(710e) Knowledge Management in Pharmaceutical Manufacturing | AIChE

(710e) Knowledge Management in Pharmaceutical Manufacturing

Authors 

Joglekar, G. - Presenter, Purdue University
Giridhar, A., Purdue University
Reklaitis, G. V., Purdue University



A large quantity of data, information, and knowledge is generated and accessed in pharmaceutical manufacturing. Over the entire life cycle of a product, from molecule discovery to formulation studies to scale-up and production, much knowledge is generated by experimentalists, mathematical modelers and practitioners. Organizing this knowledge and making it available to support decisions in a scalable way has traditionally been a challenge, due to three aspects of the knowledge to be shared: its large quantity, its high complexity, differences in the worldview of different groups and an inherently hierarchical structure of its sharing patterns. The quantity of knowledge requires solutions that scale well; its high complexity (e.g. an executable mathematical model as a single entity, as opposed to a set of equations with unstated assumptions) precludes naive database implementations; differences in worldviews require the development of a common vocabulary for describing basic building blocks and the hierarchical structure of usage patterns should ideally be recognized and incorporated.

Defining a workflow to be a pattern of knowledge-sharing, we present a framework and implementation centered around managing workflows in pharmaceutical manufacturing activities. A core idea is that there are naturally-arising workflows: e.g., from someone generating experimental data to someone building a mathematical model who uses that data to fit parameters; from a model-builder to an engineer using the model in scale-up or production decisions; from a statistician writing a DOE to a lab tech carrying out the experiment for all the points in the DOE; and so forth. Some workflows, such as most lab studies, are specifically commissioned, while others, such as using experiment results to build models, are equally driven by availability of data. In this work we describe our workflow management system and its implementation, and present case studies from tablet manufacturing and drop-on-demand printing.