Classifications

• • G—PHYSICS
  • G09—EDUCATION; CRYPTOGRAPHY; DISPLAY; ADVERTISING; SEALS
  • G09B—EDUCATIONAL OR DEMONSTRATION APPLIANCES; APPLIANCES FOR TEACHING, OR COMMUNICATING WITH, THE BLIND, DEAF OR MUTE; MODELS; PLANETARIA; GLOBES; MAPS; DIAGRAMS
  • G09B25/00—Models for purposes not provided for in G09B23/00, e.g. full-sized devices for demonstration purposes
  • G09B25/02—Models for purposes not provided for in G09B23/00, e.g. full-sized devices for demonstration purposes of industrial processes; of machinery

Definitions

• This invention relates to a method and an apparatus for training the operating personnel of a process-technical plant which includes a control station for monitoring and controlling the processes running on the plant.
• Preferred applications of the invention are offered in particular in plants for performing chemical or physical processes or industrial manufacturing methods, for instance a sintering plant, a pelletizing plant, a sulfuric-acid plant, an anode production plant, a reduction plant or some other large-scale chemical plant, which are controlled, regulated and monitored from a control station, in particular by using a control system, a distributed (process) control system (DCS) or a programmable logic controller (PLC).
• Chemical or physical processes in particular are understood to be processes in which a physical and/or chemical conversion of substances is effected.
• the chemical processes in particular also include metallurgical processes.
• control station is understood to be both an operator console with control and monitoring functions and a control system, e.g. a distributed (process) control system, or a programmable logic controller, in which regulation and/or control operations are configured or defined for the plant.
• a control station frequently includes combinations of the aforementioned elements beside possibly existing further components.
• commands issued in the control station which can be manual commands or signals from the process control system, e.g. the DCS, or the programmable logic controller, e.g. controller output signals or actuating signals, are tapped and supplied to a process simulator.
• the outputs of the process simulator are supplied to the control station, i.e. to the operator console or operating personnel, to the process control system (DCS) or to the programmable logic controller (PCL).
• the process running on the plant is simulated in the process simulator by considering in particular all commands or signals issued from the control station to the plant, which also comprise the reactions of the control system (e.g. the DCS) or the programmable logic controller.
• JP 2004021180 A Patent Abstracts of Japan
• JP 2004021180 A proposes an on-site operation training simulator, which provides for operating the training simulator with the operation monitoring means also provided for controlling the plant.
• the operation monitoring means can selectively be connected to the plant to be controlled or to the training simulator.
• the process running on the plant then is simulated.
• first principal based equations are defined as a formula-based description of direct causal and back-coupled connections or relations in a process on the basis of physical, chemical and/or process-technical equations based on laws of nature and can be parameterized and expressed by one or more equations.
• the relations generally are based on scientific conceptions or models based on expressions of laws of nature. In large-scale chemical plants, these are in practice stationary (mass, energy, enthalpy or other) balance equations with combined time behaviour terms, dynamic systems of differential equations, or mixed statements or terms including stationary and dynamic equation components.
• neural networks offer the possibility of generating a model of the underlying process on the basis of measurement data from the plant history, without a priori making assumptions on the structure of the functional relations between the different variables. Such assumptions are for instance partly necessary in linear regression statements with non-linear statement functions.
• the main applications of neural networks in the process-technical industry consist for instance in modelling processes and plants with the aim to optimize operating points, monitor measurement data, provide for online error diagnoses or expediently intervene in plant controls.
• EP 0 756 219 B1 which relates to a method for monitoring the properties of products fabricated in a manufacturing process.
• the product properties are predicted on the basis of measured process data by means of a prediction model, which at least outputs a prediction interval.
• the proposed method generates a neural network and continuously collects available process data, in order to constantly expand the limits of the prediction model on the basis of measurement data and constantly change the parameters of this model and hence improve the model quality based on the process.
• Pure data-driven models can only map those operating variants of a process which actually have occurred already in data. Because of the great number of theoretical possibilities of disturbing events, however, not all possible combinations are available at any time, so as to be able to adapt data-driven models to these possibilities.
• the process simulator uses both rigorous models and data-driven models for simulating and modelling the process in the plant and connects the same with each other via internal interfaces.
• This approach provides for using rigorous partial models in the process simulator for many properly elucidated scientific- technical relations and data-driven partial models for the relations not known accurately enough or hard to describe, which then can be assembled to a total model via the interfaces.
• the rigorous models can be based e.g. on physical equations, balance equations, reaction-kinetic differential equations and other scientifically properly elucidated formulae or models for known or examined relations.
• the individual rigorous or data-driven partial models are connected with each other by the internal interfaces provided, wherein one interface implements both transitions between rigorous and data-driven partial models and transitions each between rigorous models or data-driven models.
