First principles models with adjustable parameters alongside parameterised neural networks are applied in a commercial package to a refinery product train. This method offers simplicity of implementation as compared to the more conventional real time first principles models and the entirely machine learning approaches. The implementation is at a commercial refinery and organizational, implementation and design details are discussed.

Moderator:
Rob Confair, Marathon Petroleum Corporation

Speakers:
Manoj Shouche, Aspen Technology, Inc.
Abhay Singh, Motiva Enterprises LLC