Band 2, Heft 3

The NASA low-temperature oxidation catalyst (Pt/SnO2), originally developed for space-based carbon dioxide laser applications has been recently adapted to address formaldehyde destruction in industrial smoke stack emission streams. A mathematical model is developed that can be used to correlate the observable chemistry occurring on the surfaces of a monolith with the volumetric flow rate of the gas and cross sectional area of catalyst surfaces as well as quantifying process design variables such as pressure or temperature of an exhaust gas stream.

Generic technology is presented for the optimal screening of reaction and separation process interactions. The proposed scheme enables the use of particularly rich superstructure formulations that are comprised of two types of generic synthesis units with flexible representation modes. A reaction unit enables a detailed representation of the reaction and a conceptual representation of separation systems is facilitated through separation task units. Possible process design interactions are embedded in the superstructure formulations as combinations of generic units. The design options are explored using stochastic optimisation techniques suitable for this class of problems. The synthesis scheme supports the decision making process in early process design stages and prepares problem definitions for later stages. Conceptual screening functionalities reveal design insights into the performance of complex reaction-separation systems based on the limited modelling information available early in design. This enables the identification and inclusion of the relevant design information into the superstructure formulations employed in the subsequent design stages.

We consider the problem of predicting the future behavior of wastewater treatment plant quality indicators by creating prediction models using historical plant data. One of the main aims of this work is to be able to predict plant operational situations in advance so that corrective actions can be taken in time. Sets of historical plant data, such as BOD, COD and TSS were collected for a local wastewater treatment plant in Doha, the capital of the State of Qatar. These variables characterize the performance of any wastewater treatment plant and can be considered as quality indicators of the plant performance. Data were collected over a period of 4 years for the influent and effluent streams of the station. The plant influent and effluent predictions were performed using different techniques. These include time-series analysis, where the ARIMA (Autoregressive Integrated Moving Average) model was implemented in this case, and two Artificial Neural Networks (ANN) algorithms, namely Adaptive Linear Neuron networks (ADALINE) and Multi-layer Feedforward (ML-FF) neural networks. The predictions from the three techniques were presented and compared. The ML-FF model predictions proved to be more reliable than that of the equivalent ARIMA predictions followed by the ADALINE predictions, particularly for the finial effluent stream variables.

The Local Municipal Corporation operates a centralized treatment facility (CTF) to facilitate effective and safe disposal of hazardous biomedical waste generated from various health care facilities (HCFs) in a region. Managing these wastes is a challenging task due to unpredictable variation in amount of waste received at a CTF. In this paper a mathematical model is presented that predicts the load on CTF as a function of bed occupancy of some representative hospitals in the region. Models are available to predict waste generated from a particular hospital in terms of bed occupancy; however, it is difficult to predict total waste received at CTF that also includes wastes from HCFs other than hospitals. To develop the model, data for the waste generated at source and for waste incinerated at CTF were collected for two consecutive years on daily basis. It was observed that two hospitals are representative of the rate of biomedical waste generation in the region from all HCFs. In two consecutive years, there were significantly different trends in overall waste incinerated at CTF. The reason for variation in waste load at CTF was observed to be due to uneven load on the HCFs in two consecutive years.

A more realistic numerical technique hereafter known as Hegde's Ultimate Numerical Technique (HUNT) is developed and demonstrated on a one dimensional and a two dimensional steady state diffusion problem of heat transfer. The available numerical methods developed are based on finite difference technique neglecting the contribution of higher order terms in Taylor series expansion of the function leading to an approximation and the error in the solution. In the present effort of the HUNT, the optimization of the partial derivatives leads to the elimination of the error and justifies the stability and the convergence of the solution. The HUNT procedure based on the interface theory developed by the author, is capable of providing the ultimate optimum solution to all the partial derivatives considered as decision vectors. Even though the HUNT is demonstrated on one dimensional and two dimensional steady state diffusion equations, it does not require rigorous efforts to apply it to three dimensional problems of fluid flow and heat transfer. As pilot exercises the HUNT is demonstrated on a one dimensional circular fin and a two dimensional plate to obtain the temperature distribution. The result is compared with the analytical method and the finite volume method for which the results are available in the literature. To the knowledge of the authors, HUNT is both different and a unique example of its kind.

