WO2023139081A1 - Procédé de fabrication d'un appareil de surveillance d'une installation d'automatisation - Google Patents

Procédé de fabrication d'un appareil de surveillance d'une installation d'automatisation Download PDF

Info

Publication number
WO2023139081A1
WO2023139081A1 PCT/EP2023/051047 EP2023051047W WO2023139081A1 WO 2023139081 A1 WO2023139081 A1 WO 2023139081A1 EP 2023051047 W EP2023051047 W EP 2023051047W WO 2023139081 A1 WO2023139081 A1 WO 2023139081A1
Authority
WO
WIPO (PCT)
Prior art keywords
correlation
bridge
devices
value
assembly
Prior art date
Application number
PCT/EP2023/051047
Other languages
German (de)
English (en)
Inventor
Simon WEILANDT
Original Assignee
Siemens Aktiengesellschaft
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Siemens Aktiengesellschaft filed Critical Siemens Aktiengesellschaft
Publication of WO2023139081A1 publication Critical patent/WO2023139081A1/fr

Links

Classifications

    • G—PHYSICS
    • G05—CONTROLLING; REGULATING
    • G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
    • G05B23/00—Testing or monitoring of control systems or parts thereof
    • G05B23/02—Electric testing or monitoring
    • G05B23/0205—Electric testing or monitoring by means of a monitoring system capable of detecting and responding to faults
    • G05B23/0218—Electric testing or monitoring by means of a monitoring system capable of detecting and responding to faults characterised by the fault detection method dealing with either existing or incipient faults
    • G05B23/0224—Process history based detection method, e.g. whereby history implies the availability of large amounts of data
    • G05B23/024—Quantitative history assessment, e.g. mathematical relationships between available data; Functions therefor; Principal component analysis [PCA]; Partial least square [PLS]; Statistical classifiers, e.g. Bayesian networks, linear regression or correlation analysis; Neural networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/042—Knowledge-based neural networks; Logical representations of neural networks

