Design of Nonlinear Model‐Based Control Using Bifurcation Analysis for Solution Polymerizations Carried Out in Lumped‐Distributed Reactors

In order to implement nonlinear control, nonlinear system identification must be performed, however, there are open questions concerning this field of process control, for example, experimental planning, model structure selection, parameter estimation, and validation. Therefore, the study of nonlinear model identification is a relevant unsolved problem that needs to be handled for nonlinear control synthesis. This paper presents the use of bifurcation theory, dynamic and stability analysis for nonlinear identification, and control of polymerization reactors. Peroxide‐initiated styrene‐solution polymerization reactors (lumped‐distributed) are investigated: batch, continuous stirred‐tank reactor (CSTR), and tubular reactors. Open and closed loop analyses are carried out using jacket temperature and weight average molecular weight setpoints as the bifurcation parameters. Phenomenological mathematical models, neural network nonlinear models, and an experimental data from a polymerization unit are employed for validating the proposed methodology in order to implement confident nonlinear controllers.     

Mass spectrometry and partial least-squares regression: a tool for identification of wheat variety and end-use quality

Rapid methods for the identification of wheat varieties and their end-use quality have been developed. The methods combine the analysis of wheat protein extracts by mass spectrometry with partial least-squares regression in order to predict the variety or end-use quality of unknown wheat samples. The whole process takes approximately 30 min. Extracts of alcohol-soluble storage proteins (gliadins) from wheat were analysed by matrix-assisted laser desorption/ionization time-of-flight mass spectrometry. Partial least-squares regression was subsequently applied using these mass spectra for making models that could predict the wheat variety or end-use quality. Previously, an artificial neural network was used to identify wheat varieties based on their protein mass spectra profiles. The present study showed that partial least-squares regression is at least as useful as neural networks for this identification. Furthermore, it was demonstrated that partial least-squares regression could be used to predict wheat end-use quality, which has not been possible using neural networks.     

Identification and discrimination of bacterial strains by laser induced breakdown spectroscopy and neural networks

A method based on laser induced breakdown spectroscopy (LIBS) and neural networks (NNs) has been developed and applied to the identification and discrimination of specific bacteria strains (Pseudomonas aeroginosa, Escherichia coli and Salmonella typhimurium). Instant identification of the samples is achieved using a spectral library, which was obtained by analysis using a single laser pulse of representative samples and treatment by neural networks. The samples used in this study were divided into three groups, which were prepared on three different days. The results obtained allow the identification of the bacteria tested with a certainty of over 95%, and show that only a difference between the bacteria can cause identification. Single-shot measurements were sufficient for clear identification of the bacterial strains studied. The method can be developed for automatic real time, fast, reliable and robust measurements and can be packaged in portable systems for non-specialist users.     

Separate Determination of Homologous Sodium Alkyl Sulfates with Ion-Selective Electrodes

The possibility of the separate determination of homologous sodium alkyl sulfates with ion-selective electrodes and using multisensor analysis was studied. The optimum component ratios were found for the analysis of model mixtures with ion-selective electrodes. It was shown that nonlinear regression analysis and artificial neural networks can be used for processing analytical signals from a set of weakly selective sensors for anionic surfactants.     

Using Unfold‐PCA for batch‐to‐batch start‐up process understanding and steady‐state identification in a sequencing batch reactor

In chemical and biochemical processes, steady‐state models are widely used for process assessment, control and optimisation. In these models, parameter adjustment requires data collected under nearly steady‐state conditions. Several approaches have been developed for steady‐state identification (SSID) in continuous processes, but no attempt has been made to adapt them to the singularities of batch processes. The main aim of this paper is to propose an automated method based on batch‐wise unfolding of the three‐way batch process data followed by a principal component analysis (Unfold‐PCA) in combination with the methodology of Brown and Rhinehart 2 for SSID. A second goal of this paper is to illustrate how by using Unfold‐PCA, process understanding can be gained from the batch‐to‐batch start‐ups and transitions data analysis. The potential of the proposed methodology is illustrated using historical data from a laboratory‐scale sequencing batch reactor (SBR) operated for enhanced biological phosphorus removal (EBPR). The results demonstrate that the proposed approach can be efficiently used to detect when the batches reach the steady‐state condition, to interpret the overall batch‐to‐batch process evolution and also to isolate the causes of changes between batches using contribution plots. Copyright © 2007 John Wiley & Sons, Ltd.     

