Inferential Model Predictive Control Using Statistical Tools

With an ever increasing emphasis on reducing costs and improving quality control, the application of advanced process control in the bulk chemical and petrochemical industry is steadily rising. Two major areas of development are model-based control strategies and process sensors…


1 Introduction
1.1 Motivation
1.2 Automatic process control (APC)
1.2.1 Base control systems
1.2.2 Model predictive control (MPC)
1.3 Statistical process control (SPC)
1.3.1 Statistical hypothesis testing
1.3.2 Data pre-processing
1.3.3 Classical SPC and univariate statistical tools
1.3.4 Multivariate statistical tools
1.4 Process measurements
1.4.1 Inferential control using soft-sensors
1.4.2 General guidelines for selecting secondary measurements
1.4.3 Measurement selection using process models
1.5 Optimal control using data-based methods
1.5.1 Data-base generation
1.5.2 Control in the score space
1.6 Research philosophy
1.7 Research goals
1.8 Thesis outline
1.9 Notation
2 Test-beds for advanced process control studies
2.1 The Tennessee Eastman (TE) challenge process
2.2 The Azeotropic (AZ) tower
2.3 Summary
3 Basic controller formulation and performance evaluations
3.1 Basic controller formulation (MP-SPC)
3.2 Analysis of controller performance
3.2.1 Handling of stationary disturbances
3.2.2 Effect of feedback on cross-correlation
3.2.3 Handling of non-stationary disturbances
3.2.4 Sensitivity interpretion of the Lagrange multipliers
3.3 Controller tuning
3.4 Conclusions
4 Database generation, measurement selection, score model development and identi-fication issues
4.1 Database generation and pre-processing
4.2 Measurement selection
4.2.1 First stage
iii4.2.2 Second stage
4.2.3 Results and discussion
4.3 Orthogonal PCA model development
4.3.1 Understanding orthogonal PCA
4.4 Dynamic score model identification
4.4.1 Design of plant tests
4.4.2 Least squares calculations for model identification
4.4.3 Weighting methods
4.4.4 Results and discussion
4.5 Conclusions
5 Alternative controller formulations and extensions
5.1 On-line analyzer cascaded on score controller (MP-SPC + ACSC)
5.2 Influence of disturbance characteristics on controller performance
5.3 Alternative approaches for dealing with non-stationary disturbances
5.3.1 Using steady-state process models (MP-SPC + SS-MPC)
5.3.2 Using tests for signal stationarity (MP-SPC W)
5.4 Conclusions
6 Conclusions
6.1 Summary of Results
6.2 Principal contributions
6.3 Recommendations for Future Work
A Linear algebra
B The Propane-Propylene column (PP)
B.1 Computer code
B.1.1 compile dll.m
B.1.2 C3Splitter.c
B.1.3 BubblePR T y.c
B.1.4 PRparamsPure.c
B.1.5 PRmixParams.c
B.1.6 PRsolvEOS.c
B.1.7 PRfug.c
B.1.8 InnerIterate.c
B.1.9 cubic solve.c

Author: Dave, Kedar Himanshu

Source: University of Maryland

Reference URL 2: Visit Now

Reference URL 1: Visit Now

Leave a Comment