Definitions

  • the invention relates to a method for producing a monitoring device for an automation system.
  • the invention also relates to a computer program product that is designed to carry out such a method.
  • the invention also relates to a monitoring device that is produced using such a method and a superordinate control unit that is equipped with such a computer program product.
  • the invention also relates to an automation system that has a monitoring device that is manufactured using a corresponding method.
  • Patent application CN 111967486 A discloses a fault diagnosis method based on sensor fusion.
  • the error diagnosis method includes data being recorded by sensors during ongoing operation of a complex system.
  • the signals containing the data are processed and made available as one-dimensional data sets.
  • the one-dimensional data sets are combined into a two-dimensional data array.
  • the two-dimensional arrays of data are used as input to a self-learning algorithm to create a neural network.
  • Automation systems have an increasing number of sensors and actuators, resulting in increasing complexity for their manufacture and operation. As a result, the production costs for process automation systems increase, as does the susceptibility to errors. This applies under inter alia for monitoring devices that are intended to monitor the operation of such automation systems. There is a need for a way of speeding up the provision of such a monitoring device and at the same time reducing the associated susceptibility to errors.
  • the task is solved by a method according to the invention for producing a monitoring device for an automation system.
  • the automation system includes a plurality of devices, through the interaction of which the function of the automation system is provided.
  • the devices can be in the form of actuators which act on a system process during operation.
  • the devices can also be in the form of sensors, by means of which a variable associated with the system process can be detected.
  • the devices are connected to one another indirectly via a control unit or directly. Accordingly, the devices can be controlled and measured values can be read from them, i.e. they can be received.
  • the method includes a first step in which devices are made available to a first assembly of the automation system in an active operating state.
  • An active operating state means that the devices in the first assembly have been installed as intended and are functional.
  • the first module can be a subset of all devices that are used in the automation system.
  • the method also includes a second step in which candidate correlations are created between a plurality of pairs of devices in the first assembly.
  • the pairs of devices are consequently linked via the respective candidate correlation.
  • a candidate correlation is to be understood as a quantitative connection which, at the time of creation, is free of an empirically or theoretically created correlation value.
  • the candidate correlation stipulates for the further procedure that during the second step an at least indirect physical physical connection between the devices connected in this way is considered possible.
  • the candidate correlations are created independently of known or suspected physical relationships between the respective devices. Such physical relationships can be known beforehand, for example, by a user who carries out the method, or can be presumed by him beforehand. The creation of the candidate correlations is thus carried out for the respective devices in general.
  • the method also includes a third step in which at least one first device of the first assembly is actuated.
  • the first device is in the form of an actuator which is suitable for influencing the system process.
  • Actuating the first device causes a physical effect that can be detected in the form of measured values of the devices that are linked to the first device via candidate correlations.
  • An associated correlation value is determined for the measured values, based on the respective candidate correlation.
  • the correlation value is designed to quantitatively reflect an intensity of a connection between the actuation of the first device and the corresponding device linked via the candidate correlation.
  • the method also includes a fourth step, in which a zero correlation between the first device and a second device connected to it, ie linked via a candidate correlation, is detected.
  • the zero correlation is recognized when the associated correlation value, which is determined in the third step, falls below an adjustable correlation threshold value in terms of absolute value.
  • the correlation threshold is adjustable by a user and/or an algorithm.
  • the correlation threshold value specifies below which absolute limit an at least apparently existing physical connection between the first and second device is to be classified as negligible. Accordingly, in the third step, the candidate correlation present between the first and second device is classified as zero correlation and deleted.
  • the method has a fifth step, in which an operational correlation between the first and second device is detected if the correlation value, which is determined in the third step, exceeds the adjustable correlation threshold value in terms of absolute value.
  • the present candidate correlation is classified as operationally relevant in the fifth step. Accordingly, there is a physical connection between the first and second device, which is to be reflected in the monitoring device to be produced.
  • the candidate correlation between the first and second device is classified as an operational correlation.
  • the operational correlation between the first and second devices identified in this way belongs to a first sub-network of a neural network.
  • the neural network in turn belongs to the monitoring device to be produced.
  • the first subnet is stored in the monitoring device.
  • the method according to the invention creates a large number of candidate correlations, so that physical relationships unexpected by a user are essentially automatically recognized experimentally. Accordingly, a first sub-network created in this way or neural network of a surveillance device an increased degree of model fidelity. Because the claimed method is only run through for devices in a first assembly, the number of candidate correlations that are to be checked to determine whether they are zero correlations or operational correlations is reduced.
  • the method according to the invention can therefore also be carried out reliably on hardware with low computing power.
  • the thus manufactured monitoring device with the neural network comprising the first sub-network is further provided for use. By storing the neural network in the monitoring device, this is produced in the desired form.