A novel process for anaerobic composting of municipal solid waste

A novel process has been developed and evaluated in a pilotscale program for conversion of the biodegradable fraction of municipal solid waste (MSW) to methane via anaerobic composting. The sequential batch anaerobic composting (SEBAC) process employs leachate management to provide organisms, moisture, and nutrients required for rapid conversion of MSW and removal of inhibitory fermentation products during start-up. The biodegradable organic materials are converted to methane and carbon dioxide in 21–42 d, rather than the years required in landfills.     

An Analysis of Mixtures Using Amperometric Biosensors and Artificial Neural Networks

This paper presents a sensor system based on a combination of an amperometric biosensor acting in batch as well as flow injection analysis with the chemometric data analysis of biosensor outputs. The multivariate calibration of the biosensor signal was performed using artificial neural networks. Large amounts of biosensor calibration as well as test data were synthesized using computer simulation. Mathematical and corresponding numerical models of amperometric biosensors have been built to simulate the biosensor response to mixtures of compounds. The mathematical model is based on diffusion equations containing a non-linear term related to Michaelis–Menten kinetics of the enzymatic reaction. The principal component analysis was applied for an optimization of calibration data. Artificial neural networks were used to discriminate compounds of mixtures and to estimate the concentration of each compound. The proposed approach showed prediction of each component with recoveries greater that 99% in flow injection as well as in batch analysis when the biosensor response is under diffusion control.     

Hybrid neural modeling of bioprocesses using functional link networks

The objective of this work was to develop a model for an extractive ethanol fermentation in a simple and rapid way. This model must be sufficiently reliable to be used for posterior optimization and control studies. A hybrid neural model was developed, combining mass and energy balances with neural networks, which describe the process kinetics. To determine the best model, two structures of neural networks were compared: the functional link networks and the feedforward neural networks. The two structures are shown to describe well the process kinetics, and the advantages of using the functional link networks are discussed.     

A hybrid neural model of ethanol production by Zymomonas mobilis

A hybrid neural model was developed for the alcoholic fermentation by Zymomonas mobilis. This model is composed by the mass-balance equations of the process and neural networks, which describe the kinetic rates. Strategies that combines scarce experimental data with approximate models of the process were used to generate new data for the training of the networks, minimizing the number of experiments required. The proposed hybrid neural methodology uses all the information avail able about the process to deal with the difficulties in the development of the model.     

Modelling and prediction of retention in high-performance liquid chromatography by using neural networks

Summary Multi-layer feed-forward neural networks trained with an error back-propagation algorithm have been used to model retention behaviour of liquid chromatography as a function of the composition of the mobile phases. Conventional hydro-organic and micellar mobile phases were considered. Accurate retention modelling and prediction have been achieved using mobile phases defined by two, three and four parameters. With micellar mobile phases, the parameters involved included the concentrations of surfactant and organic modifier, pH and temperature. It is shown that neural networks provide a competitive tool to model varied inherent nonlinear relationships of retention behaviour with respect to the mobile phase parameters. The soft models defined by the weights of the networks are capable of accommodating all types of linear and nonlinear relationships, neural networks being specially useful when the relationships between retention behaviour and the mobile phase parameters are unknown. However, to train neural networks more experimental points than with hard-modelling methods are required, hence the use of the networks is recommended only for those cases where adequate theoretical or empirical models do not exist.     

Identification of combustible material with piezoelectric crystal sensor array using pattern-recognition techniques

A promising way of increasing the selectivity and sensitivity of gas sensors is to treat the signals from a number of different gas sensors with pattern recognition (PR) method. A gas sensor array with seven piezoelectric crystals each coated with a different partially selective coating material was constructed to identify four kinds of combustible materials which generate smoke containing different components. The signals from the sensors were analyzed with both conventional multivariate analysis, stepwise discriminant analysis (SDA), and artificial neural networks (ANN) models. The results show that the predictions were even better with ANN models. In our experiment, we have reported a new method for training data selection, 'training set stepwise expending method' to solve the problem that the network can not converge at the beginning of the training. We also discussed how the parameters of neural networks, learning rate eta, momentum term alpha and few bad training data affect the performance of neural networks.     

