Machine learning and genetic algorithms in pharmaceutical development and manufacturing processes

We develop an adaptive Automated Intelligent Manufacturing System (AIMS) for Case 1:to a well-understood-pharmaceutical-process to demonstrate our methodology, Case 2:with clustering, to a not-well-controlled or understood-process for seemingly identical experiments yielding disparate results, Case 3:to scale-up a process from development to manufacturing, and Case 4:to deploy AIMS adaptively, to modify the process model and reoptimize the system contemporaneously, when predictive errors are significant. The results showed AIMS had both explanatory and predictive power. We have developed the following methodological extensions: a random probe method for feature selection, a simulation approach to establish tolerances for target inputs, and an adaptive capability integrated with statistical-process-control to modify the model.