Search

A multivariate statistical process control (MSPC) method using dynamic multiway neighborhood preserving embedding (DMNPE) is proposed for fed-batch process monitoring. Different from principal component analysis (PCA) which aims at preserving the global Euclidean structure of the data set, neighborhood preserving embedding aims to preserve the local neighborhood structure of the data set. The neighborhood preserving property enables NPE to find more meaningful intrinsic information hidden in the high-dimensional observations compared with PCA. Moreover, the robustness of NPE is better than that of PCA. On the other hand, a dynamic monitoring approach based on moving window technique is employed to deal with the time-variant property of the dynamic processes. An industrial cephalosporin fed-batch fermentation process is used to demonstrate the performance of the DMNPE. The results show the advantages of DMNPE over those methods such as dynamic multiway PCA (DMPCA), static multiway NPE (SMNPE) and static multiway PCA (SMPCA) in fed-batch process monitoring. Finally, the robustness of the DMNPE monitoring is tested by adding noises to the original data sets.

A profit function is established to evaluate the economic performance of baker's yeast continuous production. A structured kinetic model for baker's yeast Saccharomyces cerevisiae is applied to estimate the key state variables of the process, such as biomass and ethanol concentration. Taking possible time variant dynamics into account, a rolling identification approach is used to identify several key model parameters online so as to adapt the changing dynamics of the process. The well predicted biomass concentration makes the profit function online predictable very well. This investigation may benefit the study on profit optimization of the process.

The electrophilic cyclization of 1,2-allenic phosphonates with in-situ-generated PhTeX to afford 4-(phenyltelluro)-2-methoxy-2,5-dihydro[1,2]-oxaphosphole 2-oxides was studied. Good yields were observed when PhTeBr was used as the electrophilic reagent. The stereochemistry of the reaction is also discussed in the paper.

A new spike sorting method based on the support vector machine (SVM) is proposed to resolve the superposition problem. The spike superposition is generally resolved by the template matching. Previous template matching methods separate the spikes through linear classifiers. The classification performance is severely influenced by the background noise included in spike trains. The nonlinear classifiers with high generation ability are required to deal with the task. A multi-class SVM classifier is therefore applied to separate the spikes, which contains several binary SVM classifiers. Every binary SVM classifier corresponding to one spike class is used to identify the single and superposition spikes. The superposition spikes are decomposed through template extraction. The experimental results on the simulated and real data demonstrate the utility of the proposed method.

Recurrent network models are widely used to characterize the behaviors of neurons in the visual cortex. However, they are seldom used to simulate neurons in the retina. In this study, two slightly different recurrent network models are introduced to describe two types of non-concentric ganglion cells, i.e., the impressed-by-contrast cell and the suppressed-by-contrast cell in the cat retina. By simulations, it is found that the additive recurrent network is able to describe qualitatively the behavior of the impressed-by-contrast cell, while the other additive recurrent network with saturation rectification is able to describe qualitatively the behavior of the suppressed-by-contrast cell.

A cell cycle model is proposed for methylotrophic yeast Pichia pastoris grown on glycerol during fed-batch cultivation. Morphological differentiation of cells, such as unbudded daughter cell, unbudded parent cell and budding cell, is depicted by the model. During the cyclic growth, cells in different cycling period are assumed to undergo sequential shifting dominantly. The input of the cell cycle model is the specific growth rate, which is calculated from the macrokinetic model proposed previously. The cell cycle related variables, such as the fraction of budding cells and the cell density are then simulated. Model validation is carried out with the experimental data of off-line assays.

In this article, a nonlinear dynamic multiway partial least squares (MPLS) based on support vector machines (SVM) is developed for on-line fault detection in batch processes. The approach, referred to as SVM-based DMPLS, integrates the SVM with the MPLS model. Process data from normal historical batches are used to develop the MPLS model, and a series of single-input-single-output SVM networks are adopted to approximate nonlinear inner relationship between input and output variables. In addition, the application of a time-lagged window technique not only makes the complementarities of unmeasured data of the monitored batch unnecessary, but also significantly reduces the computation and storage requirements in comparison with the traditional MPLS. The proposed approach is validated by a simulation study of on-line fault detection for a fed-batch penicillin production.