Combining convolutional neural networks and in‐line near‐infrared spectroscopy for real‐time monitoring of the chromatographic elution process in commercial production of notoginseng total saponins

The chromatographic elution process is a key step in the production of notoginseng total saponins. Due to quality variability of loading samples and resin capacity decreasing over cycle time, saponins, especially the five main saponins of notoginseng total saponins, need to be monitored in real time during the elution process. In this study, convolutional neural networks, one of the most popular deep learning methods, were used to develop quantitative calibration models based on in‐line near‐infrared spectroscopy for notoginsenoside R₁, ginsenosides Rg₁, Re, Rb₁ and Rd, and their sum concentration, with root mean square error of prediction values of 0.87, 2.76, 0.60, 1.57, 0.28, and 4.99 mg/mL, respectively. Partial least squares calibration models were also developed for model performance comparison. Results show predicted concentration profiles outputted by both the convolutional neural network models and partial least squares models show agreements with the real trends defined by reference measurements, and can be used for elution process monitoring and endpoint determination. To the best of our knowledge, this is the first reported case study of combining convolutional neural networks and in‐line near‐infrared spectroscopy for monitoring of the chromatographic elution process in commercial production of botanical drug products.     

Prediction of electrophoretic mobilities of sulfonamides in capillary zone electrophoresis using artificial neural networks

Artificial neural networks (ANNs) were successfully developed for the modeling and prediction of electrophoretic mobility of a series of sulfonamides in capillary zone electrophoresis. The cross-validation method was used to evaluate the prediction ability of the generated networks. The mobility of sulfonamides as positively charged species at low pH and negatively charged species at high pH was investigated. The results obtained using neural networks were compared with the experimental values as well as with those obtained using the multiple linear regression (MLR) technique. Comparison of the results shows the superiority of the neural network models over the regression models.     

Catechol determination in compost bioremediation using a laccase sensor and artificial neural networks

An electrochemical biosensor based on the immobilization of laccase on magnetic core-shell (Fe₃O₄–SiO₂) nanoparticles was combined with artificial neural networks (ANNs) for the determination of catechol concentration in compost bioremediation of municipal solid waste. The immobilization matrix provided a good microenvironment for retaining laccase bioactivity, and the combination with ANNs offered a good chemometric tool for data analysis in respect to the dynamic, nonlinear, and uncertain characteristics of the complex composting system. Catechol concentrations in compost samples were determined by using both the laccase sensor and HPLC for calibration. The detection range varied from 7.5 × 10^–7 to 4.4 × 10^–4 M, and the amperometric response current reached 95% of the steady-state current within about 70 s. The performance of the ANN model was compared with the linear regression model in respect to simulation accuracy, adaptability to uncertainty, etc. All the results showed that the combination of amperometric enzyme sensor and artificial neural networks was a rapid, sensitive, and robust method in the quantitative study of the composting system. Figure Structure of the magnetic carbon paste electrode used in the electrochemical biosensor     

Development of an electronic nose to identify and quantify volatile hazardous compounds

A new electronic nose was developed to identify the chemical compound released when a 2.5-L flask was broken inside a 3 m x 3 m x 2.5 m store-room. Flasks of 10 different hazardous compounds were initially present in the room: ammonia, propanone, hexane, acetic acid, toluene, methanol, tetrachloromethane, chloroform, ethanol and dichloromethane. Besides identification, quantification of the compound present in the air was also performed by the electronic nose, in order to evaluate the risk level for room cleaning. An array of six sensors based on coated piezoelectric quartz crystals was used. Although none of the individual sensors was specific for a single compound, an artificial neural network made it possible to identify and quantify the released vapour, among a series of 10 compounds, with six sensors. The neural network could be simplified, and the number of neurons reduced, provided it was used just for the identification task. Quantification could be performed later using the individual calibration of the sensor most sensitive to the identified compound.     

Identification of rhubarbs by using NIR spectrometry and temperature-constrained cascade correlation networks

Temperature-constrained cascade correlation networks (TCCCNs) were used to identify powdered rhubarbs based on their near-infrared spectra. Different network configurations that used multiple network models with single output (Uni-TCCCN) and single networks with multiple outputs (Multi-TCCCN) were compared. Comparative studies were made by using Latin-partitions and leave-one-out cross-validation methods. Results showed that multiple networks with single output predicted generally better than single network with multiple outputs. Better results with TCCCN models were obtained compared with conventional back propagation neural networks (BPNNs). The effects of parameters on correct identification and parameter optimizations were discussed in detail. With optimized neural network training parameters, NIR spectra from powdered rhubarb samples were classified by a TCCCN model with 100% accuracy.