissertation allows us to generate necessary data to apply to the framework thus expanding the possible molecules that can be utilized even further than the computer aided molecular design framework alone. Surrogate modeling allows us to understand complex or unknown processes to provided understanding of the process and improve designs. The work presented in this dissertation provides information about the application of those models based on the surface shape and number of inputs. Additionally, it provides information about sampling methods and sizing. Basically, this information can help make an informed decision when selecting which surrogate model, sampling method and group for each type of application. Both advances provide added depth to data analysis by enhancing current methodologies. This type of work is important because as the modern chemical engineer begins to implement data driven design techniques, the applications that are utilized will need to become more robust and accurate.