Robust compensator control of continuous fermentation processes

The paper deals with the robust compensator control of continuous fermentation processes described by a set of three non-linear differential equations. For the design purposes the non-linear model is transformed into linear one with interval parameters. Robust state space compensator is designed by the internal model principle, which ensures robust step-wise set points asymptotic tracking and external disturbances rejection in the wide working range. The effectiveness of the algorithm designed is performed by computer simulation experiments. An important feature of the proposed algorithms is their robustness over parameter uncertainties in the process models.     

Optimal control of continuous fermentation processes

Three layer control structure is proposed for optimal control of continuous fermentation processes. The start-up optimization problems are solved as a first step for optimization layer building. A steady state optimization problem is solved by a decomposition method using prediction principle. A discrete minimum time optimal control problem with state delay is formulated and a decomposition method, based on an augmented Lagrange's function is proposed to solve it. The problem is decomposed in time domain by a new coordinating vector. The obtained algorithms are used for minimum time optimal control calculation of Baker's Yeast fermentation process.List of Symbols x(t) g/l biomass concentration - s(t) g/l limiting substrate concentration - x ⁰ g/l inlet biomass concentration - s ⁰(t) g/l inlet substrate concentration - D(t) h^–1 dilution rate - (t) h^–1 specific growth rate - Y g/g yield coefficient - (t) h^–1 specific limiting substrate consumption rate - k _D h^–1 disappearing constant - w ₁, w ₂ known constant or piece-wise disturbances - _m h^–1 maximum specific growth rate - k _s g/l Michaelis-Menten's parameter - h time delay - x ₀, s ₀ g/l initial concentrations - ¯x, ¯s, ¯D optimal steady state value - V _min , V _max , v=x,s,d,t bounds of variables - t h sampling period - K number of steps in the optimization horison - Js, J _d performance indexes - L _s Lagrange's function - L _d Lagrange's functional - ₀ weighting coefficient for the amount of the limiting substrate throwing out of the fermentor - ₁, ₂ dual variables of Lagrange's function - steps in steady state coordination procedure - errors values for steady state coordination process - _v , v=x, s conjugate variables of Lagrange's functional - _v , v=x,s penalty coefficients of augmented Lagrange's functional - _v , v=x, s interconnections of the time - e _v , v=x,s, D, _x , _s gradients of Lagrange's functional - j, l indexes of calculation procedures - values of errors in calculations The researches was supported by National Scientific Research Foundation under grants No NITN428/94 and No NITN440/94     

Estimation and control techniques for continuous culture fermentation processes

This article presents results obtained when some modern estimation and control techniques are applied to a simulated fermentation process. The control structure uses a particular observer of the substrate concentration and assumes the biomass concentration is measurable. The overall structure has been tested for both external and parametric disturbances, with very good results.     

On the controllability of continuous fermentation processes

Simulation may be used as a powerful tool for accelerating bioprocess design. This paper demonstrates the use of simulations in exploring the nature and impact of the interactions that exist in a typical bioprocess for the recovery of an intracellular protein. The study shows that an integrated approach to design must be adopted in order to achieve acceptable process designs. Data from a fed-batch fermentation, with verified models for cell harvesting, cell disruption and cell debris removal have been integrated to demonstrate the consequence of process design and operating decisions on the resulting process performance. The trade-offs between product recovery and the extent of cell debris removal for a range of operating conditions have been represented through a series of windows of operation which show how process conditions must be altered in order for given process performance levels to be realised. The capacity to account for process performance including the impact of interactions is seen as a pre-requisite for rigorous bioprocess sequence design and optimisation.     

Sliding-mode control of fermentation processes

Continuous fermentation processes described by two nonlinear differential equations with uncertain parameters are considered. Sliding mode control design for these processes is proposed. The control design is carried out with direct use of nonlinear model, expert knowledge and on-line measurement of output variable only. Chattering phenomena are avoided by realizing the sliding mode with respect to the control input derivative. The excellent performance of presented control is proved through simulation investigations.     

Knowledge-based control of fermentation processes

A decade has passed since the first applications of a knowledge-based approach to the control of bioprocesses were reported. During this period, both the development and application of intelligent control in biotechnology have undergone remarkable evolution in terms of concepts, objectives, and tools. Stimulated by rapid progress in the field of real-time expert systems, knowledge-based methodology for the control of fermentation processes has now reached a more mature phase. A growing interest among the biotechnology community and intensive, realistic, and fruitful research being undertaken both in universities and in industry suggest that large-scale application of knowledge-based systems for the control of bioprocesses is inevitable. This article provides a concise summary of the main achievements in this new area and discusses recent trends, porblems, and perspectives.     

Computer control of fermentation processes

A review of computer control of fermentation processes is presented. Hardware and software technologies that have been used to implement computer control are discussed. This includes instrumentation, interfacing techniques, computer hardware configurations, data logging and documentation, displays and man-machine interaction, low-level control, back-up and error detection and programming techniques. Advanced control of fermentation processes with the utilization of modern control techniques is also presented. This topic is divided into steady state optimization and dynamic optimization. Finally, on-line estimation of bioreactor parameters for feedback control is presented.     