Material reuse/recycle has gained much attention in recent years for both economic and environmental reasons. Process integration techniques for water network synthesis have evolved rapidly in the past decade. With in-plant water reuse/recycle, fresh water and wastewater flowrates are reduced simultaneously. In this work, linear programming and mixed integer linear programming models that include piping cost and process constraints are developed to retrofit an existing water network in a paper mill that was not originally designed with process integration techniques. Five scenarios are presented, each representing different aspects of decision-making in real process integration projects. The fifth scenario makes use of fuzzy optimisation to achieve a compromise solution that considers the inherent conflict between maximising water recovery and minimising capital cost for retrofit.

A fluidized bed reactor for phosphate precipitation and removal from wastewater is modeled according to a two-step procedure. The first modeling phase, based on the development of a thermodynamic model for the computation of phosphate conversion, previously presented elsewhere is not reported here. The second step is related to the reactor modeling in the core of this paper. The pellet reactor is modeled as a reactor network involving a set of elementary cells representing ideal flow patterns. All the potential solutions are imbedded into a superstructure and the modeling problem is expressed as a MINLP problem. The MINLP problem is solved by means of the GAMS package, first for two flow rate values corresponding to two experimental fluidized bed behaviours, and then for the two flow rates considered simultaneously. In each case, the problem consists in finding an output concentration as close as possible to the experimental output concentration. Three objective functions are studied. The results are compared with those of Montastruc et al. (2004) who used a different numerical procedure. Whatever the considered case, the solutions found are structurally simpler than the ones of Montastruc et al. (2004). A major assessment of this study is that the reactor efficiency can easily be deduced, without any precise knowledge of some key parameters such as the density and thickness of the calcium phosphate layer. Finally a last numerical study concerning the superstructure definition shows that too complex a superstructure does not provide significant refinements on the solution.

Proper plant control and operation has to be based on measurements of process data at different locations within the process. The design of the instrumentation network is therefore an important step in the design phase of a given process. An advanced instrumentation network design should not only allow process control on a minimum data basis but should enable the plant operator to gain more insights into the functionality and capabilities of his process. For this purpose in-process balancing based on a suitably configured instrumentation network is necessary. To make full use of the measurements the data has to be processed by data reconciliation methods. In the present work an earlier approach by Chmielewski et al. (2002) has been applied to solids processes. This paper shows how the selected approach can be adopted to the measurement and processing of distributed variables, e.g. particle size and densities. An optimal instrumentation with regard to investment cost can be found that ensures that all necessary data for the reconciliation problem can be collected. The application to solids processes is illustrated with the examples of a simple screening-crushing process and a hydrocyclone cascade.

The problem of the on-line estimation of the reaction heat in a continuous stirred tank reactor from temperature measurements is addressed in this paper. The proposed uncertainty observer is based on differential algebraic techniques, such that the algebraic observability condition of the uncertainty from noisy temperature measurements is easily verified and the observer structure is very simple, which lead to feasible implementation. The observer proposed is robust against noisy measurements and sustained disturbances. The good performance of the observer is shown by means of numerical simulations and is compared with a nonlinear Luenberger-type observer.

In the modeling of hindered-settling bed separators, the published separation mechanisms are based on differences of particle density and size distributions, without the details of the complexity of particles-liquid interactions. A fluid dynamic model for the separator is developed using the Euler-Lagrangian approach of Computational Fluid Dynamics (CFD). Fluid motion is obtained from solving the movement of liquid governing equations. The damping effect on flow patterns caused by the movement of particles resulting in liquid-particle coupling is included in the models. Effects of particle size, particle density compositions, feed rate, feed water flow rate, and upward fluidizing water flow rate, etc., are simulated in the 2-D separation model. Flow pattern effects on the separation of fine particles in the separators with center downward-flow and side cross-flow feed systems are investigated.

