METHOD FOR PROVIDING AN INTEGRATED MODEL FOR A TECHNICAL SYSTEM

Description

[0001] The invention relates to a method for providing an integrating model for a technical system.

[0002] A technical system is, for example, a machine such as a power converter, an asynchronous machine, a synchronous machine, etc. An asynchronous machine or a synchronous machine is, for example, a motor or a generator. A technical system or machine is also, for example, a pump for fluids or gases, a compressor, a turbo compressor (e.g., for technical gases), a fan (e.g., for induced draft fans or a radiator), etc. A technical system can also be a component of a machine. For example, the stator or rotor of an asynchronous machine is one of its components. A bearing of a motor or generator, or the cooling system of a power converter, are also examples of a technical system. A technical system is also, for example, a technical process for the production and / or processing of a good.For example, a technical process might serve to produce a process gas, supply drinking water, treat drinking water, refine a hydrocarbon-containing product, etc. The present invention relates to a method for providing a model for at least one machine, in particular a machine tool. Furthermore, the present invention relates to a training system, a computer program, and a computer-readable (storage) medium. In addition, the present invention relates to a method for simulating the operation of the machine or the technical system.

[0003] Process plants, such as refineries or factories, are technical systems in which substances are altered with regard to their composition, type, or properties, and can have extremely complex structures. A plant can, for example, consist of a multitude of components, possibly interconnected and / or interdependent, such as valves, sensors, actuators, and / or the like. These components are also technical systems. Such plants are generally controlled by specialized process control systems, particularly computer-based or at least computer-aided, which can take into account the process engineering relationships between the various components. Such process control systems include automation technology, especially automation programs and operating and monitoring programs.Process control systems, also known as plant control systems, are often developed based on individual modules, each assigned to a specific component of the plant and configured to control that component. Such a module can be understood as a standardized template of the control software for a particular component type. When assembling the process control system or plant control system, the module must be adapted to the specific characteristics of the component to ensure correct control of the component and, consequently, the entire plant. Typically, plant control systems developed from individual modules, which also represent technical systems, are tested through simulation before being deployed in a real plant to ensure or, if necessary, improve the functionality of the control system. The plant simulation can be based on models (e.g., simulation models).

[0004] Models from engineering or R&D can also be used in the operational management of motors and inverters. For a drive with a power converter and a motor, for example, a temperature model for the rotor, a model for the bearing, a model for controlling the power semiconductors, etc., is necessary. The calculation and parameterization of models within drive systems and visualizations / controls are usually standalone solutions or are performed on low-level computers, which are typically developed by custom engineers. These solutions require a significant investment of time and effort in programming, maintenance, support, and expansion. If errors occur in the model, it is very time-consuming to detect or correct them during the user's ongoing process.A comparison and exchange of different but similar models for a technical system or asset is practically only possible during plant commissioning.

[0005] Models from engineering or development are used in the operational management of motors and inverters, for example, to perform analyses, comparisons, and predictions about the state of the (drive) system. Some models require continuous measurement signals with a high sampling rate (especially in the kHz range). Signals with varying sampling rates and minimum measurement durations may be needed to achieve a corresponding confidence interval for the models. Cloud-based measurement systems often have inherently low sampling rates (≤500 ms), which do not allow for "real-time analysis." Therefore, models with high sampling rate requirements for the measurement signal are currently only processed and calculated on system-level computers. This means that these models cannot be integrated into a cloud-based computing system.

[0006] US Patent 2002 / 0072828 A1 describes a method for modeling a non-linear empirical process. Based on an initial model, a non-linear network model with multiple inputs based on the initial model is to be constructed. The overall global behavior of the non-linear network model should generally conform to an initial output of the initial model. The non-linear network model is optimized based on empirical inputs, whereby the global behavior is constrained.

[0007] German patent DE 10 2006 054 425 A1 discloses a method for determining the value of a model parameter of a reference vehicle model. In this method, an estimated value of the model parameter is determined multiple times using an artificial neural network, depending on a second driving condition variable and / or a variable specified by the driver. The artificial neural network is adapted using a learning procedure.