An oxygen transfer model was established for Pichia pastoris growing on glycerol and methanol in a stirred tank bioreactor and expressing a recombinant human serum albumin (rHSA). This was based on pseudo-steady state mass balance, where the volumetric O2 transfer coefficient, k L a, was estimated as a function of power input per unit volume and aeration rate. Under pseudo-steady state, the O2 transfer rate model matched the O2 uptake rate obtained from a previous macrokinetic model. This procedure was also applied to estimate biomass concentration by using the on-line rolling identification approach.

A macrokinetic model is proposed to simulate the growth and metabolism of myeloma cell line in batch and fed-batch cultures. The model describes glycolysis, glutaminolysis, the TCA cycle, respiratory chain, formation of metabolites, as well as cell growth and death. In particular, metabolic shift in the glucose utilization is considered in response to the change of glycolysis rate during the cultivation. A closed loop regulator is introduced to approximate the induction of enzyme pool during lag phase after inoculation. Based on the model, the specific consumption rate of substrate, the specific growth rate, the specific acetyl CoA formation rate as well as the specific oxygen uptake rate are estimated. The specific substrate consumption rate and the specific growth rate are then coupled into a bioreactor model, which delivers the key variables, i.e., cell density, substrate and metabolite concentrations. Batch and fed-batch cultivations were used to validate the model.

To track the rapidly changing temperature profiles of thermal cycling of polymerase chain reaction (PCR) accurately, an innovative feedforward variable structural proportional-integral-derivative (FVSPID) controller was developed. Based on the step response test data of the heat block, a reduced first order model was established at different operating points. Based on the reduced model, the FVSPID controller combined a feedforward path with the variable structural proportional-integral-derivative (PID) control. The modified feedforward action provided directly the optimal predictive power for the desired setpoint to speed up the dynamic response. To cooperate with the feedforward action, a variable structural PID was applied, where the P mode was used in the case of the largest errors to speed up response, whereas the PD mode was used in the case of larger errors to suppress overshoot, and finally the PID mode was applied for small error conditions to eliminate the steady state offset. Experimental results illustrated that compared to the conventional PID controller, the FVSPID controller can not only reduce the time taken to complete a standard PCR protocol, but also improve the accuracy of gene amplification.

Support vector machines (SVM) are applied to the prediction of key state variables in bioprocesses, such as product concentration and biomass concentration, which commonly play an important role in bioprocess monitoring and control. A so-called rolling learning-prediction procedure is used to deal with the time variant property of the process, and to establish the training database for the SVM predictor, which is characterized with the rolling update of the training database. As an example, product concentration in industrial penicillin production is predicted, and a comparison is also made with three different artificial neural network architectures (FBNN, RBFN, and RNN). The test results indicate that a prediction accuracy of 1-3 % can be obtained for 4-40h ahead prediction using the SVM, which is better than the best of the three artificial neural networks (ANNs). Moreover, for noise-added training highly noisy data or small-sample learning, the SVM also clearly outper-forms FBNN, RBFN, and RNN.

We report a convenient and highly efficient method for the synthesis of beta-organotellurobutenolides by the aryltellurenyl halides-induced electrophilic tellurolactonization of alpha-allenoic acids under mild conditions. The resulting beta-organotellurobutenolides can be utilized as precursors for versatile butenolide derivatives through a substitution reaction with organocuprate reagent or Pd/Cu(I)-catalyzed cross-coupling with terminal alkyne.