Automatic computer control of fermentation processes

The application of an optimization algorithm to fermentation processes is described, which is based on information received by means of cascading the input quantities with pseudo random binary signals. The results in optimization of temperature, pH-value as well as productivity are presented and discussed.     

Physiological state control of fermentation processes

In this article a novel approach to the control of fermentation processes is introduced. A "physiological state control approach" has been developed using the concept of representing fermentation processes through the current physiological state of the cell culture. No conventional mathematical model is required for the synthesis of such a control system.The main idea is based on the fact that during batch, feed-batch, or even continuous cultivation the physiological characteristics of the cell population, jointly expressed by the term "physiological state", are not constant but rather variable, which is reflected in expected or unexpected changes in the behavior of the control plant, and which requires flexible alteration of the current control strategy. The proposed approach involves decomposition of the physiological state space into several subspaces called "physiological situations." In every physiological situation the cell population expresses stable characteristics, and therefore an invariant control strategy can be effectively applied. The on-line functions of the physiological state control system consist of the calculation of physiological state variables, determination of the current physiological situation as an element of a previously defined set of known physiological situations, switching of the relevant control strategy, and calculation of the control action. Attention is focused on the synthesis of the novel and nonstandard part of the control system - the algorithm for online recognition of the current physiological state. To this end an effective approach, based on artificial intelligence methods, particularly fuzzy sets theory and pattern recognition theory, was developed. Its practical realization is demonstrated using data from a continuous fermentation process for single cell protein production.     

Neural network modeling in optimisation of continuous fermentation processes

The capability of self-recurrent neural networks in dynamic modeling of continuous fermentation is investigated in this simulation study. In the past, feedforward neural networks have been successfully used as one-step-ahead predictors. However, in steady-state optimisation of continuous fermentations the neural network model has to be iterated to predict many time steps ahead into the future in order to get steady-state values of the variables involved in objective cost function, and this iteration may result in increasing errors. Therefore, as an alternative to classical feedforward neural network trained by using backpropagation method, self-recurrent multilayer neural net trained by backpropagation through time method was chosen in order to improve accuracy of long-term predictions. Prediction capabilities of the resulting neural network model is tested by implementing this into the Integrated System Optimisation and Parameter Estimation (ISOPE) optimisation algorithm. Maximisation of cellular productivity of the baker's yeast continuous fermentation was used as the goal of the proposed optimising control problem. The training and prediction results of proposed neural network and performances of resulting optimisation structure are demonstrated.     

Analysis of continuous fermentation processes in aqueous two-phase systems

Simulations of continuous ethanol or acetonobutylic fermentations in aqueous two-phase systems show that at high substrate feed concentrations it is possible to obtain solvent productivities about 25–40% higher than in conventional systems with cell recycle if the biomass bleed rate is kept about one tenth of the value of D.List of Symbols a Volumetric fraction of dextran rich phase - B h^–1 Bleed rate - D h^–1 Dilution rate - P kg m^–3 Product concentration - PD kg m^–3 h^–1 Productivity - S kg m^–3 Substrate - X kg m^–3 Biomass - Partition coefficient     

Extremum seeking scheme for continuous fermentation processes described by an unstructured fermentation model

This paper examines the problem of maximising the productivity of a class of fermentation processes described by an unstructured fermentation process model. For a given dilution rate, an extremum seeking adaptive control has been used to maximise the productivity of a fermentation process. The concept behinds the extremum seeking method is to iteratively adjust the feed substrate rate in order to steer the process to yield a maximum productivity. The main advantage of the extremum seeking adaptive control is it does not require any structural information of the modeling uncertainty.     

The logical analysis of continuous, non-linear biochemical control networks

We propose a mapping to study the qualitative properties of continuous biochemical control networks which are invariant to the parameters used to describe the networks but depend only on the logical structure of the networks. For the networks, we are able to place a lower limit on the number of steady states and strong restrictions on the phase relations between components on cycles and transients. The logical structure and the dynamical behavior for a number of simple systems of biological interest, the feedback (predator-prey) oscillator, the bistable switch, the phase dependent switch, are discussed. We discuss the possibility that these techniques may be extended to study the dynamics of large many component systems.     

Assessment of in-line near-infrared spectroscopy for continuous monitoring of fermentation processes

The application of NIR in-line to monitor and control fermentation processes was investigated. Determination of biomass, glucose, and lactic and acetic acids during fermentations of Staphylococcus xylosus ES13 was performed by an interactance fiber optic probe immersed into the culture broth and connected to a NIR instrument. Partial least squares regression (PLSR) calibration models of second derivative NIR spectra in the 700-1800 nm region gave satisfactory predictive models for all parameters of interest: biomass, glucose, and lactic and acetic acids. Batch, repeated batch, and continuous fermentations were monitored and automatically controlled by interfacing the NIR to the bioreactor control unit. The high frequency of data collection permitted an accurate study of the kinetics, supplying lots of data that describe the cultural broth composition and strengthen statistical analysis. Comparison of spectra collected throughout fermentation runs of S. xylosus ES13, Lactobacillus fermentum ES15, and Streptococcus thermophylus ES17 demonstrated the successful extension of a unique calibration model, developed for S. xylosus ES13, to other strains that were differently shaped but growing in the same medium and fermentation conditions. NIR in-line was so versatile as to measure several biochemical parameters of different bacteria by means of slightly adapted models, avoiding a separate calibration for each strain.     

