Model-based Process Control Research Articles

Automation of the Casting Process using a model-based NC Architecture

Automation of the casting process in foundries is most commonly limited to foundry lines for series manufacturing. The employed teach-in method for programming the process is time-consuming and, therefore, only suited for large lot sizes. Thus, manufacturing of small lot sizes or single parts still necessitates the handwork and expertise of workers pouring the molten metal into the mould. This creates a dangerous environment for the worker. In recent years researchers have developed a model-based process control, which allows precise casting control without having to previously teach the process. Usage of these process models has been limited to special process control systems, specifically designed for a certain machine. In this paper, a novel approach is presented, in which a NC (Numerical Control) system is enhanced by process models. Instead of programming the NC in kinematic coordinates, it will be shown that programming in process coordinates allows robust and flexible casting, even for small lot sizes. The paper will prove the feasibility of the approach by evaluation of the processing time of the model.

Neural Network Predictive Controller Based Nonlinearity Identification Case Study: Nonlinear Process Reactor - CSTR

In the last decades, a substantial amount of research has been carried out on identification of nonlinear processes. Dynamical systems can be better represented by nonlinear models, which illustrate the global behavior of the nonlinear process reactor over the entire range. CSTR is highly nonlinear chemical reactor. A compact and resourceful model which approximates both linear and nonlinear component of the process is of highly demand. Process modeling is an essential constituent in the growth of sophisticated model-based process control systems. Driven by the contemporary economical needs, developments in process design point out that deliberate operation requires better models. The neural network predictive controller is very efficient to identify complex nonlinear systems with no complete model information. Closed loop method is preferred because it is sensitive to disturbances, no need identify the transfer function model of an unstable system. In this paper identification nonlinearities for a nonlinear process reactor CSTR is approached using neural network predictive controller. KEYWORDS Continuous Stirred Tank Reactor, Multi Input Multi Output, Neural Networks, Chebyshev Neural Networks, Predictive Controller.

Analytical Solution of Free Radical Polymerization: Derivation and Validation

An elegant, simple, and exact analytical solution (AS) was obtained for a large range of elementary steps with practical importance in free radical polymerization. The AS matches excellently with the numerical solution for the four cases of monomer–polymer systems studied ranging from the slowest to the fastest. It works equally well for different initiators, different initiator and monomer concentrations, presence or absence of solvent, various solvent volume fractions, and different temperatures. It also matches quite well with experimental data reported in the literature. This AS is not only in line with previous published solutions but also extends their applicability in a natural way. Overall, the conceptual correctness as well as predictive capabilities of the derived AS are established beyond doubt. This AS has the potential to be used in various practical applications such as model based process control, CFD simulations, and so forth.

Real-time simulation of temperature in hot rolling rolls

Model-based process control is widely used in metal working. Often, simplified real-time capable numerical models replace or support measurement. In this work, we investigate the simulation of temperature in hot rolling rolls based on a 3-D model of a roll which is fast enough to act as soft sensor during operation. The involved discretization of the heat equation in cylindrical space is done via simple finite volumes and propagation in time is done via a forward Euler method. This allows a parallel and efficient implementation for GPGPUs. We validate our code against analytic solutions and measurements, provide a detailed performance analysis, and show simulation results for realistic rolls.

Model‐based derivation, analysis and control of unstable microaerobic steady‐states—Considering Rhodospirillum rubrum as an example

Microaerobic (oxygen-limited) conditions are critical for inducing many important microbial processes in industrial or environmental applications. At very low oxygen concentrations, however, the process performance often suffers from technical limitations. Available dissolved oxygen measurement techniques are not sensitive enough and thus control techniques, that can reliable handle these conditions, are lacking. Recently, we proposed a microaerobic process control strategy, which overcomes these restrictions and allows to assess different degrees of oxygen limitation in bioreactor batch cultivations. Here, we focus on the design of a control strategy for the automation of oxygen-limited continuous cultures using the microaerobic formation of photosynthetic membranes (PM) in Rhodospirillum rubrum as model phenomenon. We draw upon R. rubrum since the considered phenomenon depends on the optimal availability of mixed-carbon sources, hence on boundary conditions which make the process performance challenging. Empirically assessing these specific microaerobic conditions is scarcely practicable as such a process reacts highly sensitive to changes in the substrate composition and the oxygen availability in the culture broth. Therefore, we propose a model-based process control strategy which allows to stabilize steady-states of cultures grown under these conditions. As designing the appropriate strategy requires a detailed knowledge of the system behavior, we begin by deriving and validating an unstructured process model. This model is used to optimize the experimental conditions, and identify properties of the system which are critical for process performance. The derived model facilitates the good process performance via the proposed optimal control strategy. In summary the presented model-based control strategy allows to access and maintain microaerobic steady-states of interest and to precisely and efficiently transfer the culture from one stable microaerobic steady-state into another. Therefore, the presented approach is a valuable tool to study regulatory mechanisms of microaerobic phenomena in response to oxygen limitation alone. Biotechnol. Bioeng. 2014;111: 734-747. © 2013 Wiley Periodicals, Inc.