Constant specific growth rate control in the methanol growth phase was investigated for the fed-batch cultivation of Pichia pastoris expressing recombinant human serum albumin (rHSA). The methanol feeding strategy was determined based on an earlier proposed macrokinetic model to maintain the specific growth rate at preset levels. The experimental results demonstrate that the control strategy of constant specific growth rate is more effective than that of constant feeding rate to maximize production. Furthermore, the most productive setpoint of the specific growth rate is found between 0.005 and 0.006 h−1, which yields protein concentrations higher than 5 g l−1 at 160 h. In addition, a setpoint of 0.008 h−1 is suggested as the upper limit for specific growth rate control for the given expression system. Copyright © 2005 Society of Chemical Industry

A new method for spike sorting is proposed which partly solves the overlapping problem. Principal component analysis and subtractive clustering techniques are used to estimate the number of neurons contributing to multi-unit recording. Spike templates (i.e. waveforms) are reconstructed according to the clustering results. A template-matching procedure is then performed. Firstly all temporally displaced templates are compared with the spike event to find the best-fitting template that yields the minimum residue variance. If the residue passes the χ 2 -test, the matching procedure stops and the spike event is classified as the best-fitting template. Otherwise the spike event may be an overlapping waveform. The procedure is then repeated with all possible combinations of two templates, three templates, etc. Once one combination is found, which yields the minimum residue variance among the combinations of the same number of component templates and makes the residue pass the χ 2 -test, the matching procedure stops. It is unnecessary to check the remaining combinations of more templates. Consequently, the computational effort is reduced and the over-fitting problem can be partly avoided. A simulated spike train was used to assess the performance of the proposed method, which was also applied to a real recording of chicken retina ganglion cells.

Pichia pastoris has been widely used as a host for the expression of heterologous proteins. This paper deals with the glycerol and methanol optimal feeding control in the cultivation of P. pastoris expressing recombinant human serum albumin. An earlier proposed macrokinetic model and a cell cycle model were used to determine the substrate feeding profiles. In the glycerol growth phase, the goal of substrate feeding control is to obtain high cell density as well as the desired level of the fraction of budding cells, while in the methanol growth phase, the goal of substrate feeding is to control the specific growth rate at various set-points. The effect of both glycerol and methanol feeding control strategies were evaluated experimentally. It was found that the productivity may be enhanced if the fraction of budding cells at the end of the glycerol growth phase is minimized and if the specific growth rate during the methanol phase controlled between 0.005 and 0.006 h-1.

A structured kinetic model, which takes account of major metabolic pathways of glycerol and methanol in Pichia pastoris, is presented. Based on the combined kinetic and bioreactor model, feeding profiles of methanol are detennined with the aim of maintaining constant specific growth rate during production stage. Compared with the decreasing type of specific growth rate resulted from constant feeding profile in the standard protocol, the constant specific growth rate is believed to be advantageous for improving the productivity. Experimental results indicate that simulations of biomass and protein concentration agree well with the measured data, and the specific growth rates were successfully controlled at various set points.

To control the surface temperature on a steel ingot is of practically important task during the heat-treating process. However, it is often difficult to carry out a direct measurement in heating furnace in a long run due to the high temperature. In this work, a study on the surface temperature distribution of a steel ingot is carried out based on a three-dimensional CFD simulation. And then a temperature predictive model is proposed with the aim to predict the surface temperature on the steel ingot more accurately.

In mycelia cultivation, the moiphology of filamentous microorganisms changes greatly along with the cultivation time, which causes variations in mass transfer and metabolic behavior of the microorganisms. To give an insight into the morphology-associated time-variant process dynamics in a simple way, a cell age model is proposed for Penicillium chrysogenum fed-batch cultivation. With help of this model the average ages of the segregated cell populations may be estimated. The age model is further combined with a segregated penicillin model found in the literature. The combined model system seems to work well which is shown by the verification results with commercial penicillin fed-batch cultivations.

