Ravindra D. Gudi

Identification of complex nonlinear processes based on fuzzy decomposition of the steady state space

Abstract A methodology for identification and control of complex nonlinear plants using multi-model approach is presented in this paper. The proposed methodology is based on fuzzy decomposition of the steady state map. It is shown that such a decomposition strategy facilitates the design of input perturbation signals and helps in identifying linear or simple nonlinear models for each local region. A composition strategy to aggregate the local model predictions is proposed and shown to give excellent cross validation as well as to facilitate smooth switching between the local models. A novel control scheme that is based on the multi model strategy is proposed. The practicality of the identification and control scheme presented here is demonstrated by application to the continuous fermenter of Henson and Seborg (M.A. Henson, D.E. Seborg, Nonlinear control strategies for continuous fermenter, in: Proceedings of 1990 American Control Conference, San Diego, 1990), which exhibits severe nonlinearities and gain directionality changes.

Fault detection and isolation using correspondence analysis

In this paper, a new approach to fault detection and diagnosis that is based on correspondence analysis (CA) is proposed. CA is a powerful multivariate technique based on the generalized singular value decomposition. The merits of using CA lie in its ability to depict rows as well as columns as points in the dual lower dimensional vector space. CA has been shown to capture association between various features and events quite effectively. The key strengths of CA, for fault detection and diagnosis, are validated on data involving simulations as well as experimental data obtained from a laboratory-scale setup.

Optimal control strategies for simultaneous saccharification and fermentation of starch

Abstract Design and analysis of optimal control strategies for three types of inhibitory fed-batch bioprocesses have been discussed. These are simple saccharification (SS) of starch to glucose, simple fermentation (SF) of derived glucose to lactic acid (LA) and simultaneous saccharification and fermentation (SSF) of starch to LA. Various optimal feeding strategies have been investigated for the SSF process by manipulating starch addition rates. To avoid the complexity of solving a singular problem, the starch addition rates are expressed in terms of the broth volume, which is used as a control variable. The optimization strategy is thus solved in a nonsingular framework. Experimental studies carried out using the results of the optimization demonstrated the accuracy and utility of the approach. An increase of 20% in lactate productivity was obtained by operating the SSF process in a fed-batch mode. The focus of all the optimization studies has been to improve the performance of the SSF process. Optimal control of starch additions in the fed-batch process gave improved performance of the SSF process.

Monitoring Batch Processes Using Multivariate Statistical Tools: Extensions and Practical Issues

Abstract Extensions and practical issues in the application of a statistical technique, namely Partial Least Squares (PLS), to the monitoring, product quality prediction and fault detection of batch/semibatch processes is considered. The approach of Nomikos and MacGregor (1994a, 1994b) is explored further to include : (1) multirate sampling and (2) normal batches with varying run lengths. The theoretical development presented here is illustrated using simulated data from a fed-batch bioreactor.

Design of robust, reliable sensor networks using constraint programming

Abstract Various criteria have been considered in the literature for selection of optimal sensor networks. Amongst these, maximization of network reliability is an important criterion. While there are several approaches for designing maximum reliability networks, uncertainty in the available sensor reliability data has not been considered in these designs. In this article we present two novel formulations that incorporate robustness to uncertainties in the reliability data. Towards this end the sensor network design problem for maximizing reliability is formulated as explicit-optimization (MINLP) problem using failure rates of sensors which have better scaling properties instead of sensor reliabilities. Constraint programming (CP) has been used for solving the resulting optimization problems. Use of CP also enables easy generation of pareto front characterizing trade-offs between performance, cost and robustness for various uncertainty scenarios. The utility of the proposed approach is demonstrated on a case study taken from the literature.

