Optimizing control of a continuous stirred tank fermenter using a neural network

Feedforward neural networks are a general class of nonlinear models that can be used advantageously to model dynamic processes. In this investigation, a neural network was used to model the dynamic behaviour of a continuous stirred tank fermenter in view of using this model for predictive control. In this system, the control setpoint is not known explicitly but it is calculated in such a way to optimize an objective criterion. The results presented show that neural networks can model very accurately the dynamics of a continuous stirred tank fermenter and, the neural model, when used recursively, can predict the state variables over a long prediction horizon with sufficient accuracy. In addition, neural networks can adapt rapidly to changes in fermentation dynamics.List of Symbols F Dimensionless flow rate (F/ V₀) - F m³/h Flow rate - F ₀ m³/h Inlet flow rate - J Objective cost function - K _i Dimensionless constant in Eq. (3) (k _i /s₀) - k _i kg/m³ Substrate inhibition constant in Haldane model - k _m Dimensionless constant in Eq. (3) (k _s /s₀) - k _m kg/m³ Substrate inhibition constant in Haldane model - n prediction horizon - S Dimensionless substrate concentration (s/s₀) - s kg/m³ Substrate concentration - t h Time - v Dimensionless volume (V/V₀) - V m³ Liquid volume in fermenter - W _ij , W _jk Weight matrices in neural network - X Dimensionless biomass concentration - x kg/m³ Biomass concentration - Y Biomass/substrate yield coefficient - Weighting factor in Eq. (4) - Dimensionless specific growth rate (/ ) - 1/h Maximum specific growth rate - 1/h Specific growth rate - Dimensionless time ( t)     

Disturbance-accommodating control approach applied for the control of a continuous stirred tank reactor

This paper describes an experimental study of linear adaptive control to achieve the monitoring of a continuous stirred tank reactor. The practical control objective was the regulation of the substrate concentration at a pre-specified value in the process effluent despite local changes and/or culture physiology variations. The substrate concentration and the dilution rate have been selected as the controlled and the control variable respectively. The results obtained confirm that this approach offers the possibility to combine simplicity and effectiveness in bioprocess control.     

Biodesulphurization of mengen lignite by Rhodoccocus rhodochrous in a batch stirred and aerated tank fermenter

The biodesulphurization of Mengen lignite by a mesophilic bacterium, Rhodococcus rhodochrus ATCC 53968, was investigated in a batch stirred and aerated reactor. The experiments were carried out at 28°C with an inoculum percentage, initial pH, initial sodium acetate and lignite concentration of the biodesulphurization medium of 8% [v/v], 6.5 mM, 20 mM and 20 g/l, respectively. Variations in the sulphur contents of the lignite relative to the biodesulphurization period were monitored. The effects of the stirring and aeration rates on the removal of different sulphur forms from coal were investigated in the ranges 450–1,200 rpm and 0.1–0.53 vvm and the optimum values were found to be 500 rpm and 0.18 vvm, respectively. An increase in the total sulphur reduction with increasing biodesulphurization time was observed. The maximum total sulphur removal percentage was found to be 15.2% at 1,200 rpm after four days of incubation. The highest total sulphur removal rate was calculated on the second day of microbial desulphurization for each run. The total and organic sulphur contents of the coal after biodesulphurization were correlated with the stirring and aeration rates by using the non-linear least squares regression method. In the experimental runs lasting 8 days, the highest organic sulphur reducing percentage of 10.1% was obtained at a stirring rate of 500 rpm and an aeration rate of 0.40 vvm.     

Maximizing productivity of a continuous fermenter using nonlinear adaptive optimizing control

The control of a continuously operated fermenter at its maximum productivity level gives rise to a difficult control problem as the location of the optimum operating point changes due to the disturbances. In addition, the fermenter exhibits a change in the sign of the steady state gain near the optimum operating point. This study is aimed at developing an on-line optimizing control scheme that can track the changing location of the steady state optimum so as to maximize the fermenter productivity. A nonlinear Laguerre model, whose parameters are estimated on-line, is used for tracking the optimum operating point. The control at the optimum point is achieved using an adaptive nonlinear MPC strategy that uses the nonlinear Laguerre model for prediction. The efficiency of the proposed algorithm is demonstrated by simulating the control of a continuous fermenter that exhibits shift in the location of the optimum operating point in response to the changes in the maximum specific growth rate. The proposed on-line optimizing control strategy is shown to result in a considerable improvement in the closed loop performance even in the presence of measurement noise.     

Lipase-catalyzed selective synthesis of monolauroyl maltose using continuous stirred tank reactor

Monolauroyl maltose was synthesized by an immobilized lipase that catalyzed condensation of maltose and lauric acid in acetone using a batch reactor or a continuous stirred tank reactor. Mono- and di-lauroyl maltoses were identified by FT-IR, ¹H NMR, ¹³C NMR and MS. Monolauroyl maltose was selectively synthesized in a continuous stirred tank reactor and no diester was detected. The highest concentration of monolauroyl maltose at 28 mmol/l was obtained in 250 ml acetone when maltose was added at 4 g/d and the molar ratio of lauric acid to maltose was fixed at 4:1 at a flow rate of 0.15 ml/min for both influx and effluent without supplement of fresh molecular sieve.     

