METHOD AND SYSTEM FOR OPERATING A TECHNICAL DEVICE WITH A MODEL BASED ON ARTIFICIAL INTELLIGENCE

Description

[0001] Regardless of the grammatical gender of a particular term, persons with male, female or other gender identities are included.

[0002] The invention relates to a computer-implemented method and a system for operating a technical device with a model based on artificial intelligence, as well as a computer program product.

[0003] Devices in industrial environments are frequently operated using artificial intelligence (AI) models to perform predictive maintenance, for example. This is achieved by using a model to predict future operation based on current operating parameters, such as vibration data, acoustic data, current or voltage data, or similar information.

[0004] A model requires extensive training and is then often only valid for one specific device installation. Therefore, if an identical device is installed or arranged differently, the model must be extensively retrained according to current best practices.

[0005] It is currently very difficult to transfer an AI model trained on a device, such as a pump, located in one environment to identical or similar devices in another environment.

[0006] The environment, or operating environment, refers, for example, to the installation condition, screw connections, soil composition, or adjacent equipment. The environment can influence the data and often leads to undesirable variations in the measured values.

[0007] Publication EP 3 751 146 A1 relates to a method for training a machine learning algorithm which, by determining the damage to a first compressor of a gas turbine, enables the determination of the functionality of a second compressor of the gas turbine.

[0008] Publication EP 3 098 681 A1 describes an anomaly detection system for artificial intelligence-based health management of a host system, in which unique patterns generated in real time are monitored and system parameters indicate the performance of the host system at the system level, and the patterns are compared with predefined patterns corresponding to the system parameters in order to detect potential anomalies in the host system and its subsystems.

[0009] The object of the invention is to provide a solution in which a model for a device can be reused, i.e., independently of the operating environment, for example, independent of the installation.

[0010] The task is solved by a computer-implemented procedure for operating a technical device with a model based on artificial intelligence, comprising the following steps: a) Acquiring a first operating parameter of a reference device, which is identical in construction to the technical device, using a first measuring instrument in the form of a time-discrete first sensor signal in a first operating environment and a first permissible operating mode; b) Acquiring a second operating parameter of the reference device using a second measuring instrument in the form of a time-discrete second sensor signal in a second operating environment and a second permissible operating mode; c) Performing a first Fourier transform for the first sensor signal and a second Fourier transform for the second sensor signal using a computing device; d) Determining the quotient of the first Fourier transform and the second Fourier transform as a transfer function using the computing device.e) The computing device generates and trains a first artificial intelligence-based model for the reference device in the first operating environment; f) the computing device generates a second artificial intelligence-based model for the technical device in the second operating environment using the first model and applying the transfer function; g) the computing device operates the technical device with the second model.

[0011] This makes it possible to train an AI model, for example in a laboratory environment, and then transfer this AI model to a new installation with high accuracy of the model, without having to adapt or retrain the model (Plug & Classify).

[0012] This enables the scaling of AI models in industrial applications by allowing models to be reused without additional effort.

[0013] Although the procedure for operating a technical device with a model based on artificial intelligence is described in the present context, it should be noted that a statistical model is equally suitable for use in the aforementioned procedure and therefore represents an equivalent use.

[0014] It is understood that the described methodology can be transferred from a single reference pump to any number of identical pumps in different operating environments.

[0015] The term "structurally identical technical device" encompasses all equivalent embodiments that are essentially identical or similar in function and actual design, i.e., for example, that have the same technical object and operating characteristics (such as power consumption, performance, weight, volume, etc.) of a pump.

[0016] Operating environments can, for example, depict the assembly of the technical device, as well as cable routing or influences from neighboring devices.

[0017] The measuring means can be formed by measuring devices with appropriate sensors.

[0018] The training of the reference device can include all operating cases, both permissible operating modes and impermissible operating modes, which may, for example, cause an operating anomaly.

[0019] The Fourier transform can create a filter function based on average sensor signals. While averaging is not strictly necessary, it further improves the accuracy of the resulting model.

[0020] In a further development of the invention, it is provided that the first permissible operating mode corresponds to the second permissible operating mode.

[0021] This allows the accuracy of the second model to be further improved.

[0022] In a further development of the invention, it is provided that the application of the transfer function includes a multiplication in the frequency domain.

[0023] This allows for a simple improvement in the calculation, as significantly fewer multiplication operations are required compared to multiplication in the time domain. This operation is equivalent to convolution in the time domain.

[0024] The problem according to the invention is also solved by a system for operating a technical device with a model based on artificial intelligence, comprising a reference device which is identical in construction to the technical device, a first measuring means for recording a first operating parameter of the reference device in a first operating environment and a first permissible operating mode, a second measuring means for recording a second operating parameter of the reference device in a second operating environment and a second permissible operating mode, a computing device with a processor and a memory, wherein the computing device is also set up to operate the technical device, and the system is configured to perform the method according to one of the preceding claims.

[0025] The problem according to the invention is also solved by a computer program product with machine-readable instructions stored therein which, when executed by a processing unit, cause it to execute the method according to the invention.

[0026] The invention is explained in more detail below with reference to an embodiment illustrated in the accompanying drawings. The drawings show: Fig. 1 a schematic representation of a system for operating a technical device with a model based on artificial intelligence with the respective input variables, Fig. 2 a flowchart of the method according to the invention, Fig. 3 a simplified flowchart of the method according to the invention.

