Search

Cognitive (or intelligent) models are often superior to mechanistic models for nonideal bioreactors. Two kinds of cognitive models--cybernetic and neural--were applied recently to fed-batch fermentation by Ralstonia eutropha in a bioreactor with optimum finite dispersion. In the present work, these models have been applied in simulation studies of co-cultures of R. eutropha and Lactobacillus delbrueckii. The results for both cognitive and mechanistic models have been compared with single cultures. Neural models were the most effective for both types of cultures and mechanistic models the least effective. Simulations with co-culture fermentations predicted more PHB than single cultures with all three types of models. Significantly, the predicted enhancements in PHB concentration by cognitive methods for mixed cultures were four to five times larger than the corresponding increases in biomass concentration. Further improvements are possible through a hybrid combination of all three types of models.

Large continuous flow bioreactors are often under the influence of noise in the feed stream(s). Prior removal of noise is done by filters based either on specific algorithms or on artificial intelligence. Neither method is perfect. Hybrid filters combine both methods and thereby capitalize on their strengths while minimizing their weaknesses. In this study, a number of hybrid models have been compared for their ability to recover nearly noise-free stable oscillations of continuous flow Saccharomyces cerevisiae cultures from aberrant behavior caused by noise in the feed stream. Each hybrid filter had a different neural network in conjunction with an extended Kalman filter (EKF). The choice of the best configuration depended on the performance index. All hybrid filters were superior to both the EKF and purely neural filters. Along with previous studies of monotonic fermentations, the present results establish the suitability of hybrid neural filters for noise-affected bioreactors.

Despite its many useful properties, microbial production of poly-beta-hydroxybutyrate (PHB) is not yet commercially competitive with synthetic polymers. One reason is inadequate optimization of the fermentation under industrial conditions. In this study, a physiologically reasonable and experimentally validated kinetic model for PHB synthesis by Ralstonia eutropha was incorporated into a dispersion model to simulate a large fed-batch bioreactor. Solutions of the model indicated that cell growth and PHB synthesis were maximum at Peclet numbers (Pe) between 20 and 30, representing limited finite dispersion. At these Peclet numbers, the optimum feed rates also showed lower consumptions of the substrates than at Pe=0. Since complete dispersion was also difficult to achieve in production-scale bioreactors, these results pointed to the possibility of exploiting controlled dispersion for productivity enhancement.

Poly(beta-hydroxybutyrate) or PHB is an important member of the family of polyhydroxyalkanoates with properties that make it potentially competitive with synthetic polymers. In addition, PHB is biodegradable. While the biochemistry of PHB synthesis by microorganisms is well known, improvement of large-scale productivity requires good fermentation modeling and optimization. The latter aspect is reviewed here. Current models are of two types: (i) mechanistic and (ii) cybernetic. The models may be unstructured or structured, and they have been applied to single cultures and co-cultures. However, neither class of models expresses adequately all the important features of large-scale non-ideal fermentations. Model-independent neural networks provide faithful representations of observations, but they can be difficult to design. So hybrid models, combining mechanistic, cybernetic and neural models, offer a useful compromise. All three kinds of basic models are discussed with applications and directions toward hybrid model development.

Pilot-scale and larger bioreactors differ from small laboratory-scale reactors in terms of a greater occurrence of noise and incomplete mixing of the broth. Conventional control tries to induce good mixing and to filter out the noise as completely as possible. As such an 'ideal' operation is difficult to achieve, recent work has tried to exploit the non-ideal features to improve the performance. Using artificial neural networks, the degree of mixing, the extent of filtering of noise and the distribution of plasmid copy number (in a recombinant fermentation) can be controlled effectively on-line. This strategy generates better productivities than well-mixed noise-free operations, which suggests that deviations from ideal behaviour should be gainfully harnessed and not suppressed.

The complexity of microbial processes, coupled with the effects of incomplete mixing and the influx of noise, makes it difficult to describe quantitatively the behavior of bioreactors under realistic conditions. By circumventing the need for phenomenological equations, neural networks offer viable descriptions of such systems. However, such networks also require large amounts of carefully screened data and long training schedules. They also have limited extrapolation capability and poor physiological fidelity. Since phenomenological models are derived from physical descriptions of cellular processes, combinations of such models with neural networks, and sometimes also fuzzy logic, have provided good portrayals of nonideal bioreactor behavior while reducing the weaknesses of both approaches. However, like neural networks themselves, different architectures are possible for hybrid neural models. The rationale, architectures and illustrative applications of hybrid models for biological reactors are discussed here, with the possibility of combining them with cybernetic models, exploiting cellular intelligence, to develop self-evolving intelligent systems for optimization and control of microbial processes. [ABSTRACT FROM AUTHOR]