[0008] A modular, cloud-based vehicle simulation with an open communication structure is known from an article by Lauff Ulrich et al. entitled "Development and Testing of Distributed Functions: Simulation and Virtualization of Vehicle Systems" in the journal Automobil Elektronik 09-10 / 2017 from September 1, 2017 (2017-09-01).

[0009] One object of the present invention is to improve, and in particular simplify, the modeling of a technical system. A further object is to improve the interchangeability and enhancement of models of a technical system during operation of the technical system without direct intervention. Another object is to enable the operationalization (integration and operation) of complex models in a cloud environment without requiring in-depth programming knowledge.

[0010] The problem is solved by a method according to claim 1.

[0011] The embodiments result from claims 2 to 6.

[0012] In an integrating model for a technical system, where the technical system is in particular a machine, a component of a machine, and / or a technical process and / or a technical plant, the integrating model comprises individual models, wherein the integrating model has a machine-to-machine interface, and wherein at least one of the individual models is interchangeable. In particular, each individual model is assigned a raw model and a data preprocessor, wherein different individual models model different parts of the technical system, and wherein individual models are interchangeable for a part of the technical system, and the interchangeable individual models have different raw models and different data preprocessors. Raw models are, for example, from different manufacturers and / or are programmed using different software (e.g., Matlab, Python, C, C++).

[0013] The raw models are programmed in different programming languages. Data preprocessing makes these raw models compatible with each other, enabling data exchange between them and further processing. Depending on the perspective, data preprocessing can also be considered post-processing of data with respect to an upstream raw model.

[0014] The differences in raw models necessitate data preparation to ensure data compatibility between models. These differences might include data protocols, sampling rates, number formats, etc. For example, a technical system could be a drive system, and its components could include a power converter, air cooling, water cooling, a bearing in an electric machine (motor or generator), a motor rotor, etc. Different raw models can exist for components of a technical system, each with its own advantages and disadvantages. For instance, there might be numerous raw models for an air-cooled motor or for a bearing supporting the rotor of an electric machine. Raw models also differ in their accuracy (e.g.,In the range of 0.1% to approximately 10%, a first sampling type 1 (data streaming blocks in the kHz range, calculation only 2x / 24), a second sampling type (calculation every minute with a singular value (no streaming)), a third sampling type (calculation 2x / 24h with a singular value (no streaming)), and / or a computation time (especially from a few seconds to several minutes). The computation times are based primarily on the time constants of the raw physical and mathematical models. For example, data streaming at 8 kHz over 15 seconds can be performed for an FFT analysis of a storage condition.

[0015] Raw models can be selected from a pool. Data processing is provided to enable the use of these raw models in the integrating model. Each raw model is assigned a specific data processing module, particularly a data preprocessing module. These modules, and thus the data preprocessing, can differ in the following ways: data streaming with block formation, data streaming without block formation, averaging for short periods (minutes, hours), averaging for long periods (several days, weeks, months, years), averaging for very long periods (several years), and / or correlational, particularly time-independent, preprocessing (e.g., ratio calculation).

[0016] For example, in an electric machine, the actual values ​​(measured values) of the excitation system can be averaged over monthly periods to obtain the long-term behavior of the target-actual difference of the excitation control. Here, the interaction between different models in the motor and inverter and their systematic deviations must be considered, excluding standstill and transient processes.

[0017] For a given topic, i.e., for a part of the technical system, there can be one or more raw models. Furthermore, various integrating models exhibit a specific set of raw models, which are particularly generic.

[0018] Generic models are, for example, for the following parts of a technical system: Coolers (for both inverters and motors; water- and air-cooled), insulation systems (various motor types, transformers / chokes in inverters; etc.), semiconductors (air- or water-cooled).

[0019] In this context, "generic" means that, for example, cooler properties can be used in the same way despite different geometries and media. Furthermore, lifetime analyses of insulation systems (choke, transformer, motor winding) or bearings are also generically applied and thus applicable to different asset types (e.g., bearings of turbo compressors, process coolers of customer systems, etc.).

[0020] A switchable linking element is provided between individual models. This allows individual models to be easily connected or disconnected. Individual models with the same task, i.e., the same modeling, can also be switched by removing one model from a signal flow and adding another.