A rolling learning-prediction approach based on neural networks is proposed with the aim of on-line prediction of the product formation. Commercial-scale penicillin cultivations were taken as an example to test the product predictor. Raw data are pretreated in such a way that each input vector of the neural network consists of a series of time-discretised values on a specified transient of process variables. The output vector is composed of the amount of product at the next one and two prediction steps. The process variables involved in the predictor include carbon dioxide and product formation as well as oxygen, precursor and substrate consumption. Accumulated rather than instant values of these variables were used. A simple three-layer feedforward backpropagation neural network with a tangent sigmoidal transfer function in the hidden nodes and a linear one in the output nodes was used as the main frame of the product predictor. The proposed prediction procedure is called rolling learning-prediction because the training database is updated after each sampling interval and the learning-prediction is repeated thereafter. The robustness of the predictor was illustrated by its adaptive ability to widely scattered data sets and extra added noises. The testing results indicated that a prediction accuracy of 2–5% could be generally expected in the later phase of cultivation and reliable prediction time spans may take more than 10% of the cultivation period for penicillin production. An intrinsic problem of using neural networks—occasional trap of the network in bad local minima—is automatically detected and remedied. In addition, it was illustrated by example that the prediction error signal may be potentially used to detect extraordinary charges caused, for example, by contamination. Problems associated with the industrial application of the predictor are discussed.

In antibiotics industry, the titre in bioreactors is the most important process variable both for process supervision and scheduling. It is therefore of great significance to develop a software sensor to predict the product formation. In this contribution, a pseudo dynamic product predictor based on artificial neural network is designed. The input process variables of the predictor include substrates, precursor, nutrients and oxygen consumption, carbon dioxide evolution and also the product formation. These process variables are usually available in practice. Only the accumulated values of these variables rather than the instant ones are used. A feedforward neural network is chosen for one-step-ahead prediction of total product titre at the next step. The input vector of the neural network is the time scries of process variables over a predetermined time period. The database for network training is composed of a series of such vectors taken from historical charges which are processed as well as updated by a moving-data-window’s technique. The software sensor is tested by data of three industrial charges and ten model “generated” batches. MATLAB is used for training and testing of the neural network. The predictor is expected to be applied in optimal fermentation time scheduling for antibiotics production. Improvements and further development of the software sensor for real time applications are discussed.

The goal of bioprocess-control and-scheduling is to maximize the profit of the whole process. Maximizing of the profit implies the optimal coordination among the subindices such as productivity, yield and quality of the product. These three subindices often interact with each other since the manipulating variables may influence them simultaneously. Therefore, the best way to control the process is to optimize the profit index directly. In this paper, it is illustrated by an example of industrial penicillin fed-batch production how the profit index, named as profit function, is established. The profit function is an objective description of a process under the aspect of economy. In reality, it may be found that the profit function is often different from batch to batch. Such a fluctuation is demonstrated by gibberellic acid, penicillin and cephalosporin fed-batch processes. This suggests that optimal scheduling of process time may be an effective way to enhance the total profit. Based on a database which consists of the characteristic description of the profit function of earlier batches, the procedures of classification and optimal scheduling decision are described in details. On-line profit optimization may be done in particular if a process model is available. This is demonstrated with cephalosporin C fed-batch cultivation as an example.

An on-line profit optimization approach for baker’s yeast continuous fermentation was developed. The profit optimization was done based on the process model system and profit function which were earlier established. An on-line rolling identification method with moving-data-windows technique was used to modify the model system so as to adapt the different transients caused by different initial conditions from run to run and the time-varied dynamics of cell metabolism. With the help of such an adaptive process model, optimal molasses feeding strategies were determined by maximizing the profit function. Another primary manipulating variable, the process water feeding rate, was found to have no significant influence on the profit optimization within its normal operation range.

In baker's yeast continuous fermentation process, profit optimization is essentially the final target of all process control and scheduling measures. In this paper, an earlier proposed structured kinetic model for baker's yeast Saccharomyces cerevisiae was tested and modified with industrial fermentation data. With the support of the process model system, the profit function of the whole process is estimated This makes a further profit optimization possible.

