Search

paper deals with the application of the reinforcement learning technique (from the well-known field of machine learning) for the control of a biotechnological process – the bacterial growth. The mathematical model of the bioprocess consists of a set of nonlinear differential equations. This type of systems is difficult to control because they are strongly nonlinear, with time-varying parameters and very sensitive to small changes in initial conditions. The paper presents a control strategy based on an actor-critic structure from the reinforcement learning approach that is a combination of optimal control (based on state feedback) and adaptive control. The adaptive part of the controller is a neural network used for value function approximation. The performance of the algorithm is analyzed using numerical simulations.

This paper presents the integration on a single chip of multiple electrochemical sensors based on amperometric and potentiometric techniques, as well as of a physical sensor. The prototype comprises six sensors for in-situ monitoring of important parameters of a bioprocess; glucose, lactate, pH, cell density, dissolved oxygen, and temperature. All sensor elements have been calibrated and showed acceptable detection performance. We also demonstrate a preliminary experiment for in-situ monitoring in a yeast culture to better understand the response of the proposed microsensors towards yeast fermentations application.

This study addresses an innovative multivariable control strategy for a wastewater treatment bioprocess with activated sludge. This bioprocess takes place in a continuous reactor located at the Wastewater Treatment Plant Facai, Romania. The proposed adaptive control structure is designed under realistic assumptions: several process variables are unavailable, the reaction kinetics is time-varying and unknown, the influent pollutant flow rate varies in large limits and much more its concentration is totally unknown. The novelty of the approach consists in the insertion of a sliding mode observer (SMO) in overall control structure. The SMO provides the estimates of the unknown influent pollutant concentration. The performance of the adaptive multivariable control algorithm is assessed via numerical simulations based on data provided from the plant.

The food industry needs suitable analytical instrumentation for quality control and contamination monitoring. Such methods must be rapid, reliable, specific, and cost-effective. Biosensors, especially enzyme-based sensors, offer several advantages in monitoring bioprocessing systems. Continuous flow analysis systems with integrated biosensors are important tools in bioprocessing monitoring. This work presents the development of an affordable bioanalyzer for food analysis using enzymatic biosensors as amperometric detectors. The instrument is very versatile due to its capability to be utilised for the analysis of a great variety of analytes, such as ethanol, fructose, glucose, lactate, cholesterol, etc. It is fully controlled by a desktop application which allows you to configure all parameters of the measurement and obtain the calibration curve; then, the sample concentration may be obtained.

International audience; From genetic sequencing, dry biomass, andmetabolites measurements, the assignment of functions to the species present in chemostat experiments was solved by merging chemostat modelling and quadratic mixed integer programming. The method was tested on a nitrification bioprocess where two functions are known to drive the system. Sensitivity of the method, its advantages, and limitations are discussed.

Towards the optimization and control of bioprocesses, as the culture step of recombinant microorganisms, it is crucial to develop high-throughput monitoring techniques enabling the acquisition of a large data sets of information, as the ones based on mid infrared (MIR) spectroscopy. To maximize the knowledge associated to this this new large-scale data, it is also relevant to apply machine learning techniques, as Partial Least-Squares Discriminant Analysis (PLS-DA). In the present work, Principal Component Analysis followed by PLS-DA were applied to discriminate different growth phases of recombinant Saccharomyces cerevisiae along the production of an heterologous protein, conducted in bioreactor and monitored by high-throughput MIR spectroscopy. It was possible to derive PLS-DA models enabling to discriminate, from the MIR spectra, the yeast cells according to its metabolic status associated to the culture growth phase with an accuracy, sensitivity, and specificity between 83% and 100%. The optimised PLS-DA presented very low calibration errors, of 97% and 100% based on a cross-validation and an independent data-set, respectively. In conclusion, it was possible to build a PLS-DA model discriminating the cells metabolic status that will promote the knowledge of the bioprocess and future better control and optimization procedures.