Multi-input multi-output control of a continuous fermenter using nonlinear model based controllers

Internal Model Control (IMC) and Model Predictive Control (MPC), the two most important members of model based controllers, are favourable alternatives for control of nonlinear processes. However, the performance of these controllers deteriorates drastically in the presence of substantial process-model mismatch. Hu and Rangaiah (1998) proposed feedback augmentation for nonlinear IMC (hence named Augmented IMC, AuIMC) for improving control in the presence of modelling errors, and demonstrated its success on a neutralization process. In the present study, IMC, MPC and AuIMC strategies are tested in a more difficult case of multi-input multi-output (MIMO) operation of a highly nonlinear continuous fermenter. A new control configuration is introduced as the conventional configuration is not applicable. Simulation results for different modelling errors show that IMC is better than MPC for fermenter control. The advantage of augmentation as in AuIMC manifests in the significantly improved regulatory control of the fermenter.     

Operational stability of immobilized d-glucose isomerase in a continuous feed stirred tank reactor

d-Glucose isomerization has been studied using immobilized cells of Streptomyces phaeochromogenes in a continuous feed stirred tank reactor (CSTR) where the external film diffusion resistance was negligible. Experiments conducted with various sizes of enzyme particles indicated that a strong internal diffusion resistance improved the apparent stability of these particles. The performance equations of the CSTR were constructed by associating the material balances for the inside porous support matrix with the bulk liquid phase, and enzyme deactivation was also taken into consideration. An iterative method together with the orthogonal collocation method is proposed for the evaluation of effectiveness factor and the substrate concentration profile within the enzyme particles. The numerical results offer an alternative analytical proof for the observation that under strong internal diffusion control the apparent operational stability of immobilized enzyme is improved.     

Consecutive immobilized enzymatic reactions in continuous stirred tank reactor systems

The performance characteristics of two-enzyme reaction in a continuous stirred tank reactor (CSTR) are analytical investigated in this work. A model is formulated to describe the substrate concentration variations by taking into account the external and internal diffusion resistances. It is found that the reaction system exhibits the characteristics of reaction control or diffusion control depending on the operating conditions. The single CSTR model is also extended to describe the multiple CSTR system. The latter model enables the prediction of the number of CSTRs in series required to achieve a prescribed substrate conversion.     

Production of optically pure L-alanine in batch and continuous stirred tank reactors using immobilized Pseudomonas sp.

A process to obtain optically pure l-alanine has been developed using batch and continuous stirred tank reactors with a new l-aminoacylase-producing bacterium Pseudomonas sp. BA2 immobilized in calcium alginate beads coated with glutaraldehyde. The maximum production of l-alanine in a continuous stirred tank reactor was 11.26 g after 2 days of operation which is higher than that previously reported.     

Optimum design of a series of continuous stirred tank reactors containing immobilised growing cells

The optimum design of a series of continuously operated stirred-tank reactors containing immobilised growing cells is described. Optimal design is defined as the minimal total holding time over the reactor series to achieve a certain degree of conversion. The analysis is made under the assumptions that there is a constant and equal concentration of immobilised biomass in all bioreactors along the series, no diffusion limitation takes place, all growth of the immobilised biomass will lead to an increase in suspended biomass, and that maintenance of the immobilised cells can be neglected. It is shown that the use of more than three bioreactors in series is likely to be de trop.     

Bacterial monitoring by molecular tools of a continuous stirred tank reactor treating textile wastewater

This study was performed to examine the effect of the bacterial diversity changes on the performances of a continuously stirred tank reactor (CSTR) treating textile wastewater. The molecular fingerprint established using polymerase chain reaction-single stranded conformation polymorphism (PCR-SSCP) methods showed that bacterial community profiles changed simultaneously with the increase of the wastewater loading rates (WLR). For the two WLR of 0.28 g l(-1)d(-1) and 0.37 g l(-1)d(-1), the reactor maintained good performances, suggesting that the large bacterial community present a high specific activity. The increase of the WLR from 0.37 to 1.12 g l(-1)d(-1) decreased the colour and the chemical oxygen demand (COD) removal efficiencies from 90% to 55% and from 85% to 30%, respectively, explained by the decrease of the bacterial diversity and activity. The changes of the bacterial dominance had no affect on the reactor performances. However, the decrease of the bacterial diversity significantly affected the colour and the COD removal efficiencies. It should conclude that indigo dye-containing textile wastewater treatment required the concerted activity of multiple bacterial populations.     

