DE3933958A1 - Process modelling system - has system represented by lattice network of cells representing system characteristics - Google Patents

Abstract

Alattice network is used to establish a model that can be used to simulate the behaviour of fluid and thermodynamic processes in order to examine the effects of system disturbances. At the microscopic level a two-dimensional grid-gas model is established in terms of fluid particles, e.g steam, pressure, energy. The fundamental cells are combined by interconnections to crete supercells that allow phase changes to be considered. Various parameters, such as temp. enthalpy, speed can be processed. Hardware effects may be considered to identify changes within the process. The system may be controlled by knowledge base rules. ADVANTAGE - Offers improved dynamic modelling of systems.

Description

1. Introduction

In recent years, a lot of effort has been put into arranging artificial intelligence driven to quick plant analysis sensors or facilities for the visualization of plants conditions or special systems for early error detection to get (1, 2). Recently an expert system was based end to simulation, state estimation and fuzzy techniques, published (3).

In this post, a concept is discussed that from the combination of phenomenological, dynamic and numerical capabilities of a cellular automaton as well as the macroscopic aspects and numerics of the hydrodynamic or thermohydraulic processes, such as those in systems or compo expire.

2. Levels of knowledge generation

Based on the formalization of the system hardware and the media (water, steam, coolant) with the help of fully computer-calculable matrices, called topographies (4), a total of four sets of axioms are required for the first model level: grid gas and phase change model; Object transformations; elementary process functions; Macroscopic targets together with restrictions, see picture 1.

2.1 Microscopic lattice gas and model of phase change

In recent years, cellular automata have been used successfully Simulation of hydrodynamic phenomena used (5, 6). The hexagonal flat grid and the rules of collision of Liquid particles were adopted in this post, however some new particle types were introduced.

In addition to liquid particles (water particles) also act virtual particles (pressure particles, heat particles) as Kolisi partners, as well as steam particles (originating from phase change changes). This model must be checked and adapted with the help of numerical results from partial differ ential equations or from system measurement data stand, so not only phenomenological and qualitative ten, but ultimately also semi-quantitative and quantitative results.

Figure 2 provides a simple overview, in which a liquid-filled (water-filled) vessel is modeled using hexagonal grid axes and grid points that represent the vessel hardware or, in combination with the existing and moving liquid particles (arrows), the liquid and the liquid level .

All liquid particles only move along the lattice axes, each determined by a discrete time step. In Figure 2, all liquid particles have the same absolute speed, but larger discrete speeds must also be permitted for cases with forced pressure and forced liquid flow. The collision rules (scattering or pure transport) are shown in Figure 3.

In order to be able to treat thermohydraulic processes, vir current pressure particles and virtual heat particles, neither of which are Possess mass, introduced. That way move in one Supercell made up of a few hexagonal grid points consists of a number of all three particle types, which thus the condition of the hot and high pressure water present at that particular time. Special rules apply to the interaction between water and steam and heat particles in such a way that the physical parameters (Enthalpy, density, temperature, pressure, viscosity, surfaces energy) and the volume movement of the water are cash.

Under certain conditions, e.g. if the number of heat particles in the super cell compared to the number of pressure particles in the same super cell exceeds a certain value, the content of the super cell is subject to a phase transition, see Figure 4b. Most of the water particles escape from the super cell and move into the neighboring super cells; The same applies to some of the pressure and heat particles, as a result of which some heat particles are also consumed as a result of this phase change and some of the pressure particles are also destroyed. In the new super cell that resulted from this phase change, some vapor particles now maintain the temperature and pressure with the help of their own high speeds and impulses.

It is helpful to approximate the water and steam properties to model the steam tables (7). New results from Research on saturated liquids, according to which only measurable physical parameters are used here to use if appropriate.

2.2 Semi-microscopic knowledge of cellular processes

By averaging or summing over a large number of grid cells of the microscopic model, one obtains a cellular insight into the hydrodynamic or thermohydrau lical processes taking place. In this way, for example, the pressure increase in a closed and electrically heated and water-filled vessel as well as the chaotic eruption of water drops and steam clouds from the finally burst vessel can be clearly shown, see Figure 5.