The anaerobic digestion is an important bioprocess for the benefits it produces in energy and ambient. Yet, its monitoring and control are still hampered by the cost and lack of proper sensors. One way to mitigate this problem is the use of observers to estimate the desired variables. Prior to the observer design, an observability analysis can be done in order to determine which variables can be estimated from a given sensor configuration. Being the anaerobic digestion a highly non linear process, the usual observability analysis methods involve very complex calculations. However, analysis methods with less calculation complexity have been proposed in recent works, which also allow considering unknown inputs. Roughly, these methods consist in the dynamical study of the indistinguishable trajectories of the system, if they exist. Particularly, in this work, those methods are applied to the detectability and observability analysis of the anaerobic digestion process considering many sensor configurations. The analysis aims to obtaining a tool to facilitate the selection of sensors, to maximize the information that can be obtained from the process.

Wavelet neural network is an alternative to artificial neural network in empirical modeling of industrial processes due to efficient initialization of network parameters that reduces training time. In this paper, particle swarm optimization methods are used for initialization of dilation and translational parameters in two wavelet neural network models. Dissolved oxygen models are constructed from real bioprocess data of pilot scale fed-batch bioreactor in polyhydroxyalkanotes (PHA) production and an industrial-scale fed-batch bioreactor in penicillin production. Simulation output of dissolved oxygen and initial mean square error (IMSE) show that the distance and error between initialization and training data are small in PSO method compared to random and heuristic methods. This ensures training phase start very close to target data.

A novel multivariable adaptive control scheme for a lactic fermentation biotechnological process with unknown inputs taking place inside two sequenced Continuous Stirred Tank Reactors (CSTRs) is developed. The design of the control scheme is achieved under the hypothesis that the influent substrate concentrations and the reaction kinetics are unknown and time-varying, but also a part of the state variables are unmeasurable. The adaptive controller includes a linearizing controller connected with two state observers used for the real-time estimation of the unknown input substrate concentrations and with a parameter estimator which is used for kinetics estimation. The estimation of unknown influent substrate concentrations is realized by using the measurements of the internal substrate and lactic acid concentrations. The proposed parameter estimator employs the information which is provided on-line by two sliding-mode state observers. The performance of the designed control scheme is evaluated via simulation experiments conducted for a lactic acid production process (LAPP) carried out inside two sequenced CSTRs. The dynamics of the considered bioprocess is described by a nonlinear, incompletely known and time varying model.

The paper addresses the design and analysis of Iterative Learning Control (ILC) method for a wastewater biodegradation process. The bioprocess is considered to take place inside a continuous stirred tank bioreactor. The use of this kind of control methods is motivated by its advantages in the case of complex nonlinear systems like biotechnological processes. There were used two ILC algorithms one considered a classical approach in the field (PD-type learning algorithm), and the second one which try to exploit some characteristics of the input signal (exponential learning algorithm). These control methods are implemented for the de pollution control problem in the case of an anaerobic digestion process. This bioprocess is characterized by strongly nonlinear and not exactly known reaction rates. Furthermore, not all the state variables are measurable. The performance and effectiveness of the presented control algorithms are proven by simulation results.

This paper presents two estimation based control strategies for an aerobic process used in wastewater biological treatment carried out inside a CSTR (Continuous Stirred Tank Reactor). Our control objective is to keep the effluent pollutant concentration at a desired low level, and also to ensure that the dissolved oxygen concentration follows some imposed references. Therefore we propose two new control schemes for robust-adaptive and model predictive control, taking into account the Bayesian based weighted values of the state’s estimations provided by an interval observer and a parameter estimator of the unknown growth rate. A comparison of the effectiveness of the two approaches is illustrated by numerical simulations performed under different conditions related to some degree of uncertainness of a continuous wastewater treatment bioprocess.

With the aid of systems biology and synthetic biology, we will provide an integrated, direct and green bioprocess to address the overall challenge for universal village in production of biofuels using solar energy and captured atmospheric carbon dioxide (CO2). This will in practice be done using cyanobacteria and exemplified with the chemical/fuel ethanol as a model system. This, in addition, makes it possible to introduce waste water as a source of nutrients for the genetically engineered cyanobacteria and thereby use the developed system as a model to purify contaminated water bodies.

