Search

Commercially, the Czochralski process plays a key role in production of monocrystalline silicon for semiconductor and solar cell applications. However, it is a highly complex batch process which requires careful control throughout the whole crystal production. In the present paper, an iterative method based on model predictive control (MPC) for calculating a new and improved trajectory from a growth run to the next is explored. The method uses the results of the previous growth run in combination with an underlying model which incorporates the complex dynamic effect of the heater temperature on the pulling rate. The motivation behind this choice of strategy is to enhance the quality of the fully grown ingot from one run to the next by applying the most recent estimates of the unknown parameters. The results show that combining MPC, estimation and Run-To-Run control has enabled simulation of effective control of the Czochralski crystallization process. © 2017. This is the authors’ accepted and refereed manuscript to the article. Locked until 4.5.2019 due to copyright restrictions. This manuscript version is made available under the CC-BY-NC-ND 4.0 license http://creativecommons.org/licenses/by-nc-nd/4.0/

Glucose-insulin metabolism models are useful tools for research on diabetes, in development of diabetes-related medical devices like artificial pancreas systems, insulin pumps and continuous glucose monitors, and may also play a role in personalized decision support tools for people with diabetes. Such models are often highly nonlinear with many parameters that are person dependent. An example is the model used in the UVa/Padova T1DM simulator, which has a large number of states and parameters. It is desirable to be able to personalize such models through parameter identification based on limited glucose, meal and insulin data obtainable from free-living settings, as opposed to clinical research settings that have traditionally been required. In this paper we use the UVa-Padova T1DM simulator model in a case study to investigate observability of the model under different measurements, and the identifiability of its parameters as a function of the model's inputs and outputs. Structural identifiability is discussed and briefly investigated using the nonlinear Observability Rank Condition. Practical identifiability is discussed and investigated using sensitivity and Fisher information matrix analysis. We show how such analyses can be used to guide model reduction for improved identifiability, or to select the most proper subset of parameters to estimate. © 2019 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works.

Commercially, the Czochralski process is the most commonly applied method for production of monocrystalline silicon for semiconductor and solar cell applications. The process has traditionally been controlled using nested single-loop controllers. Due to the complex and nonlinear nature of the Cz process, this paper considers model-based control of the process with application of nonlinear model predictive controllers.

Measurement of electrodermal activity (EDA) has recently made a transition from the laboratory into daily life with the emergence of wearable devices. Movement and nongelled electrodes make these devices more susceptible to noise and artifacts. In addition, real-time interpretation of the measurement is needed for user feedback. The Kalman filter approach may conveniently deal with both these issues. This paper presents a biophysical model for EDA implemented in an extended Kalman filter. Employing the filter on data from Physionet along with simulated noise and artifacts demonstrates noise and artifact suppression while implicitly providing estimates of model states and parameters such as the sudomotor nerve activation.

An integrated system for pollution control has been developed and is installed in a fertiliser plant. The aim of the system is to improve the quality control at all levels by streamlining the information flow from the process to operators and management. The system is composed of an operator support system that supplements the distributed control system, and a reporting and documentation system that generates reports at several levels of detail, tailored to the needs of individual users. At several levels of detail, discharge data is compared to limits imposed by the authorities and by company policy. This enables rapid corrective action by operators and management. The system may form the basis of a more general tool for production quality management.

An operator support system for detection of process upsets and determining their cause is presented. An important part of the system is a total plant model where each process unit is represented as an attributed process object and where the process flows are modelled as relations between process objects in the form of process connections. For each process connection, a number of measured or calculated process variables are monitored for deviations. Once a deviation has been detected, the process object connected at the input of the process connection checks all incoming streams to see if the upset may have been caused by an upstream process object. If so, control is passed to the diagnoser associated with the upstream process unit, which in tum may pass control to a unit further upstream. This goes on until a diagnoser finds no relevant deviations in any of its upstream units. The diagnoser then decides that the cause of the upset is internal and an internal diagnosis of the process object is initiated. The diagnostic process can be controlled or overridden by the operator. The operator interaction is based on animated process flowsheet pictures where the diagnostic search is indicated by colour coded process object and process connections. The internal diagnostics of the system is based on quantitative techniques such as process models, fault tree models and Kaiman filtering methods. The topological search is rule based.

Summary In oil-rim fields, a thin oil layer lies between an aquifer and a gas cap. Oil may be produced from such fields by use of horizontal wells. Production will lower the local gas/oil contact (GOC) near the well in a process called gas coning. After gas breakthrough, the gas/oil ratio (GOR) from the well may vary strongly with the production rate. The ability to predict this dependency is essential for production optimization for such fields. We have developed a mathematical model that can predict gas-coning behavior and the resulting rate-dependent GOR for periods of several months or more. Despite a simple model structure and short computation times, the accuracy of the predictions is good. We combine a dynamic model that describes the essential reservoir behavior with a highly simplified description of the interaction between the well and the surrounding reservoir. The full model has three adjustable parameters that allow us to fit the behavior to individual wells, using historical oil-and gas-production rates. The model forms the basis of the GORM (gas/oil ratio model) computer program that, since early 2003, is in regular use for production planning and optimization at the Troll field. We also have tested the model on wells in other fields, with encouraging results. Note that the model describes the behavior of individual wells only. The focus is short-and medium-time production optimization, not reservoir management.