Modeling and simulation technology of biological reaction is the important method of theoretical studies and production control. However the greatest technical difficulties of biological reaction modeling is how to identify and discuss the undetermined parameters existing in the complex system by analyzing limited experimental data— to look for an intelligent optimization algorithm that can optimize multivariable and mult-extremum high dimensional nonlinear functions effectively. This thesis took the fed-batch fermentation model in the penicillin industrial production for example, based on the industrial control rolling simulation software, discussed the application characteristics and their improvement of the real model of four widely used optimization algorithms: Simplex Algorithm, Genetic Algorithm, Simulated Algorithm and Particle Swarm Optimization. Lastly, the thesis proposed a new combined optimization algorithm— Genetic-simplex Algorithm by analyzing the complementarity of SIM and GA. Proved by the results of industrial rolling simulation this algorithm has combined the advantages of Simplex Algorithm and Genetic Algorithm, possessing excellent optimization ability.

The cell cycling model (CCM) for S. cerevisiae proposed earlier is modified and tested with our own experimental data. Although the original CCM was well verified in steady states and exponential growth with data available in literature, some discrepancies between model predictions and experiments were found for the dynamics of fed-batch culture. The redistribution pattern of the age distribution of daughter cells is suggested as cause of the model error. With an exponential type of redistribution, instead of the original linear one, the model behaviour in transients is improved. The modified model was verified with data of fraction of budding cells and cell number for five fed-batch cultivations. The model agreed well with the experimental data. The simulation results suggest that the cell cycling process indeed is essentially in a pseudo-steady state during fed-batch cultivation, as was assumed in the model. Due to the strong correlation between the quality of baker's yeast and the state of the population in the cell cycling process, the model was applied to optimize the feeding rate of a fedbatch process with consideration of final product quality. The optimal feeding was used succesfully in a laboratory experiment, which demonstrates the validity of the model.

In fed-batch production of baker’s yeast the profile of the sugar flow must regard the Crabtree-Effect and the oxygen transfer capacity of the reactor to avoid loss of yield. In this paper, optimal fed-batch control strategies are under investigation which also consider the final product quality. The optimization is carried out based on a mathematical process model that consists of three parts: The structured growth model can predict the growth kinetics in dependence of the operating conditions of the reactor; the age-distribution model for the budding process provides the fraction of budding cells (FBC) of the yeast population which is inversly correlated to the quality parameter fermentative activity; the cost model gives information on the economics of the process operation. New parts of the model are presented and simulation results with the model are compared with experimental data. The optimum operation of the Baker’s yeast fed-batch production is then investigated. Different from usual control trajectories, consideration of the quality criterion leads to a pulsed control of substrate flow. The optimum control predicted by the model was verified in laboratory-scale experiments.

A cell cycling model for unequal budding yeast Saccharomyces cerevisiae is proposed and verified by steady state data from experiments available in the literature. This model can be used to determine the relative fraction of the cells in any cycling phase or with any genealogical age during fermentation. As the quality of yeast is strongly influenced by the cycling process, the model could therefore be used to control the quality of the harvested yeast cells. The input of the cell cycling model is the specific growth rate μ, which is obtained from a metabolic model for S. cerevisiae proposed earlier. With this extended model system not only the quality control, but also the whole economical profit optimization can be carried out. Simulations were done to optimize the profit of a commercial scale baker's yeast production process by manipulating substrate feeding rate and substrate concentration under different aeration rates, fermentation periods and other conditions applied in industry.

A macrokinetic model employing cybernetic methodology is proposed to describe mycelium growth and penicillin production. Based on the primordial and complete metabolic network of Penicillium chrysogenum found in the literature, the modeling procedure is guided by metabolic flux analysis and cybernetic modeling framework. The abstracted cybernetic model describes the transients of the consumption rates of the substrates, the assimilation rates of intermediates, the biomass growth rate, as well as the penicillin formation rate. Combined with the bioreactor model, these reaction rates are linked with the most important state variables, i.e., mycelium, substrate and product concentrations. Simplex method is used to estimate the sensitive parameters of the model. Finally, validation of the model is carried out with 20 batches of industrial-scale penicillin cultivation.