Cultivation is the paramount step of microalgae bioprocessing as a concentrated and flourished biomass broth will result in higher recoveries of its metabolites. But there is scarce model development for the prediction of microalgae biomass productivity in outdoor cultivation systems under the fluctuations of process conditions. Hence, this study develops a mathematical model based on the first principles with a nonlinear approach to forecast the microalgae growth in open ponds under non-optimal cultivation conditions in respect of the location. The model accounts for the effect of temperature, solar irradiance level, availability of nutrients in the cultivation medium and the effect of local climatic conditions over the microalgae growth. The model is validated against the experimental results from literature data and model analyses have been proceeded to determine the effect of variations in process parameters on the microalgae biomass yield.

This paper present a modeling approach of a bioprocess based on a Takagi-Sugeno formulation (TS). Using the resulted TS model, we construct a Parallel Distributed Compensation control law (PDC) to ensure the convergence of the state to a specific equilibrium points. The performance of this approach is illustrated by means of a simple example of a biomass growth process.

In this paper a sliding-mode observer for a batch bioprocess, the δ-endotoxins production of bacillus thuringiensis (BT), is presented. The proposed observer is based on the equivalent control method and a class of second-order sliding mode operators. The use of these operators in the observer design allows the fixed-time convergence of the measured variables, while the unmeasured variables converge exponentially. This structure allows to estimate the biomass in the δ-endotoxins production of BT, even, under noisy measurement conditions. Simulations show the feasibility of the proposed observer. Convergence proofs are also presented.

This paper deals with the measurement of the biomass concentration in a bioprocess, with a final purpose of building an on-line concentration transducer, for the control of a continuous bioprocess. The proposed method is based on processing the images of the target population, captured by a camera under a regular bright field microscope. The image processing algorithm for this method was developed and experimentally validated, in the case of the species Sacharomices cerevisiae. Its main properties are: no human operation required, fast computing (less than 10 sec.), low error level.

Automation of any process improves process efficiency, yield and minimizes overall cost. Bioreactors place a significant role in pharmaceutical and food processing industries. To utilize the microbes efficiently, the bioreactor should provide optimum conditions such as temperature, pH, oxygen etc. The online monitoring of bioprocess is enhanced by online control of the essential parameters to improve product yield. The effective monitoring of bioprocess is necessary to develop, optimize and maintain biological reactors at maximum efficiency. Furthermore, due to the nature, type and volume of products produced in bioprocesses, there is a strong economic incentive for process monitoring to increase yield and productivity. Most of the bioreactors are supplied with Programmable Logic Controller (PLC)based automation.Most of the PLC supports only PID controllers. The novelty in this paper is the implementation of control algorithms for multiple parameters such as temperature, pH, antifoam and Dissolved Oxygen(DO) using Lab VIEW based automation. The interfacing was done via NI Compact DAQ. Lab VIEW based automation will enhance the performance of the bioreactor with advanced control schemes. The results show that the real-time automation strategy developed in Lab VIEW is capable of simultaneously controlling the temperature, pH, DO and antifoam.

Cell culture in stem cell research is on the rise, not only for basic research but also for its potential medical and therapeutic applications. Monitoring culture process using sensors throughout the culture helps to optimize culture conditions for optimal growth and maximize yield from the cells. Further, the sensor data gives insight into developing better cell culture systems. Typical stem cell culture platforms, e.g. well plates or flasks, cannot be easily equipped with sensors and impose many challenges in periodic process measurements in a CO 2 incubator. We present an incubator compatible modular measurement system with three sensors, to monitor pH, temperature and humidity continuously throughout the culture. Sensors are assembled around a flow through cuvette for highly sterile non-contact measurements. No sample preparation or sample extraction from the incubator is needed and the measurements are carried out in a closed flow loop without wasting any medium. The modular assembly is novel, reusable and feasible for humid incubator environments. The system has been tested, validated and used in mesenchymal stem cell expansion and differentiation, for periods ranging from two to three weeks. Once the measurement has commenced at the beginning of culture, continuous measurements without sensor recalibration or special manual attention are carried out till the end of the culture. Measurement data clearly show the interplay between measured parameters, indicating a few stress sources present all through the culture. Additionally, it gives an overall picture of behavior of critical control parameters in an incubator and points out the need for bioprocess systems with automatic process monitoring and smart control for maximum yield, optimal growth and maintenance of the cells.