Control of multirate systems is a challenging problem due to several reasons such as increased complexity in the design with tighter performance specifications. In this work, an algorithm for multirate constrained predictive control (MCPC) is presented. The multirate predictive control system includes a multirate state estimator, which provides inter-sample estimates of state variables of the process from infrequent and slow measurements. Constraints are addressed rigorously in this framework. The proposed design method is verified via simulation as well as experimentation. The results of the multirate predictive control for temperature control of a stirred tank system are shown and compared with those of a proportional integral (PI) control system.

A computer package based on visual basic code is developed for single and multiple effect evaporation desalination systems (SEE and MEE). The package features design calculations of heat transfer area, power consumption, and costing. The package is user-friendly and is equipped with interactive menus for report and form printing, file saving and retrieving, help files, and tutorial. The computer code is designed to run a default design case and also allows the user to change variables within pre-specified practical ranges. A rigorous process model is used in the package, which is based on detailed fundamental material and energy balance equations, well-proven correlations for the heat transfer coefficients, physical properties, and thermodynamic losses. Illustrations of the package displays are presented together with a number of case studies. Model predictions generated from the code are validated against actual field data and they showed a very good agreement. Future plans involve development of design simulator for the multistage flash desalination (MSF) and reverse osmosis (RO).

The demanding legislation for vehicle emissions pushes the automotive industry to the commercialization of advanced exhaust systems. The careful design of each aftertreatment device fitted in exhaust line and the interactions with the engine management, makes modelling an important part of systems engineering for these applications. Almost all catalytic converter models that are found in the literature use tunable parameters which typically address the kinetic terms of chemical reactions. A great effort has been given to the definition of complex reaction schemes and mechanistic description of them in-vitro, in order to assist the advanced design of catalytic washcoats and explain their complex chemistry. This paper presents a systematic effort towards the modelling of advanced commercial exhaust treatment systems, using a unified approach for the different devices. The model used in this work describes the catalytic processes with global reactions, using apparent Arrhenius type rate expressions with L-H kinetics. The model is used to predict transient performance of commercial catalytic converters under legislated test conditions. During these tests the existing uncertainties and intrinsic variations of engine and catalyst operation make the robustness and flexibility of the model prominent targets. Comparison of computed and measured second-by-second emissions demonstrate a high level of modelling accuracy, adequate to fit the behavior of advanced commercial applications of three-way catalytic converters and NOx traps, using a minimum set of 3 tunable parameters for each device.

The deposition rates for a water spray on a plate have been studied in a rectangular box configuration to determine the effects of turbulence and droplet size in a relatively simple geometrical configuration, with the aim being to assess the ability of a CFD simulation to predict the actual wall deposition fluxes and the trends in these fluxes in such a geometry. The main air velocities that were used in this study were 1 m/s and 2 m/s inside the box. The results from the Computational Fluid Dynamics (CFD) simulations underpredicted the percentage deposition on the plate significantly (by about 75% and lower) compared with the experiments. The droplet size effects on the deposition rates were studied by varying the atomizer air pressure (air flow rate). The inlet turbulence level was changed by placing a mesh at the box inlet. The results of experiments and simulations showed similar trends when decreasing the droplet size. Increasing the turbulence intensity at the box inlet seems to result in underpredicting the actual amounts of deposition, since the movements of the fine droplets are not well predicted by RANS approaches, because the RANS approaches do not predict the gas-phase turbulence levels near the walls well. The other reasons for the underprediction in the simulations could be coalescence of droplets in the spray, dispersion of droplets due to turbulence and radial variations in droplet velocities and sizes in the spray.