With the help of such an image, which is created with the help of a compu The model computer can implement the readable pattern locate the energy source (incoming from the heater Heat elements), also determine the heat transport phenomena, furthermore the traversed paths of heat, pressure, water and Vapor particles that flow out of the vessel via the leak, he know.

The hereby from the microscopic model on semi-microscopic The level of numerical information obtained can be this semi-microscopic model can be used.

2.3 Macroscopic process knowledge

On a macroscopic level, it is useful on the one hand, individuali based media or physical parameter particles (e.g. En thalpie), which one from semi-microscopic level or macrosko received technical invoices or data. Therefore the Wandering Process Elements (WPE's) were introduced, see Section 3.5.

On the other hand there are macroscopic process abstractions created with the help of a process network, further process evolu tion with the help of object transformations and application ele mental process functions as well as goals and restrictions attention and all machine-cognitive steps are managed.

3. Learning and abstracting processes

The learning of processes is not only necessary with the help simple object transformations, but also through Application of elementary process functions that the model computer empower them to understand processes better. Are also necessary Property detectors based on both macroscopic and can be used at a semi-microscopic level. On the macrosko The process abstraction is level with the help of the WPE's and the process network generated.

3.1 Object transformations

Some simple "topographic" object transformations are shown in Figure 6. Learning at a low level of understanding can be made possible by measuring the time spans between the start of heating and bursting of the vessel or a heating defect event. Better learning is achieved by making the vessel transparent so that it can "see" the local points of contact (e.g. between heating and water) and the current water level.

3.2 Understanding the process mechanisms

A very good learning the process conditions and the process function and the dynamic behavior is achieved by means of elements of physical parameters, which are recognized by the property of the model detectors computer, see Figures 5 and 6 (below). - The preconditions for learning are shown in these pictures.

3.3 Property detectors

Fault detection devices are necessary to find the attributes of the elementary process functions. A result example of a simple property detector is shown in Figure 7 (analyzer of the vessel hardware and the media and the environment of the vessel) and in Figure 8 (display of liquid contamination).

But more complex detectors are also needed, e.g. B. to find sources and sinks of heat or pressure or transport properties, etc. This property shaft detectors are part of the current development.

3.4 Intelligent physical elements

Act the physical elements (IPE's) introduced in (4) according to physical rules, which are based on the symbolic ler nens from the processing of cellular data of the semi-microscopic levels. Then generate the IPE's dynamic processes that depend on the neighborhood of the IPE’s are. Procedures for self-adapting rules can be found in Development.

3.5 Moving process elements

The use of individualized (or individualized groups of) migrating process elements can be seen in Figure 9. The importance of the WPE's is to create an abstract dynamic macroscopic process hypothesis for each case considered by the model computer. These process hypotheses are then adopted by the process network in order to obtain a process model that is as abstract as possible at a macroscopic level, see Figure 9 (below). The mechanisms of the WPE's and the process network are now to be modeled.

4. Configuration of the plant and process model computer

A diagram of the configuration is given in Figure 10.

5. Developments

The development of the components of the model calculator shown above is in its early stages and is controlled by various computers simulations are continued (sequential and later parallel simulations), as well as the creation and testing of proto typical optoelectronic main components of the model computer. With optoelectronic components, a high-grade parallel should Information processing possible and a very fast process Modeling be possible.  

