Data processing: generic control systems or specific application – Specific application – apparatus or process – Chemical process control or monitoring system
Reexamination Certificate
2001-12-04
2004-04-06
Soderquist, Arlen (Department: 1743)
Data processing: generic control systems or specific application
Specific application, apparatus or process
Chemical process control or monitoring system
C700S030000, C700S031000, C700S032000, C700S033000, C700S034000
Reexamination Certificate
active
06718234
ABSTRACT:
FIELD OF THE INVENTION
The present invention relates to a system for on line inference and control, of the physical and chemical properties of polypropylene and its copolymers with other olefins produced in loop reactors, and, optionally, other gas-phase reactors, either isolated or combined in series, with the aid of mathematical models. More specifically, said system consists of models to infer physical and chemical properties that are not continuously measured and of relevant models to control these properties as well as other process variables of interest in the process being studied. The control system further enables the maximization of production rate and catalyst yield of producing process.
BACKGROUND INFORMATION
Usually, control of process variables in petrochemical plants is manually carried out by periodically taking product samples to be tested, while maintenance and correction actions concerning operating conditions are taken in order to obtain a product having desired characteristics. This procedure involves delays related to required corrections, since sampling and analysis are usually delayed as compared to the on line process, not to mention any human mistakes.
On the other hand, rigorous and empirical modeling techniques may be used to build process models. These mathematical models are able to infer the value of certain process variables which are measured periodically based on the values of other process variables which are measured continuously. Besides, the mathematical models can also be used to predict the future behavior of process variables resulting from modifications of the plant operating conditions.
Rigorous modeling techniques are based in the natural laws which dictate the fundamental relationships among process variables. Rigorous models are difficult to obtain and demand a high level of knowledge about the process. Besides, the complexity of the equations that constitute the rigorous model make it inadequate for on line implementation due to the long time necessary for its execution, even using computers.
In compensation, the techniques of empirical modeling do not require such a deep knowledge about the process and result in simpler mathematical models, that can be executed quickly being more adequate for real time execution. The disadvantage of empirical models is the fact they cannot be used in operating conditions different from those used in its identification. Linear and nonlinear regression models and neural nets are examples of empirical modeling techniques commonly cited in the open literature.
In the systems for inference and control described in present invention, the preferred modeling technique comprises the use of neural networks.
Neural networks are networks of elements interconnecting themselves in a unique way. Typically, networks consist of input neurons that receive signals or information from the outside of the network, output neurons that transmit signals or information to the outside of the network, and, at least, one intermediate layer of neurons that receive and pass information on to other neurons.
A system for controlling polymeric intrinsic viscosity during its production was described in the U.S. Pat. No. 3,878,379. However this technology was specific for the production of polyethylene terephtalate polyester and it comprised just a single output, excluding the control of others polymer properties.
The open literature contains a series of publications that deal with inference and control of processes in a general way using neural networks. U.S. Pat. No. 5,111,531 teaches how to implement an on line predictive control system applied to a continuous process where a neural network predicts the value of a controlled variable, the values of manipulated variables being adjusted so that the value predicted by the network closely approaches the desired value.
U.S. Pat. No. 5,175,678 teaches how to implement an on line feedforward control system using neural networks.
U.S. Pat. No. 5,282,261 teaches how to implement an inference and control system in a continuous process using neural networks.
U.S. Pat. No. 5,479,573 teaches how to develop a predictive network to operate in real time, or in training mode. Preprocessed data is used as inputs of the model for the system, which provides a predicted output that is used in the distributed control system.
U.S. Pat. No. 5,675,253 teaches how to use a neural network to develop a model, which correlates process variables together with the results of an on line nuclear magnetic resonance system to predict a desired property of the polymer.
U.S. Pat. No. 5,729,661 teaches a method and a device to preprocess input data for a neural network. Data is preprocessed to replace wrong data, or to fill in missing data and then used to train a model.
Examples of inference systems that use neural networks applied to specific processes are also found. U.S. Pat. No. 5,548,528 teaches how to build a system to continuously monitor pollutant emissions. Such system uses the neural networks to replace instruments that perform pollutant analysis in a plant. Then, the value predicted by the neural network is used as a controlled variable in a control system.
WO Patent 98/26336 teaches a method to infer the blade temperature of a vapor turbine. This invention uses temperature and pressure from other places of the turbine as inputs of a neural network that predicts blade temperature.
However, no published Patents, neither isolated nor in combination with other published literature, explains the on line inference of physical and chemical properties of polypropylene and its copolymers with other olefins produced in loop reactors and, optionally, in other gas-phase reactors, either isolated, or combined in series from measured process variables using mathematical models. Further, no Patent teaches or suggests the on line control of process variables such as production rate, density of reaction medium, and physical and chemical properties of polypropylene and its copolymers with other olefins produced in loop reactors, and, optionally, in other gas-phase reactors, either isolated or combined in series using mathematical models, such on line control and inference being enabled by the system described and claimed in present invention.
SUMMARY OF THE INVENTION
The present invention, as disclosed and claimed herein, is directed to a system for on line inference of physical and chemical properties of polypropylene and its copolymers produced in a plant using loop reactors, and, optionally, gas-phase reactor(s) with the aid of mathematical models, such system comprising the steps of:
collecting data of all variables measured in the plant, so as to build a large historical database;
out of said large database, selecting a subset including historical data on the physical and chemical properties to be inferred, as well as potential input variables for mathematical models so as to build a small database;
treating data of said small database, removing non-representative data of usual operating conditions, applying noise removal filters on signals and complementing said database with new calculated variables based on theoretical concepts such as energy and mass balances;
from the data of said small database, selecting those variables that will be the input variables of the mathematical model;
using said input variables to identify these models off line;
applying the so identified mathematical models to the real-time measurements of the input variables to the model used to calculate the on line inference of desired physical and chemical properties.
The present invention also provides an on line control system for a plant of polypropylene, and its copolymers produced in loop reactors, and, optionally, gas-phase reactor(s) using mathematical models, the system being based on a control matrix which comprises controlled variables, constrained controlled variables, manipulated variables, and disturbance variables.
Thus, the present invention provides a system for the on line inference and control of physical and chemical proper
Axelrud Celso
De Oliveira Autur Toledo Martins
Demoro Esdras Piraguacy
Birch & Stewart Kolasch & Birch, LLP
Braskem S.A.
Gakh Yelena
Soderquist Arlen
LandOfFree
System for on line inference of physical and chemical... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with System for on line inference of physical and chemical..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and System for on line inference of physical and chemical... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3217551