Supercritical fluid extraction (SCFE) has been introduced as an excellent method especially for food and pharmaceutical industries. Trout powder was selected as a source of high amount of valuable materials such as essential fatty acids (FAs). In this work, a mathematical model is introduced to simulate the transitional behavior of the extraction process. This model is able to explain the variation of FAs concentration profile in the bulk and solid phases versus the time of extraction. Effects of main operational parameters such as pressure (280 to 340 bar) and temperature (310 to 326 K) on the extraction yields were investigated in the supercritical fluid extraction system of laboratory scale and the results were compared with that information obtained from the solution of the model. In order to describe solid-solvent interactions, a linear equilibrium relationship was used in the model. Partial differential mass balance equations for the solid and fluid phases were solved numerically using an implicit method. The effective diffusivity into the particle pores and mass transfer coefficient to the bulk phase were the model parameters and they were correlated at different temperatures and pressures by fitting the results of the model to the experiments. The axial diffusivity and equilibrium coefficient of solid-solvent were kept constant in whole range of the operational conditions and they were estimated as the model constants.

This research presents a model describing the behaviour of a non-Newtonian shear-thinning fluid during aseptic filling processes, in order to determine the influence of the behaviour of fluids on the performance of filling valves in aseptic beverage plants, mainly in terms of the time required to perform the filling process. The ultimate aim of the study is to explore the possibility of improving the accuracy of industrial filling processes, so as to be able to utilise them with high viscosity fluids.The numerical model, exploiting the Finite Elements Method (FEM), was designed using the commercial software Comsol Multiphysics, and validated by comparing the steady state predictions with outcomes of filling experiments performed in industrial laboratories. Hence, subsequent numerical simulations were performed to investigate the transition from laminar to turbulent flow for shear-thinning fluids under different pressure conditions, in 3D time-dependent configurations. Results of the simulations, performed on a low fat yoghurt, show that laminar flow subsists within the whole filling system when the Metzner-Reed Reynolds number at the inlet section of the valve is lower than approx 444.

When a paint film is applied to the underside of a horizontal substrate, instabilities will lead to the formation of pendant drops. For sufficiently thick coatings, the drops may even detach from the coating before the coating dries. But even for thin paint films, with sufficiently high evaporation rates and viscosity, drop formation will lead to coating irregularities with thick regions (corresponding to the center of the drop) adjacent to much thinner regions (corresponding to troughs between drops.) In this paper we present a mathematical and numerical model based on the lubrication approximations for the drying of paint films on the underside of horizontal substrates. The paint is modeled as a multicomponent liquid with one non-volatile and one volatile component, termed the ``resin'' and ``solvent'' respectively. Our model includes the effects of surface tension and gravitational forces, and we show that surface tension gradient effects due to solvent evaporation must be considered to correctly model the post-application flow of certain paints. Simulations are given for an array of parameters, and they show that surface tension gradients can dramatically affect the post-application behavior of a model paint film, generally causing the suppression and reversal in drop growth, consequently leading to a more even final coating layer.

The objective of this study is to simulate the green petroleum coke calcining processes using the simulation program HYSYS and using actual industrial data. Because counter-current mass flow is not allowed in the HYSYS program, the kiln was described by using fictive streams and unit operations. By the simulation, it is possible to predict the operating conditions that control the contents of undesirable impurities in the calcined petroleum coke, namely, sulfur, volatile matter and moisture contents. It also gives the desirable calcined petroleum coke properties such as density. Apart from the metal contents, the simulation allows the coke calciner to utilize any type of green coke regardless of the undesirable impurities. This is done without resorting to the costly blend from different types of green cokes. Simulation is also effective in controlling and optimizing calcination processes' variables. From the simulation it was found that it is possible to process any type of green coke for varying sulfur, volatile matter and water content by adjusting the amount of tertiary air and/or fuel. Two simulation cases were studied for low and high volatile matter contents of 8.5 and 14.7 wt% in the feed. Mass flow rates of fuel and tertiary air were both increased in the first case, while no fuel was required in the second case. The benefit from this is to reduce the operating costs.