[0021] In one embodiment of the integrating model, the technical system is at least part of a drive system. For example, a first integrating model, a first model system, concerns the overarching topic of water-cooled inverters, with models, i.e., individual models or raw models, for the following subtopics: a) The subject of the first individual model is a cooler, b) the subject of the second individual model is semiconductors, c) the subject of another individual model is a deionizer ... .

[0022] A second integrating model, i.e., a second model system, concerns, for example, the overarching topic of air-cooled inverters, whereby there are models, i.e., individual models or raw models, particularly for the following subtopics: a) The topic of the first individual model is semiconductors, b) The topic of the second individual model is environmental conditions, c) The topic of the further individual model is insulation system.

[0023] A third integrating model, or third model system, concerns, for example, the overarching topic of engine plain bearings, with models, i.e., individual models or raw models, for the following subtopics: a) The topic of the first individual model is balancing, b) The topic of the second individual model is bearings (specifically plain bearings), c) The topic of the third individual model is insulation system.

[0024] A fourth integrating model, or fourth model system, concerns, for example, the overarching topic of engine plain bearings, with models, i.e., individual models or raw models, for the following subtopics: a) The topic of the first individual model is bearings (rolling bearings specifically), b) The topic of the second individual model is coolers, c) The topic of the third individual model is insulation systems.

[0025] The individual models are primarily numerical models. Examples of a machine are an asynchronous machine and a synchronous machine.

[0026] Further examples of a machine include a pump for a liquid, a turbo compressor (for example, for technical gases), a fan (for example, for an induced draft fan), etc. Examples of a technical system include an elevator, a belt conveyor, a rolling mill, a refinery, etc. A component of a machine can also be understood as a technical system. Examples of machine components, especially of an electrodynamic machine, include a stator, a rotor, a bearing, and a cooling system. The technical process serves, for example, flue gas refraction, a drinking water supply, or the monitoring of a process. A component or a technical system can be modeled by a single model. Technical systems can interact and thus form another technical system (an extended technical system). The machine-to-machine interface is, for example, an interface of the following types: REST, SOAP, WSDL, or RPC.For example, REST stands for Representational State Transfer and has been used particularly in connection with the World Wide Web. Specifically, the interface (machine-to-machine interface) is a REST API (Representational State Transfer - Application Programming Interface). The interchangeability of the individual models within the integrating model allows for flexible transmission and modification over time. The integrating model, which encompasses the individual models within it, facilitates their coordination. This is particularly advantageous when the individual models differ in at least one of the following characteristics: different accuracy, different measurement depth, a different number of measured variables, or a different number of measurement cycles.

[0027] In one embodiment of the integrating model, it includes a program module. This program module is interchangeable. The integrating model can also include multiple program modules, many of which are interchangeable. The program module(s) relate, for example, to the calculation of characteristic curves and / or the simulation of physical data and / or the parameterization of at least one individual model. The program module can be integrated into an individual model or reside separately from the individual models within the integrating model, with data exchange between at least one individual model and the program module being provided.

[0028] In one embodiment of the integrating model, it comprises a multitude of individual models. Data is processed in a first individual model, and the resulting data from this model serves as input for a second individual model. For example, the resulting data from the second individual model can then be used as input for a third individual model. Data can be processed between the individual models (preprocessing and / or postprocessing) to adapt it to the respective model.

[0029] The integrating model is cloud-based. This means it is integrated into a cloud computing environment. Therefore, the integrating model can be implemented via an IT infrastructure that is, for example, made available via the internet. In another configuration, the integrating model is edge-based. In yet another configuration, the integrating model is implemented in a hybrid form. For example, at least one data-intensive model (processing more data compared to at least one other model) can be computed at the edge, while another, less data-intensive model can be computed in the cloud.

[0030] The integrating model features data preprocessing, particularly for communication between individual models. This preprocessing specifically serves to adapt different data interfaces. These interfaces can differ, for example, in sampling rate, numerical range, etc. This allows data to be processed by an individual model to be adapted to its input interface. Data preprocessing also includes, for example, preprocessing the data entering the integrating model. In one implementation of the integrating model, postprocessing is also included. Preprocessing and / or postprocessing involves, for example, adjusting data with respect to sampling rate, unit, jitter, etc.