  • the correlation threshold value can be in the form of an absolute value or a relative value.
  • An absolute value represents a simple criterion by which zero correlations can be determined quickly.
  • empirical values from structurally identical or structurally similar first assemblies of other automation systems can be used in a targeted manner.
  • a relative value as a correlation threshold value makes it possible to define a sliding limit between zero correlations and operational correlations.
  • the relative value can be used to define that only such candidate correlations are recognized as operating correlations, that is to say classified, which significantly exceed the zero correlations with their lower correlation values in terms of absolute value.
  • parameters of statistical distributions or distribution parameters such as a Gini coefficient can be used.
  • a Gini coefficient can be used.
  • the claimed method can consequently be easily adapted in terms of speed and adaptivity.
  • each device in the first assembly can be linked to each other device in the first assembly via a candidate correlation after the second step. Consequently, in the second step, all combinatorially possible candidate correlations between the devices of the first assembly are generated. This simplifies the second step and can essentially be carried out automatically. Inputs by the user or calculations of further algorithms are therefore not necessary for the second step. Furthermore, candidate correlations are also obtained for unexpected physical relationships. begets . This prevents the user from overlooking such physical relationships, as a result of which the monitoring device to be produced has a further increased degree of imaging fidelity. Furthermore, the method is accelerated by such a configuration of the second step.
  • the adjustable correlation threshold value is adjusted until a predefinable complexity value of at least the first sub-network is reached. This can be done, for example, in several passes of the claimed method.
  • the complexity value depends on the number of devices in the first module and the operating correlations between them. The complexity value is all the higher, the higher the ratio between the number of operational correlations and the number of devices in the first sub-network or of the first assembly is .
  • the higher the correlation threshold value is set the fewer candidate correlations are categorized as operational correlations, ie classified. The fewer operational correlations in the first sub-network or the first module is present, the lower the demands on the computing power for the monitoring device to be produced.
  • the first subnet or the fewer operational correlations present in the neural network the easier and quicker it is for the user to understand during maintenance or modification of the monitoring device.
  • the more operational correlations in the first sub-network or are present in the neural network the higher the imaging fidelity of the monitoring device to be produced.
  • the adjustment of the adjustable complexity threshold value to reach the predefinable complexity value can be carried out as an automatically running optimization of the first sub-network or be formed of the neural network. Consequently, the claimed method can quickly and automatically be adjusted in terms of computing effort, and thus hardware requirements.
  • the claimed method is consequently- s can be adapted to different applications and has a wide range of possible uses.
  • the claimed method in particular the first to fifth steps, can also be carried out for a second assembly of devices. Analogous to the first module and the first sub-network, a corresponding second sub-network is produced for the second module. Devices are linked to one another in the second sub-network via operational correlations, which are identified analogously to the operational correlations in the first sub-network.
  • the method can also include a sixth step, in which candidate bridge correlations are established between at least one device in the first assembly group and a plurality of devices in the second assembly group.
  • the candidate bridge correlations connect at least the device of the first assembly in pairs with devices of the second assembly.
  • all devices in the first assembly can be linked in pairs with all devices in the second assembly via candidate bridge correlations.
  • a candidate bridge correlation is to be understood as meaning a quantitative relationship corresponding to a candidate correlation within the first assembly group, which at the time of creation is free of an empirically or theoretically created correlation value.
  • the candidate bridge correlation stipulates for the further method that during the sixth step an at least indirect physical connection between the devices connected to one another in this way is considered possible.
  • the candidate bridge correlations are created independently of known or suspected physical relationships between the respective devices. Such physical relationships can be known beforehand, for example, by a user who carries out the method, or can be presumed by him beforehand.
  • the creation of the candidate bridge correlations is thus carried out in a general way for the respective devices.
  • the method can further include a seventh step, in which the at least one device of the first assembly is actuated according to the sixth step. Furthermore, measured values are recorded from devices in the second assembly, which are linked to the at least one device in the first assembly via the candidate bridge correlations. A bridge correlation value is also determined based on the measured values.
  • the bridge correlation value corresponds functionally to the correlation value according to the fourth step.
  • the bridge correlation value is designed to quantitatively reproduce an intensity of a connection between the actuation of the at least one device of the first assembly and the corresponding device of the second assembly linked via the candidate bridge correlation.
  • the method can include an eighth step, in which a zero bridge correlation is detected if the bridge correlation value recorded in the seventh step falls below an adjustable bridge correlation threshold value in terms of absolute value.
  • the method can include a ninth step, in which an operational bridge correlation is detected if the bridge correlation value exceeds the adjustable bridge correlation threshold value in terms of absolute value.
  • the bridge correlation threshold value specifies the limit below which an at least apparently existing physical connection between the at least one device of the first module and a device of the second module is to be classified as negligible. Accordingly, in the eighth step, the candidate correlation present between the at least one device of the first module and the device of the second module is classified as zero correlation and deleted.