In this paper a technique borrowed from modern control theory, called sliding mode observers, is used as the vehicle for nonlinear parameter estimation. The observer is a dynamic system derived from the original state equations that guides, in real time, parameter values from initial estimates to optimal ones. This is done by driving errors between the estimated and "data" states to zero. Observers are developed and demonstrated for the Monod model for biokinetics, which has been used widely in bioprocess design for both aerobic and anaerobic bio-processes under both toxic and non-toxic conditions. The Monod model has four parameters; often, two of the parameters (y, yield and b, decay coefficient) are assumed or calculated from other sources such as from thermodynamic considerations. The other two parameters (k, maximum specific uptake rate of the substrate and ks, the half saturation constant for growth) are determined from experimental data. The determination of these parameters is often performed using nonlinear regression or related methods and it has been demonstrated that these are particularly difficult parameter estimation problems. Examples are presented where only k is unknown and when both k and ks are to be determined. One case where the data are corrupted with noise is considered.The primary objective of the paper is to introduce the methodology of sliding mode observers for nonlinear parameter estimation to the chemical and process engineering communities.

A new simulation method for investigating mixed aqueous electrolyte systems at thermodynamic equilibrium is presented. Based on the commercial simulation package IPSEpro, a graphical procedure for assembling equilibrium systems followed by simultaneous solution of the underlying equation system was developed. At graphical setup, species and reactions objects are connected by the information of activity and stoichiometry in a network-like manner. Consequently, the resulting equations of mass action, the electro-neutrality and balance conditions are solved incorporating the Pitzer activity coefficient model and equilibrium constants from thermodynamic standard data. The method developed allows fast and flexible application of the complex thermodynamic principles to engineering tasks. The modeling approach and all basic models for graphical input and simultaneous solution of coupled equilibrium conditions are explained. For demonstration of the simulation method, example simulations for the systems of H2SO4-H2O, CaSO4-H2SO4-H2O and NaCl-CaSO4-CO2-H2O are described and evaluated successfully.

Flexible manufacturing and processing, which are becoming crucial for companies that wish to remain competitive in today's changing market environment, require the detailed modeling of process operations, as well as the incorporation of utility consumption modeling and energy integration strategies. In this paper, a comprehensive modeling environment for heat and power integration in batch and semi-continuous processes is presented. Moreover, a general procedure concerning utility generation and utility system design optimization is described. This framework has been tested and evaluated as part of an industrial case study. The results obtained indicate that the general design methodology developed is effective for process integration of time-dependent processes with combined heat and power.

The drying rate of a mushroom undergoing microwave-vacuum (MV) drying (MVD) was controlled by moisture dissipation and was dependent on vacuum pressure levels. The main objective of this work was to develop artificial neural network (ANN) model to predict moisture ratio of MV-dried mushrooms. One-hidden-layer feed-forward ANN models were trained and validated with experimental data. The Levenberg-Marquardt algorithm was utilized in regulating the ANN model weights and biases. Inputs for ANN models were vacuum pressure and drying time. Output from ANN models was moisture ratio at a given drying time. Reduced chi-square (X 2) and root mean square error (RMSE), and residual sum of squares (RSS) of the results from ANN models were calculated and compared with those of a modified Page's model (an experimental-based mathematical model), which is commonly used in the literature. The X 2, RMSE, and RSS of the ANN model (2.272 x 10 -5, 4.023 x 10 -3, and 3.204 x 10 -3, respectively) were found to be lower than those of the modified Page's model (6.692 x 10 -4, 2.561 x 10 -2, and 12.98 x 10 -2, respectively). These results indicate that the feed-forward ANN model represented the drying characteristics of mushrooms better than the modified Page's model. Therefore, the ANN model could be considered as a better tool for estimation of the moisture content of mushrooms than by the modified Page's model.