[0031] In one embodiment of the integrating model, the integrating model features an external interface for communication between the individual models and the machine-to-machine interface. This external interface could, for example, be a REST API. Data preprocessing may also be provided for this purpose.

[0032] In one implementation of the integrating model, the machine-to-machine interface is a REST API interface. In this context, the machine-to-machine interface represents an external interface of the integrating model. "External" here means, in particular, that the data used originates from outside the integrating model and / or the data sent originates from within the integrating model and is transmitted outside of it.

[0033] In a method for providing an integrated model for a technical system, an integrating model is used, whereby different individual models are provided for modeling a part of the technical system. The technical system is, in particular, a drive, a machine, a machine component, and / or a technical process. The integrated model comprises a first individual model and a second individual model. Data from the first individual model is processed in a data preprocessing stage, and this processed data is used in the second individual model. For example, individual models with different data interfaces can be combined. The drive, in particular, comprises a power converter and / or a motor and / or a generator. The power converter is, for example, a rectifier, an inverter, or a frequency converter. The drive can be air-cooled and / or liquid-cooled.The motor or generator has bearings.

[0034] In one embodiment of the process, an individual model within the integrating model is replaced, thereby modifying the data preprocessing. For example, an individual model within the integrating model can be replaced by an improved or alternative individual model. This is particularly relevant when the interfaces differ.

[0035] Continuous measurement signals are recorded at a sampling rate in the kHz range and transmitted in blocks for use in the integrating model. This enables modeling even when real-time data transmission is not possible.

[0036] In one embodiment of the method, signals, i.e., data, with different sampling rates and / or minimum measurement durations are provided in a parameterizable block format in order to achieve a corresponding confidence interval for the individual models.

[0037] In one implementation of the method, the repetition rate of block-wise data transmission is varied. This allows, for example, adaptation to different bandwidths.

[0038] In one embodiment of the method, a sampling rate and / or a block size are changed, with the change occurring particularly depending on the individual model used. Parameters such as the sampling rate and / or the block size can be continuously changed during operation to adjust the sampling rate and the recording duration of the block sizes. The parameterization also includes, for example, event-based data management, which allows for more precise calculation of the models depending on, for example, the operating state of a (drive) system. In one embodiment of the method, a hybrid computer system is used for the integrating model.

[0039] In one implementation of the process, data from various assets (i.e., different technical systems, machines, and / or technical components or subsystems) are collected and stored. This data, combined with its integration into models—individual models within a larger, integrated model—allows for calculations to be performed in the cloud. It is possible to calculate simple and complex models (individual or integrated) of drive components such as motors, inverters, and other components in parallel with operation. These calculations allow conclusions to be drawn about the condition of the components (e.g., motors, inverters, etc.), and predictive information from the models can be used to determine maintenance or repair needs. Communication primarily takes place via a standardized API. The models can be deployed without requiring specific IT expertise.In particular, no customer-specific engineering is required, even for more complex systems. This allows for lower commissioning and / or operating costs compared to a purely plant-based solution. Maintenance and servicing, in particular, can be kept simple and are easily manageable, for example, throughout the entire plant lifecycle.

[0040] In one implementation of the process, parameterization, changes and / or adjustments to the models are carried out online in the cloud and are immediately available. This improves, for example, responsiveness to changes.

[0041] In one embodiment of the method, the models are operationalized. For example, an integrated model can model an entire powertrain, with individual models provided for the electrical components.

[0042] In one implementation of the process, individual models from standardized tools (e.g., Ansys, Varfem, Matlab, etc.) can be used. These individual models can be uploaded to the integrating model, with deployment including linking to the measurement data of the assets (property mapping). The modules and / or models are available primarily as graphical elements and can be coupled with one another. Furthermore, the measurement results from the models can be directly evaluated. These results generate notifications and values ​​that can be further processed in other applications, e.g., for KPIs or actionable recommendations.