• a characteristic of these inter- faces in particular can be the fact that the result of the preceding partial model is used as a starting point for the following partial model.
• a (total) model of the process simulator consisting of both rigorous and data- driven models (partial models), which is also referred to as hybrid model, only requires a comparatively short calculation time, as known relations are mapped by few calculation-intensive rigorous models and only those parts of the process which cannot be mapped or can only be mapped with great effort by closed equations by means of parameters are covered by data-driven models connected with the rigorous models.
• causal connections inside the process which can be described by equations on the basis of known physical, chemical and/or process-technical relations without much effort and with the required accuracy, in a first step by one or more rigorous models, and to implement the remaining causal connections for mapping the process by one or more data-driven models in a second step.
• It can also be advantageous to convert empirical substance databases to model equations by means of data-driven statements or to take over the same (e.g. adapted equations for thermal capaci- ties) and to use the same, in order to specify parameters for the rigorous models.
• apparatus characteristics can also be used or converted, in order to obtain parameters for these models.
• physicochemically determinable or characterizable intermediate conditions as internal interfaces between the rigorous models and the data-driven models. Finding or determining such inter- mediate conditions can be effected by evaluating real process data, since the intermediate conditions frequently are measurable conditions existing in the plant at certain times of the process.
• the internal interfaces in the plant define determinable intermediate conditions, which must or can be described by the partial models and also provide for checking portions of the individual partial models by measuring process quantities in the plant.
• These intermediate conditions not necessarily are uniform chemical or physical conditions, but can unambiguously be determined or characterized by various parameters, e.g. by temperature, chemical composition, retention time, distribution of particle sizes, changes in density.
• Such intermediate conditions provide an ideal starting point for the further description of the process by means of rigorous models. Therefore, it is proposed in accordance with the invention to describe all or some, but preferably many transitions between intermediate conditions, which are based on physical- chemical reactions, with rigorous models. Checking the rigorous models or the respective model parameters then can be effected by means of real, measured operating data or by laboratory experiments or pilot plant experiments. Various characterizable intermediate conditions can also be present at one place at the same time.
• the actuating variables output by the control station to the plant in particular due to control commands or automatic regulation or control interventions can be used as first external interfaces between the control station and the process simulator.
• the effects of these actuating variables frequently cannot be described in a closed, i.e. rigorous form.
• status information of the plant can be defined, which is supplied to the control station as return messages (measurements) of the plant.
• the training simulator is able to indicate process data simulated by the same in the form of sensor information in the control station and thereby achieve a realistic simulation of the plant also in the control station.
• status informa- tion frequently in particular is values of measurement sensors located in the plant.
• control station selectively controls the plant or the process simulator, wherein this is preferably not indi- cated in the control station.
• the control of the process simulator by the control station comes out as control of the plant and provides for using the training simulator under real conditions.
• the operating operators to be trained need not know that they do not control the plant, but a simulator.
• a second control station can then be integrated in the training simulator, which controls the plant or allows for instance active interventions of the trainer in the simulator when the plant is shut off, in order to possibly simulate special events or accidents and train the behaviour of the operators.
• the process running on the plant can be simulated particularly realistically by the method of the invention, because the control commands in the control station are directly and preferably completely considered in the simulation, it is particularly advantageous to also use the process simulator or parts of the process simulator for plant optimization.
• This can for instance be effected by predi- cative online optimization, by guiding the variable quantities of the process (manipulable variables) such that given technical constraints are observed and a (mostly economical) target function is maximized or minimized by predicting (simulating) effects of interventions.
• this predicative optimization can be implemented so accurately, because the model has a particularly high correspondence of the total model with the reality due to the combination of the rigorous and data-driven models.
• process simulator or parts of the process simulator for improving control strategies, in particular for optimizing and/or testing the same, by testing for instance changes of the configuration or programming of the process control system, e.g. the distributed (process) control system or of the PLC at the process simulator, before the same are employed in the real plant. Furthermore, it can also be advantageous when changes in parts of the plant are first checked by the process simulator.
• This invention also relates to an apparatus for training the operating personnel of a process-technical plant, in particular for performing the method described above.
• the plant is connected with a control station for monitoring and controlling the process running on the plant.
