Crystal wash

Second and third extractors

Vacuum still


Vacuum dryer


Vacuum still

Ceramic Ceramic filter filter


Crystalline Potassium Penicillin G

Figure 1.3. Downstream processes in industrial scale penicillin production.

culture volume, sequential growth is followed in this manner. Inoculum size is typically around 10% of the total culture volume. Since formation of secondary metabolites (in this case, penicillin) is usually not associated with cell growth, it is a common practice to grow the cells in a batch culture followed by a fed-batch operation to promote synthesis of the antibiotic. When inoculum at the desired concentration is obtained, an industrial size bioreactor (40,000-200,000 I) is inoculated. The bioreactor is operated for five to six days in fed-batch mode. After the cultivation stage, a series of product recovery techniques are applied depending on the required purity of the final product. Flow diagram for penicillin recovery process is given in Figure 1.3.

A typical time course of penicillin cultivation is represented in Figure 1.4. First, the cells are grown batchwise until they enter early stationary phase of batch growth which is also associated with the depletion of substrate. Then, the process is switched to fed-batch operation that is accompanied by penicillin production. At this stage, process is said to be in the production phase. Experimental data are displayed in [26]. Detailed discussion of various physiological phases is presented in Section 2.7.

The book consists of nine chapters. Chapters 2-5 focus on modeling. Chapter 6 presents a variety of process monitoring techniques. Chapter 7 presents control techniques for batch process operation and Chapter 8 discusses various fault diagnosis paradigms. Chapter 9 outlines recent developments that will impact fermentation process modeling, monitoring,

1.4 Outline of the Book

Glucose And Penicllin Concentrations
Figure 1.4. Time course of changes in carbohydrate (glucose, —), dissolved oxygen (% saturation/10) (a), penicillin (g.L 1 x 10) (*) and biomass (•) concentrations in a penicillin fermentation simulation [61].

and control, and speculates about the future.

Chapter 2 focuses on the development of process models based on first principles. Considering the uncertainty in some reaction and metabolic pathways, and in various parameters, both unstructured and structured kinetic models are discussed. Case studies for penicillin fermentation are presented for both types of models along with simulation results. Chapter 3 presents various concepts and techniques that deal with experimental data collection and pretreatment. Sensors and computer-based data acquisition is discussed first. Then, statistical design of experiments techniques are introduced for preliminary screening experiments. Factorial and fractional factorial designs are summarized and statistical analysis tools are presented for interpretation of results. Data pretreatment issue is divided into outlier detection and data reconciliation, and signal noise reduction. Wavelets are introduced in this section for use in noise reduction. Finally, techniques for theoretical confirmation of data such as stoichiometric balances and thermodynamics of cellular growth are presented to provide a reality check of experimental data. Chapter 4 tackles the modeling problem by focusing on data-based models. First, theoretical foundations in multivariate statistics, such as principal components analysis (PCA), multivariable regression techniques, and functional data analysis, are summarized. Then, various statistical techniques for batch process modeling (multiway PCA, multivariate covariates regression, and three-way techniques) axe presented. Extensions to nonlinear model development are discussed and artificial neural networks, nonlinear input-output modeling, and nonlinear partial least squares (PLS) modeling are introduced as alternative techniques for developing nonlinear models. Chapter 5 focuses on nonlinear model development from systems science and chaos point of view. It illustrates how the concept of correlation can be extended to the nonlinear framework and used for model development and reduction.

Chapter 6 deals with batch process monitoring problem. It starts with discussion of statistical process monitoring (SPM) tools for univariate problems (Section 6.1). Shewhart, cumulative sum (CUSUM), and exponentially weighted moving average (EWMA) charts are presented. Then, multivariate tools (PCA and PLS) for SPM of continuous processes are discussed in Section 6.2. The phase change point (landmark) detection problem and data length adjustment are discussed in Section 6.3, introducing indicator variable, dynamic time warping (DTW) and curve registration techniques. In Section 6.4, SPM of multivariable batch processes is discussed and multiway PCA, multiway PLS, multiscale SPM with wavelets techniques are introduced. Finally in Section 6.5, on-line SPM of batch processes is addressed using multiway PCA and hierarchical PCA techniques, and Kalman filters for final product quality estimation.

Chapter 7 presents various control problems in batch process operations. The first problem is the determination of the optimal reference trajectories that should be followed during the batch rim. This is an optimal open-loop control problem. A related problem is the determination of the benefits, if any, of forced periodic operation of the fermentation system and the variables and operating conditions that will maximize productivity and selectivity. The other control problems focus on closed-loop control using multi-loop, linear quadratic Gaussian, and model predictive control techniques.

Chapter 8 discusses various fault diagnosis techniques. One approach is based on determining first the variables that contribute to the increase in the statistic that indicates an out-of-control signal and then using process knowledge to reason about the source causes that will affect those variables to identify the likely causes of faulty operation. The contribution plots method is presented in the first part of the chapter. Automating the integration of the variables indicated by contribution plots and process knowledge with a knowledge-based system (KBS) is discussed in the last section of the chapter. Section 8.2 of the chapter is devoted to multivariate statistical classification techniques such as discriminant analysis and Fisher's discriminant function, and their integration with PCA. Section 8.3 focuses on a variety of model-based techniques from systems science for fault diagnosis. Generalized likelihood ratio, parity relations, observers, Kalman filter banks, and hidden Markov models are presented. Section 8.4 is devoted to model-free fault diagnosis techniques such as limit checking, hardware redundancy and KBSs. The last section outlines real-time supervisory KBSs that integrate SPM, contribution plots and KBS rules to provide powerful fault diagnosis systems.

Chapter 9 introduces some related developments in modeling, dynamic optimization, and integration of various tasks in batch process operations management. Metabolic engineering, metabolic flux analysis and metabolic control analysis concepts are introduced and their potential contributions to modeling is discussed. Dynamic optimization and its potential in industrial applications is discussed and compared with classical and advanced automatic control approaches. The integration of various tasks in process operation using a supervisory knowledge-based system is outlined for online process supervision.

Was this article helpful?

0 0

Post a comment