System sensitivity analysis (SSA) is a mathematical method used in the study of behaviour of dynamic systems. From engineering practice is well known that a dynamic system did not respond in the same way to all external stimulus (command or disturbance). The results of SSA study for anaerobic digestion bioprocess, provides important information for synthesis of control system. Also, it is possible to estimate, by simulation, quantitative changes for the system variables in the case of changes in internal structure.

Modeling of biotechnological systems is an important research area. The most challenging approach is to build non-linear state space models for these systems. In this work the parameters of a bacterial growth bioprocess are estimated using prediction error method. Prediction error methods are widely used in parameter estimation both for linear and nonlinear models and consist in minimization of the distance between measured and modeled data in a suitable norm. Because these problems are solved using numerical algorithms that are time consuming, in this paper a parallel particle swarm optimization technique is used in order to numerically solve the minimization problem. The algorithm is implemented on a multicore processor and the performances of this approach are presented by numerical simulations.

The control of bioprocesses remains an inherently complex problem. Specialist's efforts are directed towards ensuring stability and process efficiency. The necessity to obtain performances without installing a lot of redundant and dedicated expensive equipment, forces the developers to use the techniques available to processing all the information that are “hidden” in the bioprocess and to compensate the uncertain and time varying of parameters. This paper presents the design of some control methods, able to deal with the uncertainties, for a complex photoautotrophic growth bioprocess. To develop an efficiency control law it is necessary to estimate the real values of the inputs. Specifically, the methods proposed in the paper, allowing the assessment of carbon dioxide flow, input into the photobioreactor with a high degree of uncertainty.

In this paper, maximization of efficiency of the penicillin production is addressed. To overcome the issues caused by the terminal time being an optimizable parameter, two adaptations of the Hamiltonian-based gradient method are proposed. The first algorithm incorporates the terminal time directly into the set of optimized variables while the second one makes use of a time scaling with optimizable time-dependent time scale being a virtual input. Both algorithms are verified on a set of numerical experiments and the obtained results show significant improvement of the penicillin production efficiency gained by their use, which suggests their exploitation within the industrial nonlinear predictive controllers for bioprocess optimization.

Fermentation process is one of the most complicated processes, which is a large scale nonlinear modeling problem with multi-constraints and multi-variables. The successful applications of model-based advanced measurement and control techniques mostly depend on the accuracy of bioprocess model. As a swarm intelligence approach, particle swarm optimization (PSO) algorithm can explore the global optimal solution of complex problems. In recent years, there has increasingly research activities regarding the chaotic PSO (CPSO) algorithms, which performs better by instead pseudo random number with chaotic series. In this paper, a novel CPSO algorithm exploiting the properties of swarm energy is proposed for the bioprocess model parameters identification. The algorithm remains the good global searching ability by releasing the swarm energy slowly and accelerating the particles with worse fitness. Simulation experiment results show that the improved CPSO not only performs better convergence capability but also presents better solution quality as compared to the conventional CPSO.

This paper proposes a novel Model Predictive Control strategy of dissolved oxygen (DO). The control design finds application on the Bt delta-endotoxins batch process production. This strategy improves the DO control by preventing the usual drawbacks that appear with a step level change in the DO reference profile. Simulations results are provided as well as comparison results against a well-known Lyapunov based controller and a classical PID controller.

In this paper, a review of empirical correlations proposed by researchers from 1974 to nowadays is performed. The reviewed correlations allowed to determine the mass transfer coefficient (kL), the volumetric oxygen transfer coefficient (kLa), the interfacial mass transfer area (a), and the bubble diameter (dBS) for aerated tanks with circular and rectangular cross-sectional areas. Theoretical predictions of the volumetric mass transfer coefficient that have been recently proposed were described; finally, different criteria for bioreactor scale-up were considered in the light of the influence of oxygen transfer rate affecting the dissolved oxygen concentration in real bioprocess. From this review it can be concluded that, although there are several studies in this area, still there are difficulties to obtain more explicative models describing the oxygen transfer from a liquid phase to a gas phase.