Optimization of operational conditions for adipate ester synthesis in a stirred tank reactor

Esterification of adipic acid and oleyl alcohol in a solvent-free system featuring a stirred tank reactor containing commercially immobilized Candida antarctica lipase B was performed. The process was carried out using an artificial neural network (ANN) trained by the Levenberg-Marquardt (LM) algorithm. The effects of four operative variables, temperature, time, amount of enzyme, and impeller speed, on the reaction yield were studied. By examining different ANN configurations, the best network was found to consist of seven hidden nodes using a hyperbolic tangent sigmoid transfer function. The values of the coefficient of determination (R²) and root mean squared error (RMSE) between the actual and predicted responses were determined to be 1 and 0.0058178 for training and 0.99467 and 0.622540 for the testing datasets, respectively. These results imply that the developed model was capable of predicting the esterification yield. The operative variables affected the yield, and their order of contribution was as follows: time > amount of enzyme > temperature > impeller speed. A high percentage of yield (95.7%) was obtained using a low level of enzyme (2.5% w/w), and the temperature, time, and impeller speed were 66.5°C, 354 min (about 6 h), and 500 rpm, respectively. A simple protocol for efficient substrate conversion in a solvent-free system evidenced by high enzyme stability is indicative of successful ester synthesis.     

Hydrogen metabolic patterns driven by Clostridium-Streptococcus community shifts in a continuous stirred tank reactor

Applied Microbiology and Biotechnology - The hydrogen (H2) production efficiency in dark fermentation systems is strongly dependent on the occurrence of metabolic pathways derived from the...     

Conversion of benzylpenicillin to 6-aminopenicillanic acid in a batch reactor and continuous feed stirred tank reactor using immobilized penicillin amidase

A rate equation has been derived to describe the hydrolysis of benzylpenicillin to 6-aminopenicillanic acid by penicillin amidase. The integrated from of the rate equation has been shown to predict satisfactorily the progress of the reaction in a batch reactor using either soluble or immobilized penicillin amidase. The rate equation was also used to predict the performance of a continuous feed stirred tank reactor containing immobilized enzyme. There was good agreement with experimental measurements.     

Transient behavior of a continuous stirred tank biological reactor utilizing phenol as an inhibitory substrate

Transient experiments were conducted on a Pseudomomas utilizing phenol in a continuous culture by disturbing the influent substrate concentration and dilution rate. Two stable steady states existed for some ranges of the parameters. Highly damped oscillations were observed in approaching a new high conversion steady state or in returning to a new high conversion steady state following a small disturbance. When a large disturbance was applied there was a smooth (overdamped) approach to a new low conversion steady state. The observed oscillatory behavior for small disturbances was predicted by a modified Powell-Ierusalemskii bottleneck model, but could not be predicted by a Monod-Haldane model; neither model was accurate for predicting the effect of large disturbances. A constant wall growth factor was used to account for microbial film activity, and the existence of two stable states was directly due to the presence of the film.     

Self-cycling fermentation in a stirred tank reactor

Self-cycling fermentations (SCFs) were conducted in a stirred tank apparatus using Bacillus subtilis and Acinetobacter calcoaceticus. The systems were very stable and the experiments lasted through many cycles. The variation of parameters such as biomass and doubling time from cycle to cycle was small. The stirred tank reactor (STR) allowed a much better control of the working volume in the fermentor from cycle to cycle, compared to the cyclone column, and it was not necessary to make periodic corrections.The production of surfactin from B. subtilis was achieved without extending the cycle time. The harvested broth at the end of each cycle was allowed to remain in a secondary vessel, at ambient temperature, before being collected. It is exhaustion of the limiting nutrient which causes an increase in dissolved oxygen (DO). At this point, the computer, which constantly monitors the DO, triggered the harvesting sequence to end the cycle. Thus, the mature culture in the secondary vessel experienced appropriate conditions for the production of the secondary metabolite. Meanwhile, the next batch of cells was being grown in the primary reactor.The response of a gas analyzer on the effluent paralleled that of the DO measurements in the fermentor. These data for oxygen and carbon dioxide exhibited less noise than the DO readings. Either would be a more reliable parameter for feedback control of the SCF because the problem of fouling of the DO probe after extended runs of many cycles would be eliminated. (c) 1993 John Wiley & Sons, Inc.     

A Mathematical model for ethanol production by extractive fermentation in a continuous stirred tank fermentor

Extractive fermentation is a technique that can be used to reduce the effect of end product inhibition through the use of a water-immiscible phase that removes fermentation products in situ. This has the beneficial effect of not only removing inhibitory products as they are formed (thus keeping reaction rates high) but also has the potential for reducing product recovery costs. We have chosen to examine the ethanol fermentation as a model system for end product inhibition and extractive fermentation and have developed a computer model predicting the productivity enhancement possible with this technique together with other key parameters such as extraction efficiency and residual glucose concentration. The model accommodates variable liquid flowrates entering and leaving the system, since it was found that the aqueous outlet flowrate could be up to 35% lower than the inlet flowrate during extractive fermentation of concentrated glucose feeds due to the continuous removal of ethanol from the fermentation broth by solvent extraction. The model predicts a total ethanol productivity of 82.6 g/L h if a glucose feed of 750 g/L is fermented with a solvent having a distribution coefficient of 0.5 at a solvent dilution rate of 5.0 h(-1). This is more than 10 times higher than for a conventional chemostat fermentation of a 250 g/L glucose feed. The model has furthermore illustrated the possible trade-offs that exist between obtaining a high extraction efficiency and a low residual glucose concentration.