System and method for controlling an industrial process utilizing process trajectories
AssigneeIntevep, S.A.
InventorsAlexander Zambrano, Manuel Vazquez, Hugo E. Aponte, Alexis A Suarez, Jose Fernandes
Life of this patent
Estimating the term…
Abstract
A method for monitoring an industrial process, including the steps of: providing a knowledge base comprising a two-dimensional representation of operational conditions for the process; obtaining process real time data from the industrial process for a plurality of process variables; selecting relevant features from the real time data; obtaining an operational condition classification from the relevant features; transforming the operational condition classification into a point; and displaying the point on said knowledge base.
Claims14
A method for monitoring an industrial process, comprising the steps of:
- obtaining process real time data from said industrial process for a plurality of process variables;
- selecting relevant features from said real time data;
- providing a diagnosis neural network trained to correlate relevant features with operation condition classifications;
- inputting said relevant features to said diagnosis neural network to obtain said operational condition classification comprising a plurality of process conditions and a corresponding certainty factor for each of said process conditions;
- providing a projection neural network trained to correlate said process conditions and said certainty factor with a knowledge base;
- inputting said process conditions and said certainty factor to said projection neural network to obtain a two dimensional process trajectory data point;
- and
- displaying said point on the knowledge base comprising a two-dimensional process trajectory representation of operational conditions for said process.
A system for monitoring an industrial process, comprising:
- a sensor member for obtaining process real time data from said industrial process for a plurality of process variables;
- a processor unit for receiving said real time data and for selecting relevant features from said real time data;
- a diagnosis neural network trained to correlate said relevant features with operation condition classifications to obtain said operational condition classification comprising a plurality of process conditions and a corresponding certainty factor for each of said process conditions;
- a projection neural network trained to correlate said process conditions and said certainty factor with a knowledge base to obtain a two-dimensional process trajectory data point;
- said knowledge base comprising a two-dimensional process trajectory representation of operational conditions for said process and being associated with said processor unit for displaying said point of said knowledge base.
Dependencies are inferred from the claim text.