  • the bridge correlation threshold value thus corresponds functionally to the correlation threshold value according to the fourth step.
  • an operating bridge correlation recognized in the ninth step is stored. At least one detected operational bridge correlation creates a link between the first and the second sub-network.
  • the first and second partial network are stored linked to one another in this way and belong to the neural network of the monitoring device to be produced.
  • the claimed method can thus be carried out independently of one another for a first and a second assembly and the sub-networks produced in the process, ie the first and second sub-network, can be linked to form a neural network.
  • the operational bridge correlations are essentially similar to the operational correlations within the first and second assemblies. Consequently, the neural network with the first and second sub-network is homogeneous. Furthermore, the number of candidate bridge correlations is relatively small.
  • the creation of the neural network can thus be broken down into compact sub-tasks using the claimed method.
  • the claimed method is readily scalable to complex automation systems that require a large number of devices or have assemblies. Not only is the production of the monitoring device accelerated by means of the claimed method, but also the production of the associated automation system.
  • this is carried out while the first and/or second assembly is being manufactured.
  • the first or second assembly represent sections or functional blocks of the automation system that can be manufactured separately.
  • the first partial network can be produced by means of the claimed method when the first assembly has reached its functional capability in production.
  • the second sub-network in connection with the second module.
  • the functionality of the first or The second sub-network can consequently be reliably checked in an early phase of the production of the automation system. The process reliability in the production of the monitoring device, and thus the automation system, is thus further increased.
  • the bridge correlation threshold value can be adjusted in order to achieve a predefinable complexity value, in particular of the neural network.
  • the way in which the bridge correlation threshold works essentially corresponds to the correlation threshold according to the fourth or fifth step of the method.
  • the higher the bridge correlation threshold the lower the number of operational bridge correlations that are detected in the ninth step.
  • the lower the number of operational bridge correlations the simpler the neural network and the lower the computing power required to operate it.
  • the bridge correlation threshold or . the complexity value can be selected based on an indication of the available computing power of the monitoring device to be manufactured. Accordingly, the claimed method can be easily adapted to produce a high-performance neural network. As a result, the real-time capability of the monitoring device to be manufactured is increased by specifying the complexity value of the neural network.
  • a virtual sensor can be provided between the first device and the second device of the first assembly based on the operational correlation that exists between them.
  • the operational correlation can show that the first and second devices of the first assembly have the same physical relationship or show different embodiments of the same physical phenomenon.
  • a water level in a water tank corresponds to a pressure bottom of water tank . Consequently, a virtual sensor can be created based on the bridge correlation, which is suitable for the first or second device to replace in case of failure .
  • the claimed method thus implements an intelligent use of data, in particular a so-called sensor fusion. Consequently, the claimed method has increased functional integration.
  • a virtual sensor can also be provided based on the operational bridge correlation between a device of the first assembly and a device of the second assembly.
  • the virtual sensor is provided in a manner corresponding to the provision of the virtual sensor based on the first and second device of the first assembly.
  • the first to fifth steps can be carried out repeatedly and the first subnetworks determined can be displayed to the user for selection.
  • the correlation threshold value and/or the bridge correlation threshold value can be changed in the steps of the method.
  • the determined first sub-networks can be displayed to the user in a graphical representation simultaneously or one after the other, so that the complexity of the first sub-network can be clearly detected.
  • the method can also be designed so that the user selects a first subnetwork that is determined.
  • the setting of the complexity of the neural network can be influenced precisely by the user.
  • such an experience of the user with neural networks can be included in the claimed method.
  • the quality of the neural network, and thus of the monitoring device to be produced, can thus be increased.
  • second partial networks can also be displayed and made available to the user for selection.
  • the underlying task is also solved by a computer program product that is designed to create a neural network.
  • the computer program product can have a monolithic design and can thus be executed on a single hardware platform.
  • the computer program product can also have a modular structure and be designed as a plurality of subprograms that can be executed on different hardware platforms and implement the functionality of the computer program product through mutual data exchange.
  • the computer program product can also be designed as software, as firmware or hard-wired, for example as a chip, integrated circuit or FPGA, or as a combination thereof.
  • the computer program product is designed to carry out the method according to one of the embodiments outlined above.
  • the object presented above is also achieved by a monitoring device according to the invention.
  • the monitoring device is designed to monitor an automation system that includes a plurality of devices in at least one module.
  • the monitoring device has an input interface via which inputs can be made available to a digital twin of the automation system.
  • a digital twin is to be understood in particular as one outlined in more detail in US 2017/286572 A1.
  • the input interface for the digital twin is coupled to a neural network of the monitoring device.
  • the neural network is designed using a method according to one of the embodiments outlined above.
  • a higher-level control unit which is designed to create a neural network that is suitable for monitoring an automation system.