Microbioreactors with immobilized cells or enzymes are normally packed uniformly. However, uniform packing is good only when the flow is uniform throughout the microtube. In real situations this is not true. For such cases it was shown recently that, although radial variations may be ignored, a topologically optimized longitudinal distribution of immobilized pellets generated higher outputs of glucoamylase by a recombinant strain ofSaccharomyces cerevisiae. However, topology optimization (or any equation-based method) requires a good (and preferably simple) mathematical model. Such models are difficult to formulate for real nonideal conditions. Hence, a recent article by this author employed an expert system that selected between two artificial intelligence (AI) methods to determine the better method over successive intervals of time. However, such a switching strategy is complex and practically difficult. The present work, therefore, investigates a corresponding set of single (standalone) AI methods, each incorporating more biological information than earlier methods. For the sameS. cerevisiaestrain, it was seen that expressing the two input variables—feed concentration of glucose and the lengthwise distribution of immobilized cells—by genetic algorithms allowed greater flexibility and resulted in higher production rates of recombinant glucoamylase than all the methods used previously. This method is also easier to implement than topology optimization and the variable AI method. [ABSTRACT FROM AUTHOR]

Microbioreactors operated in real environments are often subject to noise from the environment. This is commonly manifested as fluctuations in the flow rates of the feed streams. Previous studies with larger bioreactors have shown that noise can seriously impair the performance. Given this possibility, the effects of noise on the performance of a microbioreactor have been analyzed for the trans-esterification of vinyl butyrate by 1-butanol by immobilized lipase B to produce butyl butyrate. As in previous work for macrobioreactors, the analysis was done with (i) no noise, (ii) unfiltered noise, and (iii) noise filtered by four different methods, and the fractal dimension of the product was used as an index of the performance. All fractal dimensions decreased with increasing dilution rates, and significant stochastic chaos was likely at low dilution rates. Of the four types of filters, the auto-associative neural filter (ANF) was the most effective in reducing chaos and restoring of smooth, nearly noise-free performance. The ANF also does not require a process model, which is a significant advantage for real systems. Simulations also revealed that even in the absence of noise, deterministic chaos is possible at low dilution rates; this underscores the importance of efficient filtering under such conditions when external noise too is present. The results thus establish the importance of noise in microbioreactor behavior and the usefulness of the fractal dimension in characterizing the effects. [ABSTRACT FROM AUTHOR]

Mechanistic models are often inadequate for non-ideal fermentation processes. This has been a major limitation in the commercial success of fed-batch fermentations for poly-β-hydroxybutyrate (PHB), a biopolymer that is potentially better than many synthetic polymers. A previous study with Ralstonia eutropha has shown that optimization through neural networks in place of mechanistic models can help to enhance PHB production substantially. Hybrid models combine the strengths of mechanistic and neural models and minimize their weaknesses. However, there is no unique hybrid neural model for a given application. Therefore, three fundamental hybrid designs have been compared with the neural and mechanistic optimizations done earlier for a fed-batch bioreactor for PHB production. Non-ideal features were introduced by adding Gaussian noise in the two feed streams--glucose and NH4Cl--and incorporating the optimum finite dispersion determined previously. All three hybrid designs were superior to the neural and mechanistic approaches. The best hybrid system, using a weighted combination of mechanistic and neural kinetic rates, generated 140% more of biomass and 330% more of PHB than conventional mechanistic optimization. This was achieved with reduced consumption of the two primary substrates. The supremacy of this hybrid system underlines the complexity of the PHB synthesis network and the merit in combining mechanistic and neural kinetics, and the performance enhancement achieved indicates the feasibility of such a system for an economically viable large-scale PHB fermentation. [ABSTRACT FROM AUTHOR]

In large-scale fermentations with oscillating microbial cultures, noise is commonly present in the feed stream(s). As this can destabilize the oscillations and even generate chaotic behavior, noise filters are employed. Here three types of filters were compared by applying them to a noise-affected continuous culture of Saccharomyces cerevisiae with chaotic oscillations. The aim was to restore the original noise-free stable oscillations. An extended Kalman filter was found to be the least efficient, a neural filter was better and a combined hybrid filter was the best. In addition, better filtering of noise was achieved in the dilution rate than in the oxygen mass transfer coefficient. These results suggest the use of hybrid filters with the dilution rate as the manipulated variable for bioreactor control. [ABSTRACT FROM AUTHOR]

Copyright of Canadian Journal of Chemical Engineering is the property of Wiley-Blackwell and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)