(259e) Smart Process Analytics and Machine Learning

Data analytics can be invaluable for improving manufacturing operations, with advanced methods expected to become increasingly important as larger and more diverse types of datasets become increasingly used in the chemicals, (bio)pharmaceuticals, and related process industries. Selecting the best method, however, requires a high level of expertise. An automated process data analytics approach is presented which empowers users to focus on objectives rather than on methods. Tools are applied to automatically interrogate the data to ascertain characteristics (e.g., nonlinearity, multicollinearity, dynamics), which are used to select among the best-in-class data analytics method. These ideas are shown to be natural extensions of the early chemical engineering literature in artificial intelligence and expert systems by Venkat Venkatasubramanian and others.