Productivity Improvement in Xanthan Gum Fermentation Using Multiple Substrate Optimization

A novel and more comprehensive formulation of the optimal control problem that reflects the operational requirements of a typical industrial fermentation has been proposed in this work. This formulation has been applied to a fed‐batch bioreactor with three control variables, i.e., feed rates of carbon source, nitrogen source, and an oxygen source, to result in a 148.7% increase in product formation. Xanthan gum production using Xanthomonas campestris has been used as the model system for this optimization study, and the liquid‐phase oxygen supply strategy has been used to supply oxygen to the fermentation. The formulated optimization problem has several constraints associated with it due to the nature of the system. A robust stochastic technique, differential evolution, has been used to solve this challenging optimization problem. The infinite dimensional optimization problem has been approximated to a finite dimensional one by control vector parametrization. The state constraints that are path constraints have been addressed by using penalty functions and by integrating them over the total duration to ensure a feasible solution. End point constraints on final working volume of the reactor and on the final residual concentrations of carbon and nitrogen sources have been included in the problem formulation. Further, the toxicity of the oxygen source, H2O2, has been addressed by imposing a constraint on its maximum usable concentration. In addition, the initial volume of the bioreactor contents and feed concentrations have been handled as decision variables, which has enabled a well‐grounded choice for their values from the optimization procedure; adhoc values are normally used in the industry. All results obtained by simulation have been validated experimentally with good agreements between experimental and simulated values.

Improved monitoring and discrimination of batch processes using correspondence analysis

Correspondence analysis (CA) has recently been proposed as a superior alternative to PCA for the tasks of monitoring and early event detection [3]. Some of the key merits of CA include improved discrimination due to the inherent nonlinear scaling as well as the ability to capture and elegantly represent the joint variable-sample associations, which make it amenable to accommodate serial correlations in the data. In this paper, we propose an extension of the CA algorithm to address some of the key problems associated with monitoring of batch processes. The task of batch process supervision is fraught with problems of serial nonlinear correlations in addition to the need to accommodate variable run lengths of the batches. It is shown in the paper that the differential weighting of points in CA offers a unique advantage of representation of the row (sample) and column (variable) points on the same plot (“Bi-plot”), which reveals the prevailing association structure between them. This joint display of observations and variables facilitates the fault diagnosis step and helps to discriminate the batches based on their performance. The paper also uses the traditional DTW for batch synchronization and a simple Euclidean distance based future data filling method during on-line monitoring. The efficacy of the proposed offline and online monitoring method is validated through a simulation study of a penicillin fed-batch fermentation process.

Nonlinear internal model control of PEM fuel cell

Abstract Proton exchange membrane fuel cells (PEMFCs) are known to exhibit strongly nonlinear dynamics and input multiplicities. This paper presents an approach for identifying GOBF-Wiener models towards representing the nonlinear dynamics. The identified model is used to synthesis a MIMO nonlinear internal model controller (NIMC). The average power density and solid temperature of the PEMFC are controlled using (a)cell voltage and inlet coolant temperatures (b) inlet molar flow rates of hydrogen and coolant. The proposed NIMC scheme is able to operate the PEMFC at the optimum power density point where the steady state gain reduces to zero and changes its sign.

Identification for Decentralized MPC

Abstract This paper addresses the problem of identifying interaction dynamics that exist between units operating in a decentralized control scheme. Identification of such interaction relationships is crucial to the deployment of coordinated decentralized control schemes. The proposed methodology is based on (i) the use of partial correlation analysis to identify the interacting channels in closed loop, and (ii) closed loop identification of the concerned dynamics using data obtained from suitable dithering. Alternative identification schemes relevant for this scenario are briefly analyzed in this paper and validation studies on a representative system is presented.

Advanced State Estimation Techniques for Packed Bed Reactors

This work focuses on the evaluation of some of the advanced recursive state estimation techniques applied to packed bed reactors. Tubular, packed or fixed bed reactors are typically modeled as infinite dimensional partial differential equations, also called Distributed Parameter Systems (DPS). Discretization methods such as orthogonal collocation or any other polynomial approximation can be used to convert these DPS into a set of Ordinary Differential Equations (ODE) or Differential Algebraic Equations (DAE), that are more suitable for application of the estimation algorithms. Here we demonstrate the superior performance of the Ensemble Kalman Filter over its other variants (EKF and UKF) for a tubular reactor model. This work also includes the evaluation of the constrained forms of these algorithms for the same tubular reactor model and the results show that the Constrained EnKF which is a nondeterministic approach is able to estimate the states accurately and is less susceptible to numerical errors unlike the Constrained UKF and Constrained EKF which encounter the problem of the covariance matrix becoming negative definite.