This paper presents the predictions of deposition patterns using CFD simulations based on transient-flow behaviour of a 1.6 m high, 0.8 m diameter, pilot-scale spray dryer, following from previous studies assessing the use of Computational Fluid Dynamics (CFD) simulations to predict the deposition on a plate in a simple box configuration. The predicted deposition fluxes here have been compared with experimental data for the deposition fluxes of skim milk, maltodextrin and water. The CFD simulation results suggested that the effect of transient air flows on the vertical patterns of deposition fluxes with distance up the dryer wall for no inlet air swirl is small. The CFD simulations underpredicted the experimental values of the deposition fluxes by approximately 50%, but the simulations predicted the same experimental trends when changing the main air flow rate through the dryer. The experimentally-measured deposition fluxes were 38%, on average, higher at a main air flow rate of 113 kg/h compared with those at a flow rate of 88 kg/h. The CFD simulations predicted an average increase in deposition flux of 26% at 113 kg/h compared with 88 kg/h, so the trends with this change in operating conditions have been predicted well by the CFD simulations. One-way particle coupling has therefore shown correct trends in the deposition fluxes with respect to both positions in the dryer and different operating conditions, and such one-way coupling is several orders of magnitude faster than the more rigorous two-way coupling.

This paper presents a Fault Detection and Isolation (FDI) method for stochastic nonlinear dynamic systems. Our contribution consists of showing another method of tackling the problem of the physical origin diagnosis of faults by combining the technique based on the innovations and the technique using the multiple Kalman filters for a nonlinear dynamic system strongly non-stationary. The usefulness of this combination is the implementation of all the faults dynamics if the decision threshold on the standardized innovation exceeds a fixed value. In the other case, one filter is enough to estimate the process state. An algorithm is described and applied to a perfectly stirred chemical reactor functioning in a semi-batch mode. In this paper, the chemical reaction used is an oxido reduction one, the oxidation of sodium thiosulfate by hydrogen peroxide.

A general expression which is useful for predicting the radial distributions or for analyzing and interpreting experimental data is derived for a liquid-solid circulating fluidized bed. The effect of both the radial liquid velocity and solids holdup distributions are taken into account in the analysis. Both effects are analyzed and are related to the operating variables of superficial liquid velocity, jl, superficial solids velocity, js, and cross-sectional average solids holdup. The results predicted by the analysis are compared with the experimental data obtained for various two-phase flow regimes, when different solids and bed dimensions are applied. The model predictions show good agreements with the experimental data and reasonable trends when different solids and bed dimensions are applied. Stronger non-uniformities in flow structures are found in larger size particles systems, with larger solids density and/or larger diameter columns. Radial solids holdup distribution is related with the cross-sectional average solids holdup. Good agreement with data and reasonable trends are observed.

Simulation of multi-stage flash (MSF) desalination processes that have a production capacity range between 50,000 to 75,000 m3/d is an intensive computational problem that requires high computer processing speed despite the availability of advanced processing computer power in hand nowadays. In this work, a comparative study is conducted to explore the performance of different numerical techniques to solve large set of nonlinear equations generated by large scale MSF models. These algorithms can be categorized into three groups, namely: conventional numerical approximation methods, multi-objective optimization based methods, and the last group comprises artificial neural networks (ANN) based models and genetic algorithms (GA) based methods. The problem of solving large sets of nonlinear equations with upper and lower constraints is accomplished successfully using all algorithms with different prediction efficiency and speed. The idea of using GA and ANN based algorithms in simulating the MSF model is basically used to generate feasible initial solution estimates that were used as starting guesses for other numerical methods in the former case and to eliminate the step of providing these initial guesses in the later case. Significant reduction of computation effort was attained using ANN-based techniques. The outcome of this work can be utilized to develop new generations of process simulators that are based on well-trained ANNs in order to achieve speedup of computations and to generate more reliable predictions without detracting from accuracy.

Dynamic simulation and optimization of a hydrogenation reactor system were developed and investigated in the present work. The process mainly consists of three adiabatic fixed bed hydrogenation reactors in series in addition to one heater before the first reactor and three coolers after each reactor for interstage cooling. The feed flow rate to the unit, the feed temperature or the carbon monoxide content of the feed may change and cause variations of the outlet temperature of reactors. Therefore, it is essential to control the inlet temperature of each reactor. There is a temperature controller before each reactor and also one after the third reactor. The tuning parameters of the controllers were optimized in three different cases, taking into account the process constraints. In each case, a special disturbance was forced to the process. Because of catalyst deactivation, the inlet temperature of reactors must be increased during the running period in order to have the desired conversion. Therefore, the optimal inlet temperature profile was determined based on SOR, MOR and EOR data. Optimization of the process was done by the use of a hybrid GA-SQP method. The new hybrid method was developed to overcome the difficulties of both methods. The genetic algorithm (GA) which is a stochastic method, is relatively slow, but is not sensitive to the initial point. In contrast, sequential quadratic programming (SQP) method is a deterministic method which is fast, but very sensitive to the initial point and gets trapped in local optima. In the newly developed hybrid method, the SQP method speeds the solving procedure, while the GA enables the algorithm to escape from local optima. An industrial acetylene hydrogenation system was used to provide the necessary data to adjust kinetics and to validate the approach.