  • the superordinate control unit can be coupled to the control unit of the automation system.
  • the higher-level control unit can be designed as a master computer or as a computer cloud, for example, and be connected to the automation system via a communicative data connection.
  • the higher-level control unit is equipped with a computer program product according to one of the embodiments presented above in order to create the neural network.
  • the higher-level control unit can also include the monitoring device or vice versa.
  • an automation system which comprises a plurality of devices which are connected to one another directly or indirectly via a control unit.
  • the automation system also includes a monitoring device that is suitable for monitoring the operation of the automation system.
  • the monitoring device is designed according to one of the embodiments outlined above.
  • the automation system according to the invention can be produced, in particular put into operation, quickly and reliably by the monitoring device used therein.
  • the claimed automation system can be designed, among other things, as a chemical production system, as a food processing system, as a refinery, as a production line, as a power plant, as a supply network, for example an electricity network, a water network, a gas network, or a telecommunications network, as a material analysis system, in particular as a gas chromatograph or as a gas analyzer, or as a traffic control system.
  • a chemical production system for example an electricity network, a water network, a gas network, or a telecommunications network
  • a material analysis system in particular as a gas chromatograph or as a gas analyzer, or as a traffic control system.
  • FIG. 1 schematically shows an embodiment of the claimed method in a first stage
  • FIG. 7 shows a neural network according to the claimed method in a monitoring mode.
  • FIG. 1 An embodiment of the claimed method 100 is shown schematically in FIG. 1 in a first stage.
  • the method 100 is used to produce a monitoring device 80 for an automation system 30, not shown in detail in FIG.
  • a plurality of devices 12 are provided therein, which belong to a first subassembly 11 of the automation system 30 to be manufactured or ready.
  • the devices 12 are designed to act directly or indirectly on a system process 32 that is to be carried out or is being carried out with the automation system 30 .
  • the devices 12 are provided in an active operating state in a first step 110 of the method 100 .
  • the devices 12 are thus installed as components of the first assembly 11 and are in a functional state.
  • the devices 12 can be embodied, for example, as actuators, sensors, control units or other assembly components of the first assembly 11 .
  • One of the devices 12 is in the form of a local control unit 41 which is assigned to the first assembly 11 .
  • a higher-level control unit 60 is provided, in which a computer program product 50 is executably stored.
  • a second step 120 is carried out, in which candidate correlations 22 are created between the devices 12. This takes place as part of an implementation of the computer program product 50 .
  • the candidate correlations 22 thus exist as objects within the computer program product 50 .
  • the devices 12 are to be understood in the same way as their representations as objects in the computer program product 50 .
  • a candidate correlation 22 expresses that within the scope of the monitoring device 80 to be produced in the sense of a neural network 40 not shown in detail in FIG.
  • the candidate correlations 22 are generated in the second step 120 between each combinatorially possible pair of devices 12 .
  • the second step 120 can thus be carried out in a simple manner. Elaborate estimates, for example from experience databases or by a user, are unnecessary.
  • the candidate correlations 22 are specified in the second step 120 by the higher-level control unit 60 .
  • FIG. 2 shows a second stage of the claimed method 100, which follows the first stage according to FIG. Consequently, FIG. 2 assumes that the second step 120 has already been completed.
  • a third step 130 of the method 100 takes place, in which a first device 14 receives a command for an actuation 15 from the superordinate control unit 60 , and thus also from the computer program product 50 .
  • the first device 14 is designed as an actuator 13 and is able to act on a physical quantity present in or in the area of the first assembly 11 .
  • a reaction is recorded by the devices 12 and transmitted to the superordinate control unit 60 in the form of measured values 23 . Based on the measured values 45, correlation values 25 for the candidate correlations 22 are determined.
  • the correlation value 25 of a candidate correlation 22 is as one Measure formed as to how far the measured value 23 of the respective device 12 follows the actuation 15 of the first device 14 .
  • the correlation values 25 for the respective candidate correlations 22 can also be compared with an adjustable correlation threshold value 27 .
  • the correlation threshold values 27 can be set for all candidate correlations 22 or individually by a user or an algorithm that belongs to the computer program product 50 in the superordinate control unit 60 .
  • a third stage of the claimed method 100 is shown in FIG.
  • the third stage follows the second stage shown in FIG.
  • the correlation values 25 determined in the third step 130 according to FIG. 2 are evaluated.
  • a zero correlation 24 is detected in a fourth step 140 for correlation values 25 which are below the adjustable correlation threshold value 27 .
  • a zero correlation 24 in the sense of the method 100 is to be understood as meaning that the direct or indirect physical connection assumed via the associated candidate correlation 22 is classified as being insignificant for the neural network 40 to be produced.
  • the zero correlations 24 are deleted in the fourth step 140 and thus do not become part of the neural network 40 to be produced.
  • a fifth step 150 takes place in the method 100, in which an operational correlation 26 is detected between two devices 12 if the associated correlation value 25 exceeds the correlation threshold value 27 in terms of absolute value.
  • the operational correlations 26 are classified within the meaning of the method 100 as direct or indirect physical relationships between the corresponding devices 12, which are relevant for the neural network 40 to be produced, and thus the monitoring device 80 to be produced, in order to show the mode of operation and operation of the first assembly 11. place .
  • the operational correlations 26 are stored and define a first sub-network 10 with the associated devices 12 as nodes.
  • the first sub-network 10 also represents a neural network.
  • a virtual sensor 28 is provided between the first device 14 and a second device 16 of the first assembly 11 .