In biopharmaceutical manufacturing, monitoring of bioprocess parameters is very crucial for maximum cell growth, viability and density, as well as biologic product concentration and yield. At the same time, disposable technologies have become popular due to their low capital costs, minimal contamination risk and lower sterilization overhead. We have developed a sensor-array that utilizes novel stimuli-responsive hydrogels intended for use as the sensing elements of disposable sensors in single use bioreactors. The response of hydrogels to changes in environmental concentration of a specific analyte is transduced using a magnetic or pressure sensor. Here we present recent results from glucose, lactate and osmolality hydrogels measured using pressure and magnetic sensors respectively. In addition to these results, we present a novel transduction method for the hydrogel response, along with a design for a sensor-transducer array for use in bioprocess control.

The design of some adaptive and robust-adaptive control strategies for a microalgae fermentation process that is carried out inside a Continuous Stirred Tank Reactor (CSTR) is presented. Initially, an indirect adaptive controller is designed by coupling a linearizing controller with a new parameter estimator used for estimation of unknown kinetics. Afterwards, under the realistic assumption that the reaction kinetics are unknown, time-varying and moreover the influent flow rate is uncertain and time-varying, but some lower and upper bounds of these uncertainties are known, two robust-adaptive control structures are proposed by combining a linearizing control law with two new appropriately parameter estimators able to estimates the unknown reaction rates. Numerical simulations are included to validate the designed controllers applied to a continuous fermentation bioprocess used for microalgae production.

This paper is concerned with the parameter estimation of a bacterial growth bioprocess using heuristic optimization. The identification algorithms used a dynamic mathematical model containing nine unknown parameters. These parameters were determinate using simulated annealing and genetic algorithms through the minimization of an evaluation function. The reaction kinetics used to simulate microbial growth are the Monod and Haldane equations. The parameter estimation problem is formulated as a multi-modal numerical optimization problem The performances of the two methods are analyzed by numerical simulations.

Infrared spectroscopy, either in the near and mid (NIR/MIR) region of the spectra, has gained great acceptance in the industry for bioprocess monitoring according to Process Analytical Technology, due to its rapid, economic, high sensitivity mode of application and versatility. Due to the relevance of cyprosin (mostly for dairy industry), and as NIR and MIR spectroscopy presents specific characteristics that ultimately may complement each other, in the present work these techniques were compared to monitor and characterize by in situ and by at-line high-throughput analysis, respectively, recombinant cyprosin production by Saccharomyces cerevisiae. Partial least-square regression models, relating NIR and MIR-spectral features with biomass, cyprosin activity, specific activity, glucose, galactose, ethanol and acetate concentration were developed, all presenting, in general, high regression coefficients and low prediction errors. In the case of biomass and glucose slight better models were achieved by in situ NIR spectroscopic analysis, while for cyprosin activity and specific activity slight better models were achieved by at-line MIR spectroscopic analysis. Therefore both techniques enabled to monitor the highly dynamic cyprosin production bioprocess, promoting by this way more efficient platforms for the bioprocess optimization and control.

State and reaction rate estimation in highly uncertain bioprocess models is an important topic for a successful control of these processes. Discontinuous observers, as for example the super-twisting estimator, offer unique properties as convergence in finite time and insensitivity with respect to non-vanishing perturbations. We present a generalization of the classical super-twisting observer and illustrate its use in the estimation of variables in a bioprocess.

Non-dominated sorting genetic programming is used to make the evaluating of several objectives impersonally. These objectives are the complexity, the oscillation and the training errors of a model. The preference of a decision-maker is integrated into non-dominated sorting GP and then a preference-based non-dominated sorting genetic programming is proposed. In order to improving the searching efficiency, decision-maker's preference is used to guide the searching direction. Last, several models are selected from the pareto front based on their performance on each objective and an integrated model is obtained. The approach is used to model the biomass concentration and its effectiveness are demonstrated.