Identification test design for multivariable model-based control: An industrial perspective

The design of plant tests to generate data for identification of dynamic models is critically important for development of model-based process control systems. Multivariable process identification tests in industry continue to rely on uncorrelated input signals, even though investigations have shown the benefits of other input designs which lead to correlated, higher-amplitude input signals. This is partly due to difficulties in formulating and solving computationally tractable problems for identification test design. In this work, related results are summarized and extended. Connections between different designs that target D-optimality or integral controllability are established. Related concepts are illustrated through simulation case studies.

Introducing uncertainty analysis of nucleation and crystal growth models in Process Analytical Technology (PAT) system design of crystallization processes

This paper presents the application of uncertainty and sensitivity analysis as part of a systematic model-based process monitoring and control (PAT) system design framework for crystallization processes. For the uncertainty analysis, the Monte Carlo procedure is used to propagate input uncertainty, while for sensitivity analysis, global methods including the standardized regression coefficients (SRC) and Morris screening are used to identify the most significant parameters. The potassium dihydrogen phosphate (KDP) crystallization process is used as a case study, both in open-loop and closed-loop operation. In the uncertainty analysis, the impact on the predicted output of uncertain parameters related to the nucleation and the crystal growth model has been investigated for both a one- and two-dimensional crystal size distribution (CSD). The open-loop results show that the input uncertainties lead to significant uncertainties on the CSD, with appearance of a secondary peak due to secondary nucleation for both cases. The sensitivity analysis indicated that the most important parameters affecting the CSDs are nucleation order and growth order constants. In the proposed PAT system design (closed-loop), the target CSD variability was successfully reduced compared to the open-loop case, also when considering uncertainty in nucleation and crystal growth model parameters. The latter forms a strong indication of the robustness of the proposed PAT system design in achieving the target CSD and encourages its transfer to full-scale implementation.

Comparison of model-based and conventional controllers on a pilot-scale heat exchanger

This pilot-scale heat exchanger demonstration compares two relatively simple nonlinear model-based control strategies to conventional proportional-integral (PI) control. The two nonlinear controllers, generic model control (GMC) and process-model based control (PMBC), use a first-principles model thereby providing characterization of the nonlinear process throughout the operating range. There are two approaches to GMC, one uses a dynamic model, the other a steady-state model. This work uses the steady-state model; accordingly, will use the term GMC-SS, which can be classified as output characterization for a PI controller, making it relatively simple to implement. PMBC uses a dynamic model and adapts to represent the process. These two nonlinear controllers were selected for this application evaluation because of their simplicity (they can be implemented in-house within many commercial control systems), diversity (steady-state and dynamic models), and demonstrated utility for control of nonlinear single-input–single-output processes. The application and results are presented and discussed.Summarizing the results: Within a small temperature operating range PI provides good control, but over the full operating range, the nonlinear and variable delay of the process lead to poor control with PI. GMC can handle the nonlinear issues, but using the convenient steady-state model; it also, provides poor control because of the variable delay associated with flow rate. PMBC was able to provide good control throughout the entire operating range. PMBC has a further advantage of only having one tuning coefficient, while PI and GMC-SS have two.

Model-Based Process Control for Optimised Banana Ripening

A dynamic model of the banana ripening process as performed in the food industry was developed. In the model, the degradation of starch to soluble sugars was described based on the carbon dioxide emission. The basis of the physico-chemical and the plant model parts were dynamic energy and mass balances. The model was evaluated by the means of experimental data produced in a pilot plant banana ripening box. The functionalities of the box follow industrial standards. The model describes the characteristic courses of the carbon dioxide concentration in the air of the process, the starch concentration in the banana pulp and the peel colour. Based on this information, the ripening state of the bananas could be estimated with the help of the model. Moreover, the model was used for an automated predictive and adaptive process control. For this purpose the Open-Loop-Feedback-Optimal (OLFO) controller was applied to the banana ripening process. Two characteristic model parameters, the starch-to-CO2 reaction rate factor and the peel colour change rate factor, were fitted to changing process courses. The control function of the banana ripening process, the set-point temperature profile, was optimised to automatically control the process to the required ripening state and the optimal storage temperature at the end of the procedure.