A simple kinetic model is developed to describe the dynamic behavior of myeloma cell growth and cell metabolism. Glucose, glutamine as well as lysine are considered as growth limiting substrates. The cell growth was restricted as soon as the extracellular lysine is exhausted and then intracellular lysine becomes a growth limiting substrate. In addition, a metabolic regulator model together with the Monod model is used to deal with the growth lag phase after inoculation or feeding. By using these models, concentrations of substrates and metabolites, as well as densities of viable and dead cells are quantitatively described. One batch cultivation and two fed-batch cultivations with pulse feeding of nutrients are used to validate the model.

This article has been retracted consistent with Elsevier Policy on Article Withdrawal. The publisher apologises for any inconvenience this may cause.

A multi-staged population balance model is proposed to describe the cell cycle dynamics of myeloma cell cultivation. In this model, the cell cycle is divided into three stages, i.e., G1, S, and G2M phases. Both DNA content and cell volume are used to differentiate each cell from other cells of the population. The probabilities of transition from G1 to S and division of G2M are assumed to be dependent on cell volume, and transition probability from S to G2M is determined by DNA content. The model can be used to simulate the dynamics of DNA content and cell volume distributions, phase fractions, and substrate and byproduct concentrations, as well as cell densities. Measurements from myeloma cell cultivations, especially the FACS data with respect to DNA distribution and cell fractions in different stages, are employed for model validation.

A two pathway spatiotemporal model is proposed to describe the function of tonic suppressed-by-contrast cells of the cat retina. The model is able to describe the experimentally determined responses of such neurons to drifting sinusoidal gratings. It is also able to predict their responses to alternating sinusoidal gratings and flashing or moving spots of light, and these predictions resemble experimental observations, at least qualitatively. The model is physiologically plausible, it can be used to summarize the dynamic responses of the tonic suppressed-by-contrast cells of the cat and potentially to account for the responses of the suppressed-by-contrast cells of other species.

In thermal cycling of polymerase chain reaction, the process recipe is characterized by high ramp rates and short temperature holds. In order to reduce the time taken to complete a standard PCR protocol, a model-based hybrid control configuration is designed to rapidly track the thermal cycling recipe. The hybrid control configuration consists of feedforward, feedback and Bang-Bang actions, of which, the feedforward control is a model predictive control action and improves the dynamic performance of temperature tracking significantly. The model of the thermocycler is identified with step response data at different operating regions. Experimental results demonstrate that the hybrid control strategy performs better than the conventional PID controller both in setpoint response and in steady-state performance of temperature control, so that the accuracy and efficiency of PCR are improved accordingly.

The characteristics of the antibiotic fermentation process with parallel lines are analyzed. The assignment of tasks, units and events is described by two sets of binary variables. Based on continuous time representation, a mixed integer linear programming (MILP) model of the antibiotic fermentation process is built. The computational example of cephalosporin fermentation process demonstrates the feasibility and validity of the scheduling model

This paper proposes a cell age model forPenicillium chrysogenum fed-batch cultivation to supply a qualitative insight into morphology-associated dynamics. The average ages of the segregated cell populations, such as growing cells, non-growing cells and intact productive cells, were estimated by this model. A combined model was obtained by incorporating the average ages of the cell sub-populations into a known but modified segregated kinetic model from literature. For simulations, no additional effort was needed for parameter identification since the cell age model has no internal parameters. Validation of the combined model was performed by 20 charges of industrial-scale penicillin cultivation. Meanwhile, only two charge-dependent parameters were required in the combined model among approximately 20 parameters in total. The model is thus easily transformed into an adaptive model for a further application in on-line state variables prediction and optimal scheduling.

A multiple modeling algorithm for antibiotic fermentation process based on fuzzy c-means (FCM) and support vector regression (SVR) is proposed. By analyzing the features of antibiotic fermentation, the mechanism of multiple modeling of the bioprocess is presented. Using FCM clustering method, the bioprocess is classified into several work states and sub-models. Then, taking advantage of the generalization properties of SVR, the multiple model of bioprocess is established and the proposed algorithm is described. Experimental data of industrial penicillin production is used to validate the model.