[0043] Continuous measurement signals are recorded at a high sampling rate (in the kHz range) and transmitted in blocks.

[0044] In one embodiment of the method, signals with different sampling rates and minimum measurement durations are provided in configurable blocks to achieve a corresponding confidence interval for the models. Further parameters include, for example, the repetition rates of the block-wise transmission.

[0045] In one embodiment of the method, parameters are continuously changed during operation to adjust the sampling rate and the recording duration of the block sizes.

[0046] In one embodiment of the procedure, the parameterization includes event-based data management, which allows for more precise calculation of the models depending on the operating state of, for example, a (drive) system.

[0047] One potential advantage of running models in the cloud lies in their rapid maintenance and update capabilities. For models calculated close to the hardware, the speed of the calculations is the primary benefit. The block-wise and parameterizable provision of measurement data enables models with demanding measurement requirements to run with relatively high accuracy in a cloud environment. The effectiveness and operating costs of models in a cloud environment are significantly lower.

[0048] In one embodiment of the procedure, the integrative model also exhibits, for example, the following: a parameterizable data provision via a standardized interface (API) and / or scalability of data-driven numerical models without additional hardware (except Standard Connect modules) and / or no need for specific parameterization of models and / or simple remote maintenance and / or high accuracy in determining the condition of the drive system using adaptive measurement methods and / or parameterizability and provision of block-wise, sampling and length-variable measured values.

[0049] A solution to the problem can also be achieved through a computer program product that has computer-executable program means and, when executed on a computer system with processor means and data storage means, is used to carry out a procedure of the type described.

[0050] If a computer program product is provided that has computer-executable program means and, when executed on a computer system with processor and data storage means, is suitable for carrying out a procedure of one of the described types, then it can be executed, for example, in the cloud, on an edge device, and / or in a hybrid manner. In this way, an underlying task can be solved by a computer program product that is designed to simulate the operational behavior of the technical system.

[0051] The invention is illustrated and explained in more detail below with reference to the figures. The features shown in the figures can be combined to form further embodiments without departing from the invention. Reference numerals have the same meaning in the different figures. They schematically show: FIG 1 a use of the integrating model; FIG 2 a data stream; FIG 3 an embedding of the integrating model in the use of a technical system; FIG 4 a use of data packets in a real-time environment; FIG 5 an integration of the integrating model; and FIG 6 a model system with interconnected individual models.

[0052] The representation according Figure 1 This demonstrates the use of the integrating model 1. This also applies in particular to the transfer of a model (e.g. from development) into a further application. Figure 1This section shows three areas: the first area 17, the second area 18, and the third area 19. The first area 17 concerns the provision of raw models 7 and 8, or the provision of program module 6. Raw models 7 and 8 can be programmed on different systems such as MATLAB or Python. The first area 17 specifically concerns models 7 and 8 and program module 6, which are used in research and development (R&D) and engineering. Program modules and / or raw models are provided, for example, via the internet and / or storage media such as flash memory, CD-ROMs, or similar devices.

[0053] The second area 18 concerns parameterization, use, and operation. Raw models as well as program modules from the first area 17 are thus transferred to the second area 18. The second area 18 contains an integrating model 1, which resides in a cloud 22, interacts with it, and / or interacts from it. The integrating model 1 contains numerical models, or is, in particular, a numerical model. Through human intervention 21, the integrating model 1 can be monitored, parameterized, programmed, and / or maintained. The second area 1 18 specifically concerns the integration of models into an application, particularly an application within a platform. Model parameters are accessible to a user 21. User access, i.e., human intervention, can occur locally and / or remotely.

[0054] The third area 19 concerns monitoring and / or prediction functions. Data from the second area 18 is therefore used in the third area 19. The third area 19 thus includes, for example, the prediction of a curve trajectory 15, a technical or monetary evaluation 16, and / or the analysis of a technical system 14, which, for instance, comprises a first machine 12 and a second machine 13. The third area 19 also specifically addresses the operation, evaluation, and continuous improvement of models. For example, a digital twin can be created. It is possible, for instance, to determine whether a model is functioning correctly. Furthermore, it is possible, for example, to propose parameter adjustments based on monitoring.The results of monitoring and / or prediction can be used, in particular, for the design of real or modeled technical systems (for example, a machine or a technical plant). Monitoring makes it possible, in particular, to calculate predictions of errors or alarm states. The models, i.e., the individual models or raw models, relate to, for example, a: . Cooler model for cooling an electric machine, Arrhenius model (describes the release of molecules from solids - used for insulation monitoring and / or aging of materials), rotor model of an electric machine, bearing model for moving parts of an electric machine.