Realistic disturbance modeling using Hidden Markov Models: Applications in model-based process control

Understanding and modeling disturbances play a critical part in designing effective advanced model-based control solutions. Existing linear, stationary disturbance models are oftentimes limiting in the face of time-varying characteristics typically witnessed in process industries. These include intermittent drifts, abrupt changes, temporary oscillations, outliers and the likes. This work proposes a Hidden Markov Model-based framework to deal with such situations that exhibit discrete, modal behavior. The usefulness of the proposed disturbance framework – from modeling to ensuring the integral action under a wide variety of scenarios – is demonstrated through several examples.

One step forward from run-to-run critical dimension control: Across-wafer level critical dimension control through lithography and etch process

Process variation on lot-to-lot and wafer-to-wafer level has been well addressed using run-to-run control in advanced process control, however, to tackle the ever increasing die-to-die level process variation at the 65 nm technology and beyond, the process control must be extended into finer domain: die-to-die level. A novel model based process control approach is proposed to reduce the critical dimension(CD) variation on die-to-die level, i.e. across-wafer level. The central idea of the proposed approach is to compensate for upstream and downstream systematic CD variation by adjusting the across-wafer post-exposure bake (PEB) temperature profile of a multi-zone bake plate. A temperature-to-offset model relating the PEB temperature profile of multi-zone bake plate to its heater zone offsets was constructed experimentally using wireless temperature sensors from OnWafer Technologies. The baseline post-etch CD signature and plasma etch bias signature were extracted from the characterization experiment. A post-etch CD variation reduction of 40% was realized in the verification experiment, which validated the efficacy of the proposed approach. This CD variation control concept was further extended to simultaneous CD variation control for multiple CD features considering that multiple CD features coexist on a real chip, and CD variation control scheme based on multi-objective and minimax optimization is proposed to realize simultaneous CD variation control for multiple CD features. We also demonstrate through simulation using minimax optimization that across-wafer CD variation of multiple CD features can be significantly reduced for the 130 nm technology node and beyond.

EFFICIENT MODEL IDENTIFICATION THROUGH SENSOR LOCATION IN BATCH DISTILLATION

Prime areas in model based process control are process identification and optimization. This paper presents their linkage for a binary batch distillation pilot plant using minimum experimental data which is necessary in practice if the batch column is to be used as a multiproduct unit. Here, minimum experimental data from plant runs is used to determine the model parameters with minimum error. The main motivation behind this work is the efficient state and parameter estimation for which measurement locations play an important role. Hence optimum measurement locations are first determined and model identification is done based on these measurements.

RUN-TO-RUN CONTROL OF MEMBRANE FILTRATION PROCESSES

Membrane filtration processes are often operated cyclically, where one cycle comprises a filtration and a backwashing phase. Due to the complex mechanisms involved, these filtration processes are mostly operated with fixed values of the manipulated variables. In this paper, a model-based process control approach is introduced, which is based upon run-to-run control theory. To evaluate the controller, a suitable model of submerged membrane filtration in wastewater applications is developed, which describes the main process phenomena while being computationally inexpensive. The model-based controller is then tested in a simulation environment employing a validated reference model. Excellent results with respect to prediction quality and optimality are obtained.

Model-based advanced process control of coagulation

The drinking water treatment industry has seen a recent increase in the use of artificial neural networks (ANNs) for process modelling and offline process control tools and applications. While conceptual frameworks for integrating the ANN technology into the real-time control of complex treatment processes have been proposed, actual working systems have yet to be developed. This paper presents development and application of an ANN model-based advanced process control system for the coagulation process at a pilot-scale water treatment facility in Edmonton, Alberta, Canada. The system was successfully used to maintain a user-defined set point for effluent quality, by automatically varying operating conditions in response to changes in influent water quality. This new technology has the potential to realize significant operational cost saving for utilities when applied in full-scale applications.

