Monitoring, Multi-rate Modeling, and Economic Model Predictive Control of Chemical Processes

In this thesis we develop methods and strategies for the design of advanced data-driven monitoring, control, and optimization algorithms for addressing the problems of handling severe faults and economically operating chemical processes while accounting for the complex process dynamics and intricate variable interactions, including issues such as lack of output measurements, process constraints and system uncertainty. In the initial phase of this research, in an effort to better detect and identify process faults, we propose a novel pattern matching based process monitoring approach. Traditional multivariate statistical processes monitoring (MSPM) techniques like principal component analysis (PCA) and partial least squares (PLS) are notwell-suited inmonitoring non-Gaussian processes because the derivation ofT2 and SPE indices requires the approximate multivariate Gaussian distribution of the process data. In this work, a novel pattern analysis driven dissimilarity approach is developed by integrating multidimensional mutual information (MMI) with independent component analysis (ICA) in order to quantitatively evaluate the statistical dependency between the independent component subspaces of the normal benchmark and monitored data sets. The new MMI based ICA dissimilarity index is derived from the higher-order statistics so that the non-Gaussian process features can be extracted efficiently. Moreover, the moving-window strategy is used to deal with process dynamics. The multidimensional mutual information based ICA dissimilarity method is applied to the Tennessee Eastman Chemical process. The process monitoring results of the proposed Page vii of xxi method are demonstrated to be superior to those of the regular PCA, PCA dissimilarity, regular ICA and angle based ICA dissimilarity approaches. Next, we address the problemof the unavailability of reliable andcomputationallymanageable first-principles-based models by developing a data-based multi-rate modeling and control approach. To this end, we consider the problem of multi-rate modeling and economic model predictive control (EMPC) of electric arc furnaces (EAF), which are widely used in the steel industry to produce molten steel from scrap metal. The two main challenges that we address are the multi-rate nature of the measurement availability, and the requirement to achieve final product of a desired characteristic, while minimizing the operation cost. To this end, multi-rate models are identified that include predictions for both the infrequently and frequently measured process variables. The models comprise local linear models and an appropriate weighting scheme to capture the nonlinear nature of the EAF. The resulting model is integrated into a two-tiered predictive controller that enables the target end-point to be achieved while minimizing the associated cost. The EMPC is implemented on the EAF process and the closed-loop simulation results subject to the limited availability of process measurements and noise illustrate the improvement in economic performance over existing trajectory-tracking approaches. Finally, we consider the problem of variable duration economic model predictive control (EMPC) of batch processes subject to multi-rate and missing data. To this end, we first generalize a recently developed subspace-based model identification approach for batch processes to handle multi-rate and missing data by utilizing the incremental singular value decomposition technique. Exploiting the fact that the proposed identification approach is capable of handling inconsistent batch lengths, the resulting dynamic model is integrated into a tiered EMPC formulation that optimizes process economics (including batch duration). Simulation case studies involving application to the energy intensive electric arc furnace Page viii of xxi process demonstrate the efficacy of the proposed approach compared to a traditional trajectory tracking approach subject to the limited availability of process measurements, missing data, measurement noise and constraints.