The work outlined in this paper is towards developing a portable, economy blood glucose meter for self-monitoring of blood glucose, which has high performance-price ratio with multifunction and is attractive to the average people. Its functions include blood glucose concentration measurement, LCD display, historical records restore, calendar/timer. It enables the patients with diabetes to detect blood glucose with a glucose oxidase electrode independently at home. This paper will outline hardware configuration in the implementation and software management. Especially, graphic trend curve on LCD allows individuals to easily see how actual blood glucose readings compare with historical records stored in the meter and show how individuals stay within their specific target range so that they can determine how well they are controlling their diabetes. Experimental results indicate that this device is easy-to-use and suitable for home blood glucose monitoring for patients.

A macrokinetic model for Pichia pastoris expressing recombinant human serum albumin is proposed. The model describes the balances of some key metabolites, ATP and NADH, during glycerol and methanol metabolism. In the glycerol growth phase, the metabolic pathways mainly include phosphorylation, glycolysis, tricarboxylic acid cycle, and respiratory chain. In the methanol growth phase, methanol is oxidized to formaldehyde at first. Then, while a part of formaldehyde is oxidized to formate, the rest is condensed with xylulose-5-monophosphate to form glyceraldehyde-3-phosphate, and further assimilated to form cell constituents. The metabolic pathways following glyceraldehyde-3-phosphate were assumed to be similar to those in the glycerol growth phase. Based on the model, the macrokinetic bioreaction rates such as the specific substrate consumption rate, the specific growth rate, the specific acetyl-CoA formation rate as well as the specific oxygen uptake rate are obtained. The specific substrate consumption rate and the specific growth rate are then coupled into a bioreactor model such that the relationship between substrate feeding rates and the main state variables, i.e., the medium volume, the concentrations of the biomass, the substrate, and the product, is set up. Experimental results demonstrate that the model can describe the cell growth and the protein production with reasonable accuracy.

Spike sorting is a prerequisite for studying the multi-channel extracellular recordings. A new method is developed for spike sorting in multi-channel extracellular recordings of retinas. Based on the assumption that noise follows Gaussian distribution, spike templates are constructed based on principal component analysis and subtractive clustering, then a template matching procedure is applied to choose the best-fit template or templates combination for overlapping spikes so that after subtracting the template or templates combination from the spike event, the sum of squared residue follows Chi-square distribution. Artificial spike trains are used to assess the performance of the proposed method.

Robust controllers that ensure closed-loop pole placement are designed for uncertain discrete-time switched systems. A robust switching rule is proposed combined with the switched static output feedback controllers to ensure 1) quadratic stabilisation of the resulting closed-loop system, 2) pole location inside a circle for each linear mode of operation. The switching rule is state-dependent and does not rely on any uncertainties. All the results are expressed in terms of linear matrix inequalities (LMIs), which can be easily tested with efficient algorithms.

A macrokinetic model for a myeloma cell line is proposed. The model describes the dynamic balances of lactate, alanine, ATP and NADH during the metabolsim of glucose, glutamine and other amino acids. The metabolic pathways mainly include glycolysis, glutaminolysis, the trcicarboxylic acid cycle, the formation and utilization of amino acids, the respiratory chain, cell growth and cell death. The metabolic shift of glucose is especially considered because of a change in the rate of glycolysis. Thus the model functions in three modes to describe the behaviour of the myeloma cell line. On the basis of this model the macrokinetic bioreaction rates such as the specific substrate consumption rate, the specific growth rate, the specific acetyl-CoA formation rate, as well as the specific oxygen uptake rate, are estimated. The specific substrate consumption rate and the specific growth rate are then coupled into a bioreactor model such that the key variables, i.e., the cell density, the substrate and metabolite concentrations, are obtained. Experiments with batch and fed-batch cultures of a myeloma cell line (X63-Ag8.653) were used to validate the model. The prediction of the model was simulated by the rolling prediction approach.