This work deals with membrane processes that stand for an emerging technology in chemical and bioprocess industry used both in production and down-stream processing. Membrane processes receive growing attention mainly due to reduced energy demands and higher efficiency of the achieved separation or processing goals. These systems, however, did not receive much attention from process optimization community and that is why they provide many opportunities for development of optimal operation design. Membrane processes, their common features, and principles will be discussed in details. Mathematical models suitable for modelling, simulation, and control will derived. The existing industrial operation and control will be explained. Theory of automatic control will be employed to suggest improvements over the existing state of control and to propose new control strategies. Selected case studies will demonstrate the advantages of the proposed approach.

Neural network and genetic algorithm with altered genetic operators have been implemented into bioprocess field to improve the modelling approach. Enzyme streptokinase is secreted by β-hemolytic streptococcus sp. often used to treat acute myocardial infarction, being a vital drug it is required to be produced in recombinant form for thrombolytic therapy. The unstructured model factors are found to apparently influence the existence of active cells in bioreactor environment. The structural features of a cell itself can be assumed to be an entire functional structured system with various sub-cellular entities. Our endeavour is to make out obligatory constraints that deal with plasmid instability. On simulation of process the patterns obtained noticeably depicts the role of few relevant parameters governing such system, including, metabolite concentration and dilution rate, to find out segregational instability and competitive dynamics in growing cell population. An adjustment of genetic operators to deal with two key factors of continuous culture model is taken into account. Neural network based technique has been employed to constitute logical models prior to serve as tool for optimizing process parameters of the dynamical system.

In this paper the widely used benchmark simulation model NO.1 (BSM1) is extended to combine the phosphorus removal by adding two anaerobic tanks and applying the AMS2d as bioprocess. A default control loop and a series of 10 control strategies are implemented in the new simulation plant to test the performance as well as the control strategies. The performance of the plant and the cost of operation are also given to compare the control strategies.

This paper deals with on-line estimation and control of a biological wastewater treatment plant process. The objective of the control is to force the residual substrate and the dissolved oxygen to track a given set point despite the disturbances and system parameter uncertainties. The control law is a closed loop paradigm control based on the positive invariance. The control approach structure is combined with a discrete minimal-order observer, for the on-line reconstruction of unmeasured biological states and unknown parameters of the bioprocess. The efficiency of both the control and estimation are demonstrated via computer simulations.

This paper deals with the modeling and controlling of an uncertain biotechnological process. More precisely, in the first part we estimate the biomass concentrations in a biotechnological process by using interval observers and the reaction rates using an OBE. Since the modeling of these processes is uncertain, only some bounds of the biomass concentrations can be estimated by using two appropriate interval observers. The second part refers to designing a robust - adaptive control structure for the process. The effectiveness and performance of these strategies are illustrated by numerical simulations.

This paper presents the design of adaptive and robust-adaptive control strategies for an anaerobic bioprocess with biogas production. The design schemes are developed under the realistic assumption that both bacterial growth rates and influent flow rates are time-varying and uncertain, but some lower and upper bounds of these uncertainties are known. The proposed control structures are achieved by combining a linearizing control law with a state asymptotic observer or an interval observer, and with a parameter estimator used for on-line estimation of unknown kinetics of the process. Numerical simulations are performed to test the proposed algorithms.

Bioprocess scale-up from optimized flask cultivations to large industrial fermentations carries technical challenges and economical risks. Essentially, the prediction of optimal process conditions in large fermentations based on small scale experiments is non-trivial. For example, common statistical methods encounter problems with the high-dimensional, small sample size and, on the other hand, the use of various scale-up criteria requires a priori knowledge that may be difficult to obtain. We propose a novel computational scale-up approach applicable to various bioprocesses. The method bases on regularized linear and logistic regression. With embedded feature selection, it automatically identifies the most influential parameters and predicts their values in large scale. In addition, the method predicts the resulting large scale yield. As a case study, we examined the production of a cytotoxic compound. We predicted scale-up from flask and 2L to 30L fermentations and found that, in both cases, the product yield predictions are close to experimentally observed yields.