• the apparatus of the invention includes external interfaces for tapping the com- mands or signals issued in the control station and/or for supplying status information to the control station, and a process simulator with a calculation unit.
• the external interfaces can be configured as first interfaces for tapping the commands issued in the control station and second interfaces for supplying status information to the control station, wherein these interfaces can also be com- bined to a common interface or interface unit.
• the calculation unit of the process simulator is equipped to simulate the process running on the plant by means of rigorous models and data-driven models, wherein at least one, preferably a plurality of internal interfaces are formed between the rigorous and data-driven models.
• These internal interfaces serve to connect rigorous and data-driven models which each other, which serve as partial models, wherein in accordance with the invention two or more data-driven models and/or two or more rigorous models can also be connected with each other via an internal interface.
• the calculation unit represents an implementation of the method described above by means of data processing programs, to which this invention relates as well.
• the external interfaces which can comprise the first and/or the second external interfaces, are interposable in an interface or interface unit provided between the control station and the plant.
• the same can then be configur- able such that the data exchanged between the control system and the plant via the interface selectively can be tapped in parallel in the external interface of the apparatus of the invention or can be changed and be output again by the process simulator.
• the process simulator also receives all control commands, for instance in the form of actuating variables, and status information of the plant, for instance in the form of sensor data.
• a simulation operation and a regulation or control operation can also be performed in parallel.
• the process simulator can include a separate, second control station, which provides for a plant control in the case of training.
• This control station can for instance be implemented in the apparatus of the invention as a pure software implementation.
• a training case which simulates a critical plant situation, during the ongoing operation of the plant, and to further control the plant by the training simulator of the invention itself during this training case.
• a particularly authentic training situation is achieved for the operating personnel or the operators of a process- technical plant.
• FIG. 1 schematically shows the execution of the process of the invention and the structure of the apparatus of the invention for training the operating personnel of a plant.
• a process-technical plant 1 in which a chemical process 2, for instance a sintering process or a crystallization takes place.
• the chemical process 2 is schematically symbolized by a sequence of arrows and rectangles, wherein the rectangles represent physicochemically characterizable intermediate conditions 3 and the arrows represent causal connections 4 of the process during a transition between the intermediate conditions 3.
• the causal connections can for instance be transitions between two intermediate conditions, which are caused by chemical reactions or physical regularities, i.e. transi- tions each between a starting condition and a final condition.
• the kind of causal connections is arbitrary and not restricted to this typical case.
• the plant 1 is connected with a control station or control system 7, via which operating per- sonnel (operators) can control and monitor the process 2 running on the plant 1.
• controllers 8 are provided, which supply actuating variables to the plant via the first external interfaces 5.
• These actuating variables influence plant parameters, such as pressure, temperature or the like, in certain parts of the plant and thereby the causal connections 4 in the plant 1.
• These intermediate condi- tions 3 preferably are defined such that they are correlated with a defined physical-chemical condition of the plant, which can be detected for instance by means of one or more sensors or auxiliary variables in the intermediate conditions 3.
• These sensors transmit their sensor values via the second external interfaces 6 to the control station 7, in which the sensor values are represented on displays 9 for characterizing the plant condition.
• a tool should be provided for the operating personnel or the operators, in order to facilitate learning how to operate the process 2 running on the plant 1 and to continuously train the operating personnel also during operation of the plant 1.
• the process simulator 10 is integrated in an apparatus 11 for training the operating personnel of a process-technical plant 1 with first interfaces 5 for tapping the commands issued in the control station 7 and second interfaces 6 for supplying status information to the control station 7.
• the process simulator 10 includes a calculation unit, which is equipped to simulate the process 2 running on the plant 1 by means of rigorous models 12 and data-driven models 13, wherein internal interfaces 14 are formed between the rigorous models 12 and the data-driven models 13.
• These interfaces 14 can correspond in particular to the defined or characterizable interme- diate conditions 3 of the process 2 and thus indicate defined physical-chemical conditions.
• a command issued in the control station 7 by a controller 8 is tapped by the process simulator 10 as actuating variable in one of the first external interfaces 5 and is available as input variable for the simulator.
• distributed (process) control systems and programmable logic controllers can be integrated in the control station 7, which in turn generate commands and signals which can be tapped as actuating variables in the first external interfaces 5.
• the actuating variables tapped in the external interfaces 5 are connected with the internal interfaces 14 by means of data-driven models, wherein in the illustrated example the actuating variable a of the controller 8 has an influence on the conditions A, B and C of the internal interfaces 14.
