Process control using on-line instrumentation and process models
AssigneeIneos USA LLC
InventorsRamaswamy Vaidyanathan, Ronald S. Hurlbut, William D. Stephens, David R. Van Hare
Life of this patent
Estimating the term…
Abstract
A method for providing improved estimates of properties of a chemical manufacturing process is disclosed. The method employs a process model that includes, or is modified by, scores or other gains obtained from the mathematical transformation of data obtained from an on-line analyzer. Chemical manufacturing processes using the method also are disclosed.
Claims58
A method for obtaining an improved estimate of a property useful in a chemical manufacturing process employing an on-line analyzer, the method comprising the steps of:
- mathematically transforming data obtained from an on-line analyzer to obtain scores correlative to the property;
- and
- using the scores in connection with a process model to generate an estimate of the property.
A chemical manufacturing process comprising the steps of:
- using an on-line analyzer to collect data related to a chemical material at a point within the manufacturing process;
- mathematically transforming the data to produce scores related to the chemical material;
- using the scores in connection with a process model to produce an estimate of a property of interest for the chemical material;
- and
- inputting the estimated property into a controller used in the chemical manufacturing process.
A method for obtaining an improved estimate of a product or process property useful in a chemical manufacturing process employing a chemical reactor and an on-line analyzer, the method comprising the steps of:
- mathematically transforming data obtained from an on-line analyzer to obtain gains selected from the group consisting of coefficients from multivariate curve fits and weights or outputs from neural network analysis correlative to the property;
- and
- using the selected gains in connection with a process model to generate an estimate of the property.
Dependencies are inferred from the claim text.