6. References.

(1) W. Bastl, H. Märkl: The Key Role of Advanced Man-Machine Systems for Future Nuclear Power Plants. IAEA Int. Conf. on Man-Machine Interface in the Nuclear Industry, Tokyo, Feb. 1988
(2) K. Kato, et. al .: MITI Project on Advanced Man-Machine System for Nuclear Power Plants. Proc Int. Seminar on Human Interface, Kyoto, Feb. 1988
(3) JA Hassberger, JC Lee: Simulation-Based Expert System for Nuclear Power Plant Diagnostics. Nucl. Sc. Engn. 102: 153-171 (1989)
(4) D. Vetterkind: Automatic Modeling of Plant Disturbances and Failure Limitation Procedures by a Calculus of Functional Nets. IAEA Int. Conf. on Man-Machine Interface in the Nuclear Industry, Tokyo, Feb. 1988
(5) U. Frisch, B. Hasslacher, Y. Pomeau: Lattice-Gas Automata for the Navier Stokes Equation. Phys. Rev. Lett. 56: 1505-1508 (1986)
(6) S. Wolfram: Cellular Automaton Fluids 1: Basic Theory. J. of Statist. Physics 45: 471-525 (1986)
(7) U. Grigull, J. Straub, P.ieberer: Staem Tables in SI-Units. Springer, 1984
(8) A. Elsner: Interaction Potentials of a Saturated Fluid. Phys. Lett. 138: 168-172 (1989)
(9) JL Horner: Optical Signal Processing. Academic Press, 1987

Claims (15)

The invention is a process model computer for hydrodynamic and thermohydraulic processes and process disturbances in technical systems, characterized in that 1. at the microscopic level using the mechanisms of cellular automata in a two-dimensional grid-gas model, which is already known (see quotes in the description: 5, 6), in addition to the liquid particles also steam particles, Pressure particles and energy particles (preferably thermal energy or kinetic energy particles) can be used. 2. that the particles given according to A1 different discrete Can have speeds. 3. from the grid named according to A1 and A2, which is preferably hexagonal, supercells are now formed in a grid arranged in parallel therewith with the aid of coupling points and raking points, which contain several grid points of the first grid and in which the phase transitions are liquid ↔ vaporous be carried out (see Fig. 4) in such a way that the parameters of the steam pressure tables (quotation 7 of the description) apply approximately. 4. in a semi-microscopic cell model from the mi microscopic model (A1 to A3) process data obtained wisely averaged or added up or to physical parameters tern (temperature, enthalpy, speed, etc.) processed will. 5. to evaluate the cell model specified according to A4 Various property detectors can be used both simple hardware or media constellations (e.g. Liquid level in a vessel (represented by a compu Readable matrix, see A6)) or process functions (source, Sink, transport, loss, parameter difference, open / closed circuit, cross-circuit (e.g. from heat transfer Medium and heat), etc.) in the cell model (A4), available. 6. intelligent physical elements (IPE’s) who introduced at the semi-microscopic level, which is based on the input and / or learned rules in the cell model (A4) accordingly behave in their immediate neighborhood.   7. in a macroscopic process model related formalized components (e.g. steam generator) or replacement components (e.g. liquid leak) the migrating process elements elements (WPE's) are used according to the specified rule sentence or learned rule sets according to their neighbor shaft behave. 8. that the hardware and working media of the modeled technical system formalized as computer-readable matrices will. 9. an error generator is used, which can be used with the Orders: remove, destroy, create, separate, connect, trans portiere, source, sink, loss / leak, return, stop, reverse direction, walk together, swap, etc., Predefinable errors in the computer-readable named after A8 Matrices (called topographies) so that the model can also model process disturbances. 10. Using object transformations on a semi-microscope ical, macroscopic or even at the microscopic level given hardware and working media images changed can be that after these changes in the model computer ongoing model processes (e.g. IPE processes or WPE processes or processes in the microscopic model) of the properties detectors (see A5) then recognized in a process evolution module then lead to learning processes; these object transformations are also from the above Process evolution model itself initiated. 11. A process abstraction takes place in a process network with the help of predefined or learned process functions (see Fig. 9). 12. the necessary with the help of a test and management module Introduction of target parameters and further target information as there are also restrictions and moreover here from this Module from the desired interaction of all other parts of the Model computer is managed. 13. With the help of a diagnostic module the incoming current Process data (of the technical system to be modeled) be prepared (in collaboration with the other parts of the model computer) to troubleshoot the fault in progress process. 14. with the help of the module "Countermeasures" (in cooperation with the other parts of the model computer) countermeasures are proposed and developed to address the Eliminate process problems or limit their effects.