Data acquisition and control of a continuous fermentation unit

Summary Data acquisition and computer control ofZymomonas mobilis fermentation for ethanol production has been studied. An HP 200 series microcomputer system was used in conjunction with an HP 3497A data acquisition unit. On-line ethanol, glucose and cell mass were measured for use as possible control variables. Dilution rate was used as the manipulative variable. A versatile, user-friendly data acquisition program was written to gather, control and analyze data from the continuous fermentation. The program allows user-given control and calibration algorithms so that sophisticated control policies, e.g., self-tuning regulator (STR) and instrumentation, can be implemented with relative ease.     

Ageing vessel design and optimization for continuous very-high-gravity ethanol fermentation processes

A continuous very-high-gravity (VHG) ethanol fermentation process design, consisting of a chemostat vessel connected to several equal-sized ageing vessels configured in parallel, was developed. The objective of the developed process is to have complete glucose utilization during fermentation stage. The process design integrates the conservation of mass principle and the experimental data of collected residual glucose profiles measured under VHG conditions. An ageing vessel involves three consecutive time periods: filling, ageing and operating. The ageing time is biological relevance, and is affected by the initial glucose concentration, the ethanol concentration, and the yeast viability in an ageing vessel. The operating time period is adjustable; a short operating time means a high discharge rate in order to empty an ageing vessel. The filling time links to the selection of the number of equal-sized ageing vessels that are installed downstream to a chemostat device. The developed process features the use of equal sized fermenters for all chemostat and ageing vessels so that the vessel exchangeability and the flexibility of fermentation operation are increased.     

Evaluation of sensors for the control of a continuous ethanol fermentation

A method is presented for the evaluation of sensors used in the control of continuous fermentations. Simulations of open-loop response to input disturbance provided a starting point for the choice of sensor type. This was evaluated quantitatively through a sensitivity ratio. It was shown that in the case of ethanol fermentation, there existed three regions where different sensors could be used for the process control depending on the inlet sugar concentration. Sugar sensors were preferable above an inlet sugar concentration of 50 kg/m³, while ethanol sensors were preferable below 25 kg/m³. In the intermediate region, sugar and ethanol sensors demonstrated equally good performance. A controllability study of a continuous ethanol fermentation was also made. A single-stage continuous stirred-tank fermentor was simulated operating at a dilution rate of 0.1 1/h and inlet glucose concentration of 160 kg/m³. The outlet glucose concentration was controlled with a PI controller. Mean square error of the controller input signal during the first five hours after introducing input disturbance was taken as a measure of the controllability. This was studied in the relation to the two key sensor characteristics, sampling time and accuracy.List of Symbols c _p kg/m³ ethanol concentration - c _p kg/m³ fermentor ethanol concentration corresponding to c _si and D - c _s kg/m³ substrate (glucose) concentration - c _s kg/m³ fermentor glucose concentration corresponding to c _si and D - c _si kg/m³ inlet substrate (glucose) concentration - c _si kg/m³ inlet glucose concentration value used for sensitivity evaluation - c _sm kg/m³ glucose concentration — measured value - c _ss kg/m³ glucose concentration setpoint value - c _x kg/m³ biomass concentration - D 1/h dilution rate - D 1/h dilution rate value used for sensitivity evaluation - D _i 1/h dilution rate at ith sampling interval - D ₀ 1/h dilution rate at steady state - K _c m³/kgh controller gain - K _p kg/m³ product inhibition constant - K _s kg/m³ Monod constant - n ₁, n ₂ random numbers - r _p kg/m³ h ethanol production rate - r _s kg/m³ h substrate (glucose) consumption rate - r _x kg/m³ h biomass growth rate - vector of independent variables - y _i ith dependent variable - Y _ps ethanol yield - Y _xs biomass yield - parameter vector - _j jth parameter - _ij sensitivity of y_i with respect to _j - _p sensitivity of fermentor ethanol concentration - _s sensitivity of fermentor glucose concentration - sensitivity ratio - c _p kg/m³ ethanol concentration difference corresponding to a change of c _si by 5% - c _s kg/m³ glucose concentration difference corresponding to a change of c _si by 5% - c _si kg/m³ concentration difference added to c _si - _i kg/m³ error at ith sampling interval - 1/h specific growth rate - _m 1/h maximum specific growth rate - _s kg/m³ standard deviation of monitored glucose concentration - _I h min kg/m³ integral time - _s min sampling period The Swedish Ethanol Foundation and the National Board for Technical Development (NUTEK) are kindly acknowledged for the financial support of this project. The authors wish to thank Peter Warkentin for the linguistic advice.