• the actuating variable b influences the conditions A and C of the interfaces 14.
• the actuating variable c finally influences the conditions B and D of the internal interfaces 14.
• the direct influence of the controllers 8 on the conditions of the plant mostly is only difficult to describe by means of rigorous models 12, which typically rather are applied for the physical-chemical procedure.
• the method proposed in accordance with the invention thus forms a network between internal interfaces 14, which are each connected with each other by rig- orous models 12 and/or data-driven models 13.
• These networks can of course have a much more complex structure than is illustrated in the example in a simplified way.
• the hybrid models proposed in accordance with the invention now implement the simply rigorously describable relations or models with known equations, mostly in the form of stationary balance equations, differential equations and combined (time) terms.
• the complex model parts which are based on a model which is only difficult to represent mathematically in a closed form, are implemented with data-driven partial models, wherein the data-driven partial models in particular can also comprise self-configuring artificial neural networks.
• the same can also have a complicated and complex structure, but generate themselves from data obtained in the plant and supplied for instance via the external interfaces 6, and can be stored and evaluated in the process simulator 10 for configuration of the data-driven models 13. These methods are known to one of skill in the art and need therefore not be explained in detail (see e.g. Chemical Engineering and Processing 44 (2005), pp. 581-592 or pp. 855-868).
• What is essential for the proposed process simulator 10 is the fact that the rigorous models 12 and the data-driven models 13 generate a total hybrid model of the plant 1 via the internal interfaces 14, via which they communicate with each other.
• the communication between the process simulator 10 or the apparatus 11 and the control station 7 is effected via an interface containing the external interfaces 5, 6.
• This interface can for instance constitute an OPC interface, which represents a standardized interface for communication between the hardware of the control station 7 and the calculation unit of the process simulator 10.
• OPC interface represents a standardized interface for communication between the hardware of the control station 7 and the calculation unit of the process simulator 10.
• the operating and observing means 8, 9 of the control station 7 thus can selectively be connected by the trainer to the real plant control or to the training simulator itself. Ideally, a train- ing situation is possible, which is not recognized as such by the operating personnel or the operators and therefore provides a particularly realistic possibility for training and practicing dangerous situations.
• a separate, second control station 16 can be connected to the process simulator 10, with which the trainer or other operating personnel controls the plant 1 when the control station 7 is used for training purposes.
• the second control station 16 and the input and output unit 15 can constitute a common means, as in particular a second control station 16 also enables the trainer to specify an arbitrary training situation in the plant, to which the trained operating personnel at the control station 7 then must react.
• the process simulator can also be used for predicting future operating conditions. This prediction can be used for optimizing the existing plant, for testing e.g. alternative feed materials, changed parts of the plant or new operating conditions. This is possible, because the process simulator 10 gets access to the current process data of the process 2 running in the plant 1 via the external interfaces 5, 6. Since these interfaces 5, 6 allow parallel tapping of the signals, the process simulator 10 serving as training simulator therefore can also be used as optimizer for the plant, which makes accurate suggestions for settings of certain control elements of the control station 7, parallel to the ongoing operation. Concrete examples of such systems include training simulators for sintering or pelletizing plants.
• the time behaviour initially is considered or removed from among the operating data of the plant 1 , so that the data records describe stationary conditions, which in particular form the characterizable intermediate conditions 3 and the internal interfaces 14.
• known stationary equations then are inserted between the data records.
• a sintering plant for instance, such rigorously describable relation exists between the bed height of a sintering belt, the thickness of the ore mixture, the belt speed and the quantity of the ore mixture charged.
• Unknown relations such as the temperature distribution of the sintering process along the entire length of the sintering machine in dependence on the formulation of the ore mixture and the belt speed, are mapped with data- driven models 13, for instance artificial neural networks.
• the results of the process for instance the amount of returns
• the partial model describing the amount of returns can for instance be used to support the plant operator in the conduct of the process and ensure an optimum execution of the process.

Landscapes

• Engineering & Computer Science (AREA)
• Business, Economics & Management (AREA)
• Physics & Mathematics (AREA)
• Educational Administration (AREA)
• Educational Technology (AREA)
• General Physics & Mathematics (AREA)
• Theoretical Computer Science (AREA)
• Testing And Monitoring For Control Systems (AREA)
• Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Applications Claiming Priority (4)

Publications (1)

Family

ID=40560855

Family Applications (1)

Country Status (2)

Cited By (1)

Families Citing this family (3)

Citations (1)

Family Cites Families (9)

• 2007
  • 2007-12-11 DE DE102007059582A patent/DE102007059582B4/de not_active Expired - Fee Related
• 2008
  • 2008-11-12 WO PCT/EP2008/009538 patent/WO2009062680A1/en active Application Filing

Patent Citations (1)

Non-Patent Citations (1)

Also Published As

Similar Documents

Legal Events