[0055] Parameterization can be performed internally by a user or externally via an interface, such as an API. Parameterization affects, for example, the physical properties of at least one asset, i.e., a component. This component could be, for example, a bearing, a rotor, a power converter, a residual current monitor, a machine, etc. The API can, for instance, access a database and retrieve the parameters for the asset from there. Machine data, for example, might contain information about the types of bearings used. Bearing types can then be used to query data such as whether the bearing is a ball bearing or a needle bearing, and how many balls or needles it has.

[0056] The representation according Figure 2This shows a visualization of data flows, as well as a visualization of key figures, recommendations, and / or the linking of properties. The integrating model 1 is shown. This model comprises a multitude of raw models 7, 8, 9, 10, and 11. These raw models 7, 8, 9, and 10 are integrated into individual models 2, 3, 4, and 5. The individual models 2, 3, 4, and 5 have modules for data processing 25, 26, 27, and 28. The individual models 2, 3, 4, and 5 are designed to allow data exchange 53 between them. The individual models 2, 3, 4, and 5 can be interconnected serially. The interconnection of the individual models 2, 3, 4, and 5, i.e., their data connection, takes place in the model interconnection 100. A prediction device 15 can, for example, record or continue a time series. A parameterization device can also be provided. The integrating model 1 has external interfaces 23 and 24.External interfaces 23 and 24 are, for example, of the REST API type. Thus, data from a data input 51 can be routed via input interface 23 to individual models 2, 3, 4, and 5. The data can then be fed to an evaluation module 16 via output interface 24. Within this evaluation module, a drive system, for example, can be analyzed. Furthermore, notifications, recommendations, emails, and / or generated key performance indicators (KPIs) can be displayed alphanumerically or graphically.

[0057] The representation according Figure 3Figure 1 shows the embedding of the integrating model in the use of a technical system. An electric machine 12, such as an electric motor, has at least one sensor and a data connection. The electric machine 12 sends data 54 to the cloud 22. There, the data is initially stored, for example. Afterwards, the data 55 is offered via a data interface 29 of the type REST API. The data can be requested via an external interface 23 of the type REST API of the integrating model 1. The integrating model 1 can use the data to perform modeling and, for example, calculate an approximation of whether a shutdown is necessary. The integrating model 1 can forward the calculated data via an external interface 24. This also occurs, for example, via a REST API interface. This exported data can also be stored in the cloud and / or locally. The data is then used for a purpose 19.This is, for example, an analysis device for a drive system. Functions 31, 32 and 33, such as alarms, warnings, predictions, timelines, reports, recommendations, etc., can be created and displayed.

[0058] The representation according Figure 4This illustrates the use of data packets in a real-time environment within the model. Sensor data 36, ​​37 are generated by machine 12. Sampled sensor data is transferred to the cloud and stored there as data blocks 34, 35. These data blocks are offered for querying via the REST API data interface 29. The integrating model 1 queries data blocks 34, 35 via a query cycle 56. This provides the model with sufficient data to start a simulation. The data blocks are primarily BLOBs (Binary Large Objects). Data exchange is configurable via properties / measurement names. The measured values ​​can be, for example, high-resolution (data BLOBs), time series, or individual state bits, depending on the application. Of course, combinations of measured values ​​are also possible, such as exchanging individual bits and high-resolution data, for example, to initiate an event-based calculation.

