DEVELOPMENT OF PREDICTOR MODELS.
Abstract
The typical control loop is installed to adjust the process inputs in such a manner that the process remains close to its desired operating point in face of unmeasured disturbance inputs. Box and Jenkins presented a technique for developing a control strategy based on a stochastic predictor model for the disturbances and a deterministic process model from which the process inputs to give a desired output could be calculated. This report considers the model development phase of Box and Jenkins' procedure, although their entire technique for control system design will be reviewed for the reader's cconvenience. The model development simply consists of determining the parameters in the model that give the best fit of the experimental data. This requires a multidimensional search strategy, several of which are compared in this work. It is essential to be able to implement the resulting procedure on present-day process control computers. Some practical considerations of the type of input and amount of data required are also discussed. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Jun 01, 1969
- Accession Number
- AD0688797
Entities
People
- Brian L. Ramaker
- Cecil L. Smith
- Paul W. Murrill
Organizations
- Louisiana State University