M. Ramasamy

System Identification Using Orthonormal Basis Filters

Models are extensively used in the design and implementation of advanced process control systems. In model predictive control (MPC), model of the plant is used to predict the future output of the plant using the current and future optimal inputs and past outputs. Therefore, the design of MPC, essentially, includes the development of an effective plant model that can be used for predicting the future output of the plant with good accuracy (Camacho & Bordon, 2004; Rawlings, 2000). Models can be developed either from purely theoretical analysis (conservation principles, thermodynamics, etc.) or from experimental data or somewhere in between. The process of model development from experimental data is known as system identification. The identification test can be conducted either in open-loop (open-loop identification) or while the plant is under feedback control (closed-loop identification).

Effect of Bulk Temperature and Heating Regime on Crude Oil Fouling: An Analysis

Semi-empirical threshold fouling models predict higher fouling rates at high surface or film temperatures. Several experimental fouling data reported in literature and from our study were analyzed with respect to increase in surface and bulk temperatures that showed a decrease in fouling rates. The existing threshold fouling models do not adequately describe the phenomenon of decreasing fouling rates with increase in surface or bulk temperatures. The possible causes including the effect of temperature difference, heating regime and solubility of fouling precursors were analyzed and reported.

Effect of Bulk Temperature on Formation of Crude Oil Fouling Precursors on Heat Transfer Surfaces

Temperature plays a very important role in the formation of fouling precursors in crude oils which is considered to be the first step before the precursors are either attached to the wall as a deposit or transferred back to the bulk fluid by diffusion. In order to investigate the formation characteristics of fouling precursors in crude oils at different bulk temperatures, a custom-design thin film microreactor is constructed. It is observed during the experiments that tendency to form fouling precursors is higher at higher surface temperatures. The precursor particles once formed continue to grow in size with time at constant surface temperatures. It is also observed that the particles tend to grow in size while it is cooled when the temperatures are below 55 oC.

Development of Heat Exchanger Fouling Model and Preventive Maintenance Diagnostic Tool

The Crude Preheat Train (CPT) is a set of large heat exchangers which recover the waste heat from product streams back to preheat the crude oil. The overall heat transfer coefficient in these heat exchangers may be significantly reduced due to fouling. One of the major impacts of fouling in CPT operation is the reduced heat transfer efficiency. The objective of this paper is to develop a predictive model using statistical methods which can a priori predict the rate of the fouling and the decrease in heat transfer efficiency in a heat exchanger in a crude preheat train. This predictive model will then be integrated into a preventive maintenance diagnostic tool to plan the cleaning of the heat exchanger to remove the fouling and bring back the heat exchanger efficiency to their peak values. The fouling model was developed using historical plant operating data and is based on Neural Network. Results show that the predictive model is able to predict the shell and tube outlet temperatures with excellent accuracy, where the Root Mean Square Error (RMSE) obtained is less than 1%, correlation coefficient R2 of approximately 0.98 and Correct Directional Change (CDC) values of more than 90%. A preliminary case study shows promising indication that the predictive model may be integrated into a preventive maintenance scheduling for the heat exchanger cleaning.

Iterative closed-loop identification of MIMO systems using ARX-based Leaky Least Mean Square Algorithm

Closed-loop identification of MIMO systems is considered. An iterative Leaky Least Mean Squares (LLMS) algorithm is proposed for the development of ARX structure. The performance of the proposed algorithm with respect to the existing recursive algorithms is investigated in a simulation study. The simulation results show that the proposed algorithm can produce more accurate parameter estimates than the conventional recursive algorithms.

System Identification Based Proxy Model of a Reservoir under Water Injection

Simulation of numerical reservoir models with thousands and millions of grid blocks may consume a significant amount of time and effort, even when high performance processors are used. In cases where the simulation runs are required for sensitivity analysis, dynamic control, and optimization, the act needs to be repeated several times by continuously changing parameters. This makes it even more time-consuming. Currently, proxy models that are based on response surface are being used to lessen the time required for running simulations during sensitivity analysis and optimization. Proxy models are lighter mathematical models that run faster and perform in place of heavier models that require large computations. Nevertheless, to acquire data for modeling and validation and develop the proxy model itself, hundreds of simulation runs are required. In this paper, a system identification based proxy model that requires only a single simulation run and a properly designed excitation signal was proposed and evaluated using a benchmark case study. The results show that, with proper design of excitation signal and proper selection of model structure, system identification based proxy models are found to be practical and efficient alternatives for mimicking the performance of numerical reservoir models. The resulting proxy models have potential applications for dynamic well control and optimization.

Isolation of Interacting Channels in Decentralized Control Systems Using Instrumental Variables Method

This paper describes the use of partial correlation based instrumental variables method for the identification and isolation of weak interaction dynamics between subsystems in decentralized control systems. Unlike the available methods based on the ordinary least square, the proposed method clearly discriminates the interaction channels that have significant contribution to the interconnected subsystem from the ones which do not by reducing the model error that arises due to the process inputs correlation. The efficacy of the proposed method is illustrated through a case study.

Dynamics of Wall Heated Packed Bed Reactors

The multi-tubular packed bed reactors in shell tube heat exchanger configuration are widely used for carrying out exo-/endo-thermic reactions. To design control strategies for stable and safe operation of such reactors, information on dynamic response of such units are needed. Dynamic response of a pilot scale multi-tubular packed bed reactors in shell tube heat exchanger configuration performing dehydrogenation of isopropyl alcohol are presented.

Identification of Multi-Input Multi-Output Systems Using Combined Direct and Indirect Methods

Successful deployment of cooperative decentralized model predicative control needs reasonably accurate subsystem interactions models. Processes in which open-loop tests are not permitted, closed-loop identification of subsystems interactions is crucial. An approach that combines the direct and indirect methods of closed-loop identification is proposed in this paper. It is shown that full dynamics of MIMO systems can be determined following a two-steps identification procedure. A representative case study is used to demonstrate the efficacy of the proposed approach.

A Comparison Study between Integrated OBFARX-NN and OBF-NN for Modeling of Nonlinear Systems in Extended Regions of Operation

In this paper the combination of linear and nonlinear models in parallel for nonlinear system identification is investigated. A residuals-based sequential identification algorithm using parallel integration of linear Orthornormal basis filters-Auto regressive with exogenous input (OBFARX) and a nonlinear neural network (NN) models is developed. The model performance is then compared against previously developed parallel OBF-NN model in a nonlinear CSTR case study in extended regions of operation (i.e. extrapolation capability).