This paper presents the design and analysis of adaptive and robust-adaptive control strategies for a nonlinear propagation bioprocesses carried out inside a fixed bed reactor. The design schemes are developed under the realistic assumption that both bacterial growth rates and influent flow rates are time-varying and uncertain, but some lower and upper bounds of these uncertainties are known. The proposed control structures are achieved by combining a linearizing control law with an appropriately state asymptotic observer or an interval observer and with a parameter estimator used for on-line estimation of unknown kinetics. The effectiveness of the proposed algorithms is validated by numerical simulations applied in the case of a prototype fixed bed bioprocess.

Demand on modeling an accurate process model as well as optimization of biochemical process has increased as it is vital for sustainable development in bioprocess industries. Therefore, this paper is concerned about the empirical modeling of enzymatic hydrolysis using xylanase for the production of xylose from rice straw. The parameters investigated in this research were temperature, agitation speed of incubator shaker and xylanase concentration, to obtain the production of xylose. Feed-forward neural network (FANN) was employed to describe the relationship of the input and output of the process. Then the genetic algorithm (GA) method was applied to optimize the process condition. The initial data is split into training and validation before re-sampling the data with bootstrap re-sampling method. The training data again was then split into training and testing data.The neural network model was developed with one hidden layer and 6 number of hidden nodes. The correlation coefficient of training and testing set was found to be 0.9970 and 0.9975 respectively, though the correlation coefficient of validation was obtained as 0.8501. The optimization of the parameters namely temperature, agitation speed and xylanase concentration of the xylose production using the GA method was found to be 50.3111°C, 153.5140 rpm and 1.6944 g/l with the optimum xylose production predicted is 0.1845 g/l.

This paper uses a model predictive controller for nonlinear wastewater process described by Takagi-Sugeno model. An approach to design MPC based on fuzzy Kalman filter (FKF) has been proposed. The aim of this suggested method is solving a depollution control problem with controlling two reactant concentrations with two control inputs. Simulation results on the mentioned bioprocess show the performance of the designed controller.

The hydrogen can be produced in a biological production process by dark and photo-fermentation of organic substrates. Under anaerobic conditions, hydrogen is produced during conversion of organic substrate into organic acids using fermentative bacteria and during conversion of organic acids into H2 and CO2 using photo- fermentative bacteria. This bioprocess has been studied with a number of microorganisms, it is a very complex process and influenced by many factors. In order dark and photo-fermentation process is an important approach for bio-hydrogen production. In this study, different factors haves been examined to enhance biohydrogen production by these organisms, either as a combined or sequential using dark and photo-fermentation process. The effect of each factor on biohydrogen production efficiency is reported. A comparison of hydrogen production efficiency between dark-fermentation, photo-fermentation and two stage processes was investigated.

Unlike prevailing contemporary approach considering the influx feed flow as the only crucial input influencing the final product concentration in bioprocess cultivation, the current paper considers other (non-nutritional) inputs as well. Among them, the volume withdrawal is an important option. Two types of control strategies employing this second input are considered here, each of them using it in a different way. First, the so-called quasi-double-input strategies exploit the broth withdrawal to follow a pre-determined volume profile. Second, the true-double-input strategies consider the second input as a completely independent optimization variable. Main analysis tool here is the gradient optimization numerical algorithm and its adaptations. All resulting strategies are summarized and formulated in the context of optimization, those introduced earlier (i.e., the quasi-double-input) are unified into a common general strategy. Practical complication (inadmissible volume decrease) related to withdrawal introduction is successfully addressed and the summary results show indisputable improvement gained by the non-nutrient input introduction.

Continuous estimation of bioprocess state during yeast fed-batch fermentation requires using accurate mathematical model and parameters that can describe the actual state of the yeast growth during fermentation. Application of functional state modelling approach for the yeast Saccharomyces cerevisiae fed-batch fermentations is presented. The main advantage of the modelling approach over global model with complex structure is that the parameters of each local model can be estimated separately from the other local model parameters. Moreover, this approach makes it easier to perform model simulation and parameter estimation compared with the complex global model. Obtained experimental data from yeast fed-batch fermentations show sufficiently good match with the curves simulated using functional state fermentation model.