The Morse potential energy function (PEF) is considered regarding the characterization of interaction forces of particles with tuning parameters. Phase coexistence of Morse fluids is predicted for different steepness and range of the PEF parameters using the grand-canonical transition matrix Monte Carlo (GC-TMMC) method, with quantification of the parameter S, which is the product of a constant with a unit of reciprocal length and the equilibrium distance between two molecules. We found that a lower limit of S exists bounded by infinite critical temperature. The critical properties of the vapor-liquid equilibrium curves are estimated using a rectilinear diameter method and a scaling law approach. A Clausius-Clayperon type relation of S and critical temperature is derived in this work. Vapor-liquid surface tension of Morse fluids by finite size scaling and GC-TMMC is also reported. Surface tensions are found to be higher at lower S.

Temperature (T) and moisture content (MC) of non-homogenous food undergoing microwave-vacuum (MV) drying (MVD) are directly dependent on microwave power, vacuum pressure, and the product's physical properties. A two-hidden-layer Artificial Neural Network (ANN) model was developed in an earlier study to predict temperature and moisture content of the product at a given time based on the present state of product conditions and process control parameters. This approach either provided lowest error in temperature prediction or in moisture content prediction but not the lowest error in both the prediction parameters simultaneously. The main objective of this work was to improve the performance of the ANN model for temperature and moisture content predictions in MV dried samples. Experimental data obtained from MVD of tomato slices at different drying conditions was normalized and divided into two groups for training and validating. The parallel dynamic ANN model consisted of two double-hidden-layer feed-forward ANN models with varying node numbers (10, 20, and 30). These models were separately trained, simultaneously for moisture content as well as temperature, with the Levenberg-Marquardt algorithm. Inputs for the ANN models were magnetron on-off status, vacuum pressure, temperature, and moisture content at time `ti'. The previous temperature and moisture content data at time `ti-1, i-2, …, i-n' where n = 0, 10, 20, and 30 were also added to the input layer. Outputs from the ANN models were temperature and moisture content at time `ti+1'. The results indicated that the dynamic ANN model working in parallel with the previous temperature and moisture content data provided results that are more accurate and required less training time than those of ordinary ANN models. Model simulation may supply essential information regarding temperature and moisture content of non-homogenous foods corresponding to microwave power and vacuum pressure levels to the predictive control system. Therefore, improved drying efficiencies and thermal damage prevention may be achieved.

Distillation columns are widely used in chemical processes and exhibit nonlinear dynamic behavior. In order to gain optimum performance of the distillation column, an effective control strategy is needed. In recent years, model based control strategies such as internal model control (IMC) and model predictive control (MPC) have been revealed as better control systems compared to the conventional method. But one of the major challenges in developing this effective control strategy is to construct a model which is utilized to describe the process under consideration. The purpose of this paper is to provide a review of the models that have been implemented in continuous distillation columns. These models are categorized under three major groups: fundamental models, which are derived from mass, energy and momentum balances of the process, empirical models, which are derived from input-output data of the process, and hybrid models which combine both the fundamental and the empirical model. The advantages and limitations of each group are discussed and compared. The review reveals a remarkable prospect of developing a nonlinear model in this research area. It also shows the discovery of new advance methods in an attempt to gain a nonlinear model that is able to be used in industries. Neural network models have become the most popular framework in nonlinear model development over the last decade even though hybrid models are the most promising method to be applied for future nonlinear model development.