  • the computer program product 50 recognizes, based on the correlation values 2 according to FIG. 2, that the first device 14 and the second device 16 record different embodiments of the same technical process. This can be recognized, for example, using a virtualization threshold value with which the correlation value 25 is compared.
  • the operating correlations 26 identified in the fifth step 150 are stored as a first sub-network 10 and stored in the superordinate control unit 60 as part of the neural network 40 to be produced.
  • the neural network 40 which includes the first sub-network 10, is suitable for use in the monitoring device 80 to be produced.
  • a fourth stage of the embodiment of the claimed method 100 is also shown in FIG. It is assumed that the first step 110 to the fifth step 150 have already been completed and a first sub-network 10 has been created, which includes devices 12 of the first assembly 11 of the automation system 30 to be produced. It is also assumed that the first step 110 to the fifth step 150 is also carried out analogously in connection with devices 12 of a second assembly 21 of the automation system 30 . Accordingly, there is a second sub-network 20 which is assigned to the second module 21 and includes its devices 12 . In the second sub-network 20 , the associated devices 12 are linked to one another via operational correlations 26 , analogously to the first sub-network 10 .
  • a sixth step 160 of the method 100 takes place, in which in pairs between a device 12 of the first assembly 10 candidate bridge correlations 32 to a plurality of devices 12 of the second assembly 21 are produced.
  • a candidate bridge correlation 32 corresponding to a candidate correlation 22 as shown in FIG. 1 and FIG. 2, represents an assumption that an indirect or direct physical relationship between the corresponding devices 12 is classified as possible.
  • each combinatorially possible candidate bridge correlation 32 is created between the device 12 of the first assembly 11 and the devices of the second assembly 21 . This is done using the computer program product 50 , which is stored in an executable form in the superordinate control unit 60 .
  • the sixth step 160 can also be carried out analogously for each device 12 of the first assembly 11 . These corresponding executions of the sixth step 160 are not shown in FIG. 6 merely for the sake of better clarity.
  • FIG. 5 shows a fifth stage of the embodiment of the method 100 claimed.
  • the fifth stage follows the fourth stage according to FIG.
  • the device 12 of the first assembly 11, which is linked via candidate bridge correlations 32 to devices 12 of the second assembly 21, is actuated.
  • the actuation 15 of this device 12 is brought about in a seventh step 170 by the computer program product 50 in the superordinate control unit 60 .
  • the actuated device 12 of the first assembly 11 is designed as an actuator 13 .
  • the actuation 15 acts on a physical variable that can be directly or indirectly detected by the devices 12 of the second assembly 21 with varying degrees of accuracy.
  • the actuation 15 in the seventh step 170 acts essentially analogously to the actuation 15 in the third step 130 .
  • the reactions of the individual devices 12 of the second assembly 21 are transmitted to the superordinate control unit 60 in the form of measured values 23 .
  • a bridge correlation value 35 is determined for each of the candidate bridge correlations 32 based on the respective measured values 23 .
  • the bridge correlation value 35 is a quantitative measure of how closely the response detected on the respective device 12 of the second assembly 21 follows the actuation 15 in the seventh step 170 . je higher the bridge correlation value 35 determined in this way is in absolute terms, the closer is the indirect or direct physical connection between the actuation 15 and the measured value 23 .
  • the bridge correlation values 35 can be compared with adjustable bridge correlation threshold values 37 .
  • the bridge correlation threshold values 37 according to FIG. 5 are predefined for the candidate bridge correlations 32 .
  • the respective bridge correlation values 35 achieved are compared with the respective bridge correlation threshold values 37 . If a bridge correlation value 35 falls below the associated bridge correlation threshold value 37 in absolute terms, the presence of a zero bridge correlation 34 is detected in an eighth step 180 . If the bridge correlation value 35 exceeds the respective bridge correlation threshold value 37 in terms of absolute value, the presence of an operational bridge correlation 36 is detected in a ninth step 190 .
  • a sixth stage of the claimed method 100 is shown in FIG.
  • the sixth stage follows the fifth stage according to FIG.
  • the zero bridge correlations 34 recognized in the eighth step 180 are deleted therein.
  • the operating bridge correlations 36 identified in the ninth step 190 are stored, so that the first and second partial networks 10 , 20 are connected to one another to form the neural network 40 .
  • the neural network 40 which includes the first and second partial network 10, 20, is stored in the seventh stage in the higher-level control unit 60 and made available for the monitoring device 80 to be produced.
  • the monitoring device 80 is suitable for monitoring the operation of the automation system 30, which includes the first and second assemblies 11, 21.
  • FIG. 7 shows the neural network 40, which is produced according to the first to ninth steps 110, 190 and belongs to a monitoring device 80, in a monitoring operation of the associated automation system 30.
  • the intended start-up layer process 33 so that measured values 23 are generated that characterize the system process 33 in more detail.
  • the neural network 40 which includes the first and second partial network 10 , 20 , uses the measured values 23 to determine a result that is provided as an input 42 for a digital twin 70 .
  • the neural network 40 thus serves as an input interface for the digital twin 70 and is at least functionally assigned to the digital twin 70 .
  • the automation system 30 can be simulated in advance by the digital twin 70, ie future operating states can be calculated in advance.
  • the digital twin 70 also belongs to the monitoring device 80 which is manufactured by means of the claimed method 100 .
  • the monitoring device 80 is included in the superordinate control unit 60 which is used to create the monitoring device 80 .
  • the monitoring device 80 is thus functionally a component of the superordinate control unit 60 and is executed on the same hardware platform.
  • the monitoring device 80 can be produced quickly and reliably using the claimed method 100 .
  • increased imaging fidelity is achieved by means of the neural network 40 when the operation of the automation system 30 is simulated.

