Abstract:In refining and chemical industries, there is a growing demand for efficient model identification methods in MPC commissioning and maintenance. In this work we will introduce a modern identification technology called Tai-Ji ID. Key components of the method are: 1) automatic and multivariable tests, can be open loop or closed-loop, 2) automatic parametric model estimation and 3) automatic model validation/selection. The advantages of the technology include reduced test time (70%), reduced process disturbance, accurate models and user-friendliness. An application of the technology at Rohm and Haas is reported. The Tai-Ji ID was chosen to do the re-identification for a multivariable controller that has been successfully in operation for over a year. Initially, the existing models were identified using traditional step test data and the step test lasted for many days around the clock. The identification test with Tai-Ji ID for the dynamic models took a sorter duration where all the 8 MVs were moved (stepped) simultaneously. Part of the inferential models was also identified using the multivariable test data. The reduction of test time by over 70% was realized compared to the initial installation. The tests had not effected unit operation. The performance of the controller with the new model is satisfactory.

Simple Control-Relevant Identification Test Methods for A Class of Ill-Conditioned Processes

Some Study on Identification of Ill-Conditioned Processes for Control

Experience-based Identification and Model Predictive Control for a Methanol Recovery Distillation Column

System Identification for Process Control: Recent Experience and Outlook.

Progress in MPC Identification: A Case Study on Totally Closed-Loop Plant Test

Multivariable process identification for mpc: the asymptotic method and its applications

Mimo Process Identification for Controller-Design - Test Signals, Nominal Model and Error-Bounds

New Developments in Industrial MPC Identification

Some notes on MPC relevant identification

MPC Closed-Loop Identification Without Excitation

How Much Can We Increase the Efficiency of MPC Identification?

A method of LPV model identification for control

Model Predictive Control Based On Experience Model For The S-Trioxane Concentration Column

Toward a low cost and high performance MPC: The role of system identification

Technical Communique: Optimal closed-loop identification test design for internal model control

Parametric Versus Nonparametric Models in Mpc Process Identification

Case Studies on Closed-Loop Identification for MPC

Robust PID Tuning Using Closed-Loop Identification

System identification: New developments of the asymptotic theory/method

Comparison of Three Classes of Identification Methods

Multivariable Process Identification Based on Frequency Domain Measures