Expert systems in bioprocess control: Requisite features

Konstantinov, Konstantin; Aarts, Robert J.; Yoshida, Takemi

doi:10.1007/bfb0007199

Cited by 13 publications

(3 citation statements)

References 29 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…), change of setpoints, and modification of the control algorithm parameters. Because of the complexity and the knowledge-intensive nature of the PS identification, expert system techniques are suitable for the design of the higher level (Konstantinov and Yoshida, 1992a,b;Konstantinov et al, 1993aKonstantinov et al, , 1994bOrr, 1992;Stephanopoulos et al:, 1993). The lower system level performs standard functions (such as pH, DO, and feeding control) and can be designed using the traditional control techniques.…”

Section: Architecture Of the Ps Control Systemmentioning

confidence: 99%

Monitoring and control of the physiological state of cell cultures

Konstantinov

2000

Biotechnol. Bioeng.

Self Cite

View full text Add to dashboard Cite

Advances in bioprocess engineering depends ultimately on the level of understanding and control of the physiological state of the cell population. Process efficiency is strongly influenced by changes in the cellular state which should be monitored, interpreted, and, if possible, properly manipulated. In most control systems this function is not explicitly considered, which hampers process development and optimization. Conventional control logic is based on direct mapping of the growth environment into process efficiency, thereby bypassing the cell state as an intermediate control objective. Today, this limitation is well realized, and explicit monitoring and control of cellular physiology are considered to be among the most challenging tasks of modern bioprocess engineering. We present here a generic methodology for the design of systems capable of performing these advanced monitoring and control functions.The term "physiological state" is quantified by a vector composed of several process variables that convey significant information about cellular state. These variables can be selected among different classes, including specific metabolic rates, metabolic rate ratios, degees of limitation, and others. The real-time monitoring of many of these is possible using commercial sensors. The definition and calculation of representative sets of physiological state variables is demonstrated with examples from several fermentor cultures: recombinant Escherichia coli for phenylalanine production, bioluminescent E. coli (harboring lux genes driven by a heat shock protein promoter) for detection of environmental pollutants, plant cell culture of Perilla frutescensfor anthocyanin production, and perfusion cultures of recombinant mammalian cells (NS0 and CHO) for therapeutic protein production.If the physiological state vector is on-line calculated, the fermentation process can be described by its trajectory in a space defined by the vector components. Then, the goal of the control system is to maintain the physiological state of the cell as close as possible to the trajectory, providing maximum efficiency. A control structure meant to perform this function is proposed, along with the mechanism for its design. In contrast to conventional systems which work in a closed loop in respect to the cell environment, this scheme operates in a closed loop in respect to the cell state. The discussed control concept has been successfully applied to the recombinant phenylalanine production, resulting in physiologically consistent operation, total computer control, and high process efficiency. Initial results from the application of the method to perfusion mammalian cell cultures are also presented.

show abstract

Section: Architecture Of the Ps Control Systemmentioning

confidence: 99%

Monitoring and control of the physiological state of cell cultures

Konstantinov

2000

Biotechnol. Bioeng.

Self Cite

View full text Add to dashboard Cite

show abstract

“…Experts' knowledge is expressed as linguistic conditional statements (i.e., if±then rules). Fuzzy logic inference has been used to estimate the model parameters (Aynsley et al, 1993;Konstantinov et al, 1993, Kosko, 1992Simutis et al, 1992Simutis et al, , 1993Sugeno, 1985). Fuzzy rules have also been combined with a parametric model, where the data is partitioned into several domains using expert knowledge (Zhang et al, 1994).…”

Section: Introductionmentioning

confidence: 99%

Hybrid model building methodology using unsupervised fuzzy clustering and supervised neural networks

Ronen

Shabtai

Guterman

2002

Biotech & Bioengineering

View full text Add to dashboard Cite

This paper suggests a model building methodology for dealing with new processes. The methodology, called Hybrid Fuzzy Neural Networks (HFNN), combines unsupervised fuzzy clustering and supervised neural networks in order to create simple and flexible models. Fuzzy clustering was used to define relevant domains on the input space. Then, sets of multilayer perceptrons (MLP) were trained (one for each domain) to map input-output relations, creating, in the process, a set of specified sub-models. The estimated output of the model was obtained by fusing the different sub-model outputs weighted by their predicted possibilities. On-line reinforcement learning enabled improvement of the model. The determination of the optimal number of clusters is fundamental to the success of the HFNN approach. The effectiveness of several validity measures was compared to the generalization capability of the model and information criteria. The validity measures were tested with fermentation simulations and real fermentations of a yeast-like fungus, Aureobasidium pullulans. The results outline the criteria limitations. The learning capability of the HFNN was tested with the fermentation data. The results underline the advantages of HFNN over a single neural network.

show abstract

“…More specifically, artificial intelligence techniques such as artificial neural networks or fuzzy logic have been developed and applied in the last two years. The particular feature of these techniques is the interpretation of expertise in the control system, which constitutes an "expert system" when a rule base is implemented (Konstantinov et al, 1993). However, only few works on the advantages of artificial intelligence techniques can be found in the literature for AWT processes (Marsili-Libelli et al, 1996;Steyer et al, 1997;Giraldo-Gómez and Duque, 1998;Pullammanappallil et al, 1998;Genovesi et al, 1999Genovesi et al, 2000).…”

Section: Introductionmentioning

confidence: 99%

Expert system for the on-line diagnosis of anaerobic wastewater treatment plants

Puñal

Rodríguez

Carrasco

et al. 2002

Water Science and Technology

View full text Add to dashboard Cite

A diagnosis system for anaerobic wastewater treatment processes is presented. The system is able to recognise the state and trend of the operation and suggest the appropriate control action. The on-line variables measured were gas flow rate and composition (methane and carbon monoxide), feed and recycling flow rates, temperature and pH, while the manipulable inputs are feed, recycling and buffer-addition flow rates. The diagnosis system comprises a structured rule base, incorporating expert knowledge using fuzzy logic features. The structure of the system is based on the classification of information related to the process in three categories: i) the state of the process, ii) its trend and iii) the recommended set-point values for the inputs manipulated: feeding, buffer and recycling pumps. The system was applied to diagnose the operation of a 1.1 m3 hybrid UASB-UAF treating wastewater from a fibreboard production factory under different conditions (overload and underload), corresponding to some of the typical causes of destabilisation in anaerobic wastewater treatment plants. These situations require control action in order to maintain the stability and the treatment capacity of the reactor. The application of the system developed for the purpose of managing the situation proved to be reliable for supplying the actual state and trend of the process, as well as the adequate set point values to recover stable operation and/or to avoid further destabilisation.

show abstract

Expert systems in bioprocess control: Requisite features

Cited by 13 publications

References 29 publications

Monitoring and control of the physiological state of cell cultures

Monitoring and control of the physiological state of cell cultures

Hybrid model building methodology using unsupervised fuzzy clustering and supervised neural networks

Expert system for the on-line diagnosis of anaerobic wastewater treatment plants

Contact Info

Product

Resources

About