Dynamic model generation for implementing hybrid linear/non-linear controller

US9170572B2 · kind B2 · utility

Assignee

• HONEYWELL INTERNATIONAL INC. · US

Inventors

• Ward MacArthur · Scottsdale, US
• Sriram Hallihole · Kanchinakote, IN
• Ranganathan Srinivasan · San Jose, US
• Madhukar Madhavamurthy Gundappa · Savanuru, IN
• Mandar Vartak · Bengaluru, IN
• Gobinath Pandurangan · Savanuru, IN
• S. Chandrakanth Vittal · Bengaluru, IN
• Lucy Liu · Chandler, US
• Sanjay K. Dave · Savanuru, IN
• Avijit Das · Savanuru, IN
• Sreesathya Sathyabhama Sreekantan · Poonamallee, IN
• Roshan Yohannan · Savanuru, IN
• Rajni Jain · Savanuru, IN

Key dates

• Filing date: Jul 6, 2011
• Grant date: Oct 27, 2015
• Expiry date: Dec 5, 2032

Classification

• Technology area (CPC G): Physics
• CPC primary: G06F30/13
• WIPO field: Computer technology
• WIPO sector: Electrical engineering

Abstract

A method of dynamic model selection for hybrid linear/non-linear process control includes developing a plurality of process models including at least one linear process model and at least one non-linear process model from inputs including dynamic process data from a processing system that runs a physical process. At least two of the plurality of process models are selected based on a performance comparison based on at least one metric, wherein the selected process models number less than a number of the plurality of process models received. A multi-model controller is generated that includes the selected process models. The physical process is simulated using the multi-model controller by applying the selected process models to obtain closed loop performance test data for each of the selected models. The performance test data is compared. A selected process model is then selected.