This paper gives an overview of identification techniques for use in model based predictive control (MPC) applications. Use of parametric and nonparametric models are discussed such as impulse- and step reponse models, statespace models, innovation form models and ARMAX/ARIMAX models. Prediction error methods for on-line and offline identification are presented, control relevant identification techniques based on multistep prediction error identification are presented and identifiability measures and test signal selection are discussed.

An interactive program package for synthesising test signals for system identification has been developed. The signals are generated by repeated maximization of the information return over a time horizon which is shorter than the total length of the experiment. This simplification reduces the computational burden vastly, at the expense of the obtained test signals not being optimal in the global sense. However, the method seems to be useful in practical identification problems, and some examples of its use are given.

The paper describes an adaptive autopilot. The autopilot is based on a mathematical model which uses a priori information to find initial parameter estimates. The autopilot also includes a model of wave induced law-motion. The estimates of the oscillating yaw components are used to minimize their influence on the rudder motions. It is shown in the paper that a Kalman filter gain for a vessel of unit length can be computed once and for all. Based upon this unit vessel Kalman filter gain, formulas for computing the gain for a general vessel are given. A vector of variable forgetting factors are introduced. This completely eliminates the blow up problem. The autopilot is tested on board the coastal steamer M/s MIDNATSOL and results from these tests are given.

This paper is concerned with analysis and design of an adaptive autopilot for ships. The design is based on a low and high frequency model of the vessel motion adequate to ship steering. The low frequency model describes the vessel response to rudder control and slowly varying environmental forces. The high frequency model represents the wave induced oscillatory part of the yaw motion. The models are used in a Kalmanfilter and the rudder control is computed from linear quadratic theory based on the low frequency part of the state vector. This yields a very effective filtering of the wave component of the yaw motion. Proper operation of this filter/controller structure requires knowledge of the vessel model parameters and the dominating wave frequency. The vessel parameters are estimated on line by a recursive prediction error method. In order to reduce the computing requirements, the state estimator is operated using scheduled gains. This results in an easy and robust design. The Convergence properties are investigated by using the method of Ljung. The performance is exemplified by simulation experiments.

A mathematical model of a fluidized bed reactor for roasting of pyrite is formulated, based on physical and chemical relationships. Several unknown or uncertain model parameters are estimated from experiments on a pilot plant reactor. The estimation algorithm is based on the maximum likelihood method. A multivariable control system is constructed using Kalman filtering and optimal control theory. The control system is implemented on the actual pilot plant reactor. Operating records indicate an excellent control system performance.

This paper describes a computer-based, dynamic positioning system for floating vessels. The system is based on a detailed mathematical model of vessel motion in response to forces from thrusters, wind, waves and water current. The system uses a Kalman filter for optimal estimation of vessel motions and environmental forces from wind, waves and current. The control system is based on feedback from the motion variables where the oscillatory, wave-induced component is removed by the estimator. Feedback from the water current estimate provides the integral action of the system and feed forward from the wind force estimates are implemented. Simulation results and recordings from actual operation of the system indicate an excellent system performance.

A splitting algorithm for fast and slow modes of ocean hydrodynamics is presented. The purpose of the splitting is to reduce the large amount of computational work needed for simulating long real-time periods. The essential point of the splitting is that the external gravity wave terms are extracted from the fully three-dimensional equations of horizontal motion, allowing the reduced equations to be integrated with a larger time step than the original model. The fast external gravity waves are traced by a depth integrated system which is weakly coupled to the reduced three-dimensional momentum equations. The split model shows a radical decrease in computational time and the accuracy is of the same order as in the non-split case.

This paper addresses the blow-up problem associated with the parameter estimation part of an adaptive controller. A partial solution to the problem has been devised by the introduction of a variable forgetting factor. However, this does not eliminate the blow-up possibility. This is shown by simulation experiments on two different models.

This paper describes computer-based, dynamic positioning system for floating vessels. The system is based on a detailed mathematical model of vessel motion in response to forces from thrusters, wind, waves and water current. The system uses a Kalman filter for optimal estimation of vessel motions and environmental forces from wind, waves and current. The control system is based on feedback from the motion variables where the oscillatory, wave-induced component is removed by the estimator. Feedback from the water current estimate provides the integral action of the system and feed forward from the wind force estimates are implemented. Simulation results and recordings from actual operation of the system indicate an excellent system performance. Reference is given to installations made on actual vessels.

This paper addresses the blow-up problem associated with the estimation part of an adaptive controller. A partial solution of the problem has been deviced by introduction of a variable forgetting factor. It does, however, not eliminate the blow-up possibility. This is shown by simultation experiments on two different models. Two new methods based on a vector variable forgetting factor are presented. One of these methods completely solves the blow-up problem, whereas the other reduces the possibility of blow-up tendencies. Simulation experiments using both methods are compared with the single variable forgetting factor case.