[0059] The representation according Figure 5Figure 1 shows an integration of the integrating model 1. In a plant / factory 41, there is a power converter 39 and a motor 40. The power converter 39 has a data box 48. The motor 40 has a data box 49. The data box 48 is connected to a bus system 43. A controller 46 and a protection system 47 are also connected to this bus 43. The data box 48 is connected to a connection module 52 via a data connection 44 (e.g., via fiber optic cable). The connection module 52 is used for processing, collecting, and forwarding data. The motor 40 has sensors 42. Data from the sensors 42 can be sent to the data box 49. The data box 49 sends the data via a data connection 45 (e.g., via a fiber optic connection) to the connection module 52. The connection module 52 sends the data to the cloud 22. The integrating model 1 is located in the cloud 22, or the integrating model 1 is reachable via the cloud 22.The integrating model 1 comprises various individual models 2, 3, ..., 6, which can be secured against unauthorized access. A user interface 50 is provided for one user to operate the integrating model 1 or to query data from the integrating model 1.

[0060] The representation according Figure 6Figure 1 shows a model interconnection 100. To generate the model interconnection 100, the desired raw models are selected from a pool 111 of raw models 7, 8', 8", 8', 9', 9" and integrated into the model interconnection 100 together with associated data preprocessing operations 25, 26', 26", 26', 27', 27" from another pool 110. Each group of raw models relates to a specific topic to be modeled (for example: rotor temperature, bearing current, load torque, etc.). A first topic, 101, is assigned raw model 7 and data preprocessor 25. For a second topic, 102, for example, three raw models 8', 8", 8‴ and three data preprocessors 26', 26", 26‴ are available, of which only two are selected and integrated for the model interconnection. For a third topic, 103, for example, two raw models 9', 9" and two data preprocessors 27', 27" are available, of which only all are selected and integrated for the model interconnection 100.Within the model interconnection 100, the individual models 2, 3', 3", 4', 4" are interconnected data-wise, using switches 104 to 109 (switchable linking elements). This allows individual models to be quickly connected or disconnected without having to add further raw models to the interconnection. In this way, for example, a comparison of models with different raw models can be easily generated.

Claims

1. Method for providing an integrating model (1) for a technical system (14), wherein the integrating model (1) has individual models (2, 3, 4, 5, 6), wherein the integrating model (1) has a machine-machine interface (23), wherein at least one of the individual models (2, 3, 4, 5, 6) is interchangeable, wherein an individual model (2, 3, 4, 5, 6) is respectively assigned a raw model (7, 8, 9, 10) and data preprocessing (25, 26, 27, 28), wherein different individual models (2, 3, 4, 5, 6) model different parts of the technical system (14), wherein individual models (3, 3', 4, 4', 4") are interchangeable for a part of the technical system (14), wherein the interchangeable individual models (3, 3', 4, 4', 4") have different raw models (7, 8, 9, 10) and different data preprocessing (25, 26, 27, 28), wherein the raw models (7, 8, 9, 10) are programmed in different programming languages, wherein a switchable combinational logic element (104, 105, 106, 107, 108, 109) is provided between individual models (2, 3, 3', 4, 4', 4", 5, 6), wherein the integrating model (1) is cloud-based, characterized in that the technical system (14) is an electrical machine, namely a converter, a motor and / or a generator, and continuous measurement signals from a sensor of the electrical machine are recorded at a sampling rate in the kHz range and are transmitted block-by-block for use in the integrating model, thus allowing modelling even when real-time transmission of the data is not possible.

2. Method according to Claim 1, wherein an individual model (2, 3, 4, 5, 6) of the integrating model (1) is interchanged, wherein the data preprocessing (25) is changed.

3. Method according to Claim 1, wherein the sampling rate and / or the block size is / are changed, wherein the change takes place in particular on the basis of an individual model (2, 3, 4, 5, 6) used.

4. Method according to one of Claims 1 to 3, wherein an individual model (2, 3, 4, 5, 6) is activated or deactivated by means of a switch (104-109).

5. Method according to one of Claims 1 to 4, wherein the data preprocessing (25, 26, 27, 28) is provided for communication between individual models (2, 3, 4, 5, 6).

6. Method according to one of Claims 1 to 5, wherein the machine-machine interface (23) is a REST API interface.

7. Computer program product which comprises computer-executable program means and, when executed on a computer device with processor means and data storage means, is configured to carry out the method according to one of Claims 1 to 6.