A computer system solution for integration of a distributed bioreactor monitoring and control instrumentation on the laboratory scale is described. Bioreactors equipped with on-line analyzers for mass spectrometry, near-infrared spectroscopy, electrochemical probes and multi-array gas sensors and their respective software were networked through a real-time expert systems platform. The system allowed data transmission of more than 1800 different signals from the instrumentation, including signals from gas sensors, electrodes, spectrometer detectors, balances, flowmeters, etc., and were used for processing and carrying out a number of computational tasks such as partial least-square regression, principal component analysis, artificial neural network modelling, heuristic decision-making and adaptive control. The system was demonstrated on different cultivations/fermentations which illustrated sensor fusion control, multivariate statistical process monitoring, adaptive glucose control and adaptive multivariate control. The performance of these examples showed high operational stability and reliable function and meet typical requirements for production safety and quality. © 2003 Elsevier B.V. All rights reserved.

An electronic nose, a gas-phase multisensor system, was used to monitor precultivations of a recombinant tryptophan-producing Escherichia coli strain. The electronic nose signals showed a high correlation toward the main stages of the precultivations, namely, exponential growth, oxygen-limited growth, and glucose depletion. Principal component analysis (PCA) of the electronic nose signals was performed and shown to be useful for monitoring preculture progression. More importantly, PCA also allowed a qualitative assessment of the preculture performance during subsequent fed-batch cultivations. The electronic nose signals from the precultures showed, furthermore, a high correlation to the time of phosphate limitation and the tryptophan yield coefficient of the subsequent fed-batch cultivations, which allowed an accurate prediction of these process variables using partial least squares (PLS). The results demonstrate on data from 12 cultivations how the electronic nose can be a useful tool for the assessment of inoculum quality, thereby providing means of reducing batch-to-batch variation and increasing the productivity of bioprocesses.

Measurement data from an electronic nose (EN), a near-infrared spectrometer (NIRS) and standard bioreactor probes were used to follow the course of lab-scale yoghurt fermentation. The sensor signals were fused using a cascade neural network: a primary network predicted quantitative process variables, including lactose, galactose and lactate, a secondary network predicted a qualitative process state variable describing critical process phases, such as the onset of coagulation or the harvest time. Although the accuracy of the neural network prediction was acceptable and comparable with the off-line reference assay, its stability and performance were significantly improved by correction of faulty data. The results demonstrate that on-line sensor fusion with the chosen analyzers improves monitoring and quality control of yoghurt fermentation with implications to other fermentation processes. ⌐ 2002 Elsevier Science B.V. All rights reserved.

Multivariate statistical process control (MSPC) was for the first time applied to analyse data from a bioprocess on-line multi-analyser system consisting of an electronic nose (EN), a near-infrared spectroscope (NIRS), a mass spectrometer (MS) and standard bioreactor probes. One hundred and fifty sensor signals from the electronic nose, 1050 wavelength signals from the NIRS, carbon dioxide evolution rate calculated from mass spectrometer signals and standard bioreactor data (eg amount of substrate fed) were interrogated for their ability to model a bioprocess using MSPC. The models obtained were validated on a recombinant Escherichia coli fed-batch process for tryptophan production. Limiting trajectories were defined in the MSPC models for warning, action, and process experience with respect to biomass and tryptophan concentrations. The results showed the capacity and robustness of MSPC models for monitoring with multi-analysers and allowed a comparison of the different analysers' suitability for this kind of data processing. Furthermore, the results demonstrate that MSPC models provide a functional and versatile framework for coping with large information flows and are also suited to a variety of other bioprocessing monitoring and control tasks. ⌐ 2002 Society of Chemical Industry.

A multivariate bioprocess control approach, capable of tracking a pre-set process trajectory correlated to the biomass or product concentration in the bioprocess is described. The trajectory was either a latent variable derived from multivariate statistical process monitoring (MSPC) based on partial least squares (PLS) modeling, or the absolute value of the process variable. In the control algorithm the substrate feed pump rate was calculated from on-line analyzer data. The only parameters needed were the substrate feed concentration and the substrate yield of the growth-limiting substrate. On-line near-infrared spectroscopy data were used to demonstrate the performance of the control algorithm on an Escherichia coli fed-batch cultivation for tryptophan production. The controller showed good ability to track a defined biomass trajectory during varying process dynamics. The robustness of the control was high, despite significant external disturbances on the cultivation and control parameters.