Nonlinear model-based process control [Book Review

Nonlinear model-based process control [Book Review

Generic model control of induced protein expression in high cell density cultivation of Escherichia coli using on-line GFP-fusion monitoring

A model-based control algorithm (generic model control) is presented for fed-batch cultivation of recombinant Escherichia coli producing either transcriptional or translational fusion products. With the recent development of translational and operon fusions using green fluorescent protein (GFP) [Albano et al. (1998) Biotechnol Prog 14:351–354] along with an on-line GFP sensor [Randers-Eichhorn et al. (1997) Biotechnol Bioeng 55:921–926], real-time measurements of foreign protein level are now possible. A mathematical model is presented that is both accurate and simple so as to ensure that all state variables remain observable during cultivation. A balance between model accuracy and mathematical tractability was obtained to facilitate the formulation of the control algorithm. Generic model control (GMC) is a process model-based control algorithm incorporating a process model directly within the control structure. GMC was desirable since linearization of the process model was not necessary and robust performance could be obtained despite process disturbances or plant/model mismatch. Furthermore, a time-delay compensator was built into the control law to account for the observed 90-min lag associated with GFP fluorescence. The feasibility of the GMC algorithm was demonstrated by simulations.

Sensitivity studies for volume averaged models of plasma etch reactors

A well-stirred reactor model is employed to model the etching of silicon in low pressure chlorine/argon plasmas. The model gives the spatially averaged species composition and etch rate in a plasma etch reactor by solving conservation equations for species, mass and electron energy distribution function (EEDF). Systematic sensitivity analyses, made possible by a new iterative approach, allows the identification of key parameters for improved fault detection and model based process control of plasma reactors.

Sensitivity studies of silicon etching in chlorine/argon plasmas

In this article a well-stirred reactor model is utilized to model the etching of silicon in low-pressure chlorine/argon plasmas. Well-stirred reactor models are increasingly common in the literature due to their low requirements of computer resources for detailed chemical kinetics calculations. The model predicts the spatially averaged species composition and etch rate in a plasma etch reactor by solving conservation equations for species, mass, and the electron energy distribution function (EEDF). The reactor is characterized by a chamber volume, surface area, surface area fraction of the wafer, mass flow, pressure, power deposition, and composition of the feed gas. In such plasma etch models, assumptions on the EEDF which are needed to determine reaction rate coefficients for electron-impact reactions, are crucial for a prediction of steady state conditions. The model presented in a recent article [P. Ahlrichs, U. Riedel, and J. Warnatz, J. Vac. Sci. Technol. A 16, 1560 (1998)] is extended to describe the etching of the wafer with a special set of reactions occurring on a certain area fraction of the total reactor surface. A modified numerical procedure to solve the species conservation equations and the EEDF is presented, which needs considerably less computation time than the approach previously taken. Systematic sensitivity studies are presented to identify the connection between input parameters, outflow composition, and etch rate of the process. Such numerical studies are an important step towards fault detection and model based process control of plasma reactors.

Tai-ji ID Automatic Closed-Loop Identification Package for Model Based Process Control

Tai-Ji ID is an identification software that carries out identification automatically from test design to model validation. Data collected from both open-loop and closed-loop operations can be used. Tai-Ji ID is developed for model based process control such as model predictive control (MPC) and linear robust control. Special attention is paid to deal with industrial processes which have complex dynamics, high disturbance level and many process variables. Tai-Ji ID is based on the so called ASYM method of identification. Tai-Ji ID is designed for process control engineers and process engineers who are involved in model based process control and monitoring. No specific knowledge on process identification is required in order to use Tai-Ji ID. On the other hand, users' knowledge on process unit operation can be incorporated in data preparation step and in model validation step during a Tai-Ji ID session.

State Profile Estimation for a Nonisothermal Adsorption Process by a Nonlinear Distributed Parameter Observer

A major dynamical characteristic of pressure swing adsorption plants are nonlinear wave propagation phenomena travelling forwards and backwards within the fixed beds due to the change in flow direction. The control of product purity by an advanced process control strategy requires knowledge of the spatial concentration profiles. In this paper, the reconstruction of the spatial profiles of a single fixed bed of an oxygen production plant with a nonlinear distributed parameter observer is examined as a contribution to the design and the realization of a model based process control strategy. Assuming a binary mixture, the distributed parameter model describing the nonisothermal adsorption process comprises six space and time dependent states: temperature, pressure, oxygen mole fraction in the gaseous phase, the adsorbed amounts of oxygen and nitrogen, and the molar flux. The location of sensors is examined and discussed with respect to observer convergence and a nonlinear distributed parameter observer is designed for the adsorption process injecting correction terms in selected balance equations. The placement of the injected correction terms as well as their tuning parameters are of major importance concerning the convergence properties of the observer. The observer is implemented with the simulation environment DIVA and examined with respect to its convergence behavior.