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Artificial Intelligence (AREA)
  • Evolutionary Computation (AREA)
  • Mathematical Physics (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Biophysics (AREA)
  • Molecular Biology (AREA)
  • Biomedical Technology (AREA)
  • Health & Medical Sciences (AREA)
  • Computational Linguistics (AREA)
  • Data Mining & Analysis (AREA)
  • General Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Computing Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Software Systems (AREA)
  • Automation & Control Theory (AREA)
  • Testing And Monitoring For Control Systems (AREA)
  • General Factory Administration (AREA)

Abstract

L'invention concerne un procédé (100) de fabrication d'un appareil de surveillance (80) pour une installation d'automatisation (30). L'installation d'automatisation (30) comprend une pluralité de dispositifs (12) qui sont reliés les uns aux autres indirectement ou directement par l'intermédiaire d'une unité de commande (41. Le procédé (100) comprend une première étape (110) dans laquelle les dispositifs (12) d'un premier ensemble (11) de l'installation d'automatisation (30) sont dans un état de fonctionnement actif. Le procédé (100) comprend une deuxième étape (120) dans laquelle des corrélations candidates (22) entre une pluralité de paires de dispositifs (12) dans le premier ensemble (11) sont créées. Le procédé (100) comprend également une troisième étape (130) dans laquelle au moins un premier dispositif (12) sous la forme d'un actionneur (13) est actionné. Des valeurs de mesure (23) des dispositifs (12) reliés à l'actionneur (13) sont enregistrées et une valeur de corrélation (25) associée est déterminée. Le procédé (100) comprend en outre une quatrième étape (140) qui consiste à identifier une corrélation nulle (24) entre le premier dispositif (12) et un second dispositif (16) connecté à ce dernier si la valeur absolue de la valeur de corrélation (25) tombe sous une valeur seuil de corrélation réglable (27), et à supprimer la corrélation nulle (24). En variante, le procédé (100) comprend une cinquième étape (150) dans laquelle une corrélation de fonctionnement (26) entre les premier et second dispositifs (14, 16) est identifiée si la valeur absolue de la valeur de corrélation (25) dépasse la valeur seuil de corrélation réglable (27). Dans le procédé (100), la corrélation de fonctionnement (26) est également stockée dans un premier sous-réseau (10) d'un réseau neuronal (40) de l'appareil de surveillance (80).
PCT/EP2023/051047 2022-01-21 2023-01-18 Procédé de fabrication d'un appareil de surveillance d'une installation d'automatisation WO2023139081A1 (fr)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
DE102022200694.1A DE102022200694A1 (de) 2022-01-21 2022-01-21 Herstellungsverfahren für eine Überwachungsvorrichtung einer Automatisierungsanlage
DE102022200694.1 2022-01-21

Publications (1)

Publication Number Publication Date
WO2023139081A1 true WO2023139081A1 (fr) 2023-07-27

Family

ID=85199513

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/EP2023/051047 WO2023139081A1 (fr) 2022-01-21 2023-01-18 Procédé de fabrication d'un appareil de surveillance d'une installation d'automatisation

Country Status (2)

Country Link
DE (1) DE102022200694A1 (fr)
WO (1) WO2023139081A1 (fr)

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP4293457A1 (fr) 2022-06-15 2023-12-20 Siemens Aktiengesellschaft Procédé de surveillance, produit programme informatique, unité de surveillance, dispositif d'analyse de gaz et utilisation d'une intelligence artificielle

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20060206291A1 (en) * 2005-03-11 2006-09-14 Bash Cullen E Commissioning of sensors
WO2008064616A1 (fr) * 2006-11-27 2008-06-05 Siemens Aktiengesellschaft Procédé et système de diagnostic pour le diagnostic d'un système technique
US20170286572A1 (en) 2016-03-31 2017-10-05 General Electric Company Digital twin of twinned physical system
CN111967486A (zh) 2020-06-02 2020-11-20 安徽三禾一信息科技有限公司 一种基于多传感器融合的复杂装备故障诊断方法

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2013000504A1 (fr) 2011-06-28 2013-01-03 Siemens Aktiengesellschaft Procédé d'aide à la mise en service d'un système technique

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20060206291A1 (en) * 2005-03-11 2006-09-14 Bash Cullen E Commissioning of sensors
WO2008064616A1 (fr) * 2006-11-27 2008-06-05 Siemens Aktiengesellschaft Procédé et système de diagnostic pour le diagnostic d'un système technique
US20170286572A1 (en) 2016-03-31 2017-10-05 General Electric Company Digital twin of twinned physical system
CN111967486A (zh) 2020-06-02 2020-11-20 安徽三禾一信息科技有限公司 一种基于多传感器融合的复杂装备故障诊断方法

Also Published As

Publication number Publication date
DE102022200694A1 (de) 2023-08-10

Similar Documents

Publication Publication Date Title
DE19742446B4 (de) Fehlerdiagnoseverfahren
EP0894304B2 (fr) Procede de diagnostic automatique de systemes techniques prenant en consideration une acquisition de connaissances efficace et un traitement efficace concernant un temps de fonctionnement
EP0893746A2 (fr) Système et méthode de diagnostic d'un processus
EP2685382B1 (fr) Procédé et dispositif de création et de test d'un programme d'appareil de commande
DE102012110710A1 (de) Verfahren zur Integration von Modellen eines Gesundheitsmanagementsystems für Verkehrsmittel
EP2990892B1 (fr) Procédé de liaison d'une interface d'entrée/de sortie d'un appareil d'essai destiné à l'essai d'un appareil de commande
DE4438854C2 (de) Überwachungssystem für eine technische Anlage
EP0966703B1 (fr) Procede d'analyse assistee par ordinateur de defaillances de capteurs et/ou d'actionneurs dans un systeme technique
WO2023139081A1 (fr) Procédé de fabrication d'un appareil de surveillance d'une installation d'automatisation
DE102016124585A1 (de) Verfahren und Vorrichtungen zur Verwendung einer analytischen/statistischen Modellierung für eine kontinuierliche Prozessüberprüfung (Continued Process Verification, CPV)
DE102014223810A1 (de) Verfahren und Assistenzsystem zur Erkennung einer Störung in einer Anlage
EP1947568A1 (fr) Procédé destiné à l'observation d'un appareil de commande
WO2000033209A2 (fr) Procede et dispositif de conception d'un systeme technique
WO2014140253A1 (fr) Développement d'un modèle d'ordre supérieur
EP1313030A2 (fr) Méthode pour la détermination de structure complexe de corrélation à partir de données de processus et d'installation
DE102009018785A1 (de) Verfahren und Vorrichtungen für eine virtuelle Testzelle
EP1717651B1 (fr) Méthode et système d'analyse des événements relatés à l'opération d'un véhicule
EP1157317B1 (fr) Procede et dispositif pour la reduction d'une pluralite de valeurs mesurees d'un systeme industriel
EP4193227A1 (fr) Dispositif et procédé de détection d'anomalies dans une installation industrielle pour l'exécution d'un processus de production
EP4060433A1 (fr) Procédé et système de prévision du fonctionnement d'une installation technique
EP3944158A1 (fr) Procédé mis en uvre par ordinateur destiné à l'apprentissage d'une machine permettant de faire fonctionner un composant de véhicule et procédé de fonctionnement d'un composant de véhicule
DE19914819A1 (de) Entwicklungswerkzeug
EP3553679A1 (fr) Procédé de diagnostic de panne assisté par ordinateur pour un système technique
DE102017207036A1 (de) Verfahren zur rechnergestützten Analyse des Betriebs eines Produktionssystems
WO2009053282A1 (fr) Procédé de fourniture de modèles

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 23703700

Country of ref document: EP

Kind code of ref document: A1