[0027] Fig. 1 shows a schematic representation of a system for operating a technical device with a model based on artificial intelligence and the respective input variables.

[0028] In this embodiment, three systems S1-S3 are represented by corresponding AI models.

[0029] The systems each comprise an identical pump P, which is represented by a pump model p(t), and a separate operating environment E1-E3, which are represented by separate operating environment models e 1 (t), e 2 (t), e 3 (t).

[0030] The same input signals xh< (t) are fed to the systems S1-S3 and individual time-continuous output signals y 1 h< (t), y 2 h< (t), y 3 h< (t) are formed by the systems S1-S3.

[0031] Time-discrete output signals y 1 h< [k], y 2 h< [k], y 3 h< [k] are derived from the time-continuous output signals y 1 h< (t), y 2 h< (t), y 3 h< [k] by sampling units S.

[0032] The condition for these procedural steps is that for systems S1-S3, in particular for the operating environments L a of the pumps P, elements of the time-discrete output signal y 1 [n], y 2 [n], y 3 [n] are each captured in a permissible operating environment lh, i.e., they do not represent an operating anomaly.

[0033] A permissible operating environment includes, for example, a permissible operating temperature range, a permissible operating humidity range, a permissible speed range or torque range of a pump, and a permissible current or power consumption.

[0034] Operating parameters include, for example, a temperature value, a current, a voltage, a vibration characteristic, an acoustic characteristic or the like for a technical device, which are determined with appropriate sensors or using indirect methods, such as imaging techniques.

[0035] Fig. 2 Figure 1 represents an example of a flowchart of the method according to the invention, wherein the explanations of the preceding figure apply accordingly.

[0036] In a first operating environment E1, time-discrete first sensor data SD1 is acquired from a first pump P1, stored as a time-discrete first sensor signal y 1 [k] with k elements and as a Fourier-transformed first sensor signal Y 1 [n] calculated by a first Fourier transform FFT1, in particular by a Fast Fourier transform.

[0037] In the present context, instead of a Fast Fourier Transform, for example a z-Transformation, a Discrete Fourier Transform or a Discrete Cosine Transform can also be used.

[0038] The first pump P1 forms a reference pump, based on which a very accurate operational AI model can be calculated for further pumps, without having to retrain a model for these further pumps themselves.

[0039] In a first operating environment E1, a second pump P2, which is identical in construction to the first pump P1, acquires discrete-time second sensor data SD2, as a discrete-time second sensor signal y 2 [k] with k elements and stores the spectrum calculated by a second Fourier transform FFT2, in particular by a Fast Fourier transform, as a Fourier-transformed second sensor signal Y 2 [n].

[0040] A computing device with a processor and a memory now forms the quotient of the Fourier-transformed first sensor signal Y 1 [n] and the Fourier-transformed second sensor signal Y 2 [n] and stores it as a transfer function F[n] with n elements.

[0041] If an operational AI model is created for the first pump P1 and trained in a training process T, especially with permissible and impermissible operating environments in which anomalies can occur, this trained model can be transferred to other pumps, such as the second pump P2 in this example, without having to retrain the model.

[0042] Now, an AI model can be determined for the second pump P2 using the transfer function F[n], without having to train an AI model for the second pump P2, and can be used for an accurate prediction P of the operating state of the second pump P2.

[0043] The figure also represents a corresponding system for operating a technical device P2 with a model based on artificial intelligence.

[0044] It includes a reference device in the form of pump P1, which is identical in construction to the technical device in the form of pump P2. Furthermore, it includes a first measuring device MD1 for recording a first operating parameter of the reference device in a first operating environment and a first permissible operating mode.

[0045] Furthermore, a second measuring instrument MD2 is provided for recording a second operating parameter of the reference device in a second operating environment and a second permissible operating mode.

[0046] The system also includes a computing device PD with a processor and memory, which is configured to operate the technical equipment. It is clear that the computing device PD can be designed as a distributed system.

[0047] Fig. 3 shows an example of a simplified flowchart of the method according to the invention.

[0048] The computer-implemented procedure for operating a technical device with an artificial intelligence-based model includes the following steps: a) Acquiring a first operating parameter of a reference device P1, which is identical in construction to the technical device P2, using a first measuring instrument MD1 in the form of a time-discrete first sensor signal in a first operating environment E1 and a first permissible operating mode; b) Acquiring a second operating parameter of the reference device P1 using a second measuring instrument MD2 in the form of a time-discrete second sensor signal in a second operating environment E2 and a second permissible operating mode; c) Performing a first Fourier transform FFT1 for the first sensor signal and a second Fourier transform FFT1 for the second sensor signal using a computing device PD; d) Determining the quotient of the first Fourier transform and the second Fourier transform as a transfer function F using the computing device PD.e) Generating and training a first artificial intelligence-based model for the reference device P1 in the first operating environment E1 using the computing device PD, f) Generating a second artificial intelligence-based model for the technical device P2 in the second operating environment E2 using the first model and applying the transfer function F using the computing device PD, g) Operating the technical device P2 with the second model using the computing device PD.

[0049] In this example, the first permissible operating mode corresponds to the second permissible operating mode in order to obtain a particularly high accuracy of the pump model.

[0050] In a preferred embodiment, the application of the transfer function includes multiplication in the frequency domain, thereby reducing the number of multiplications required by a computing device. Reference symbol list:

[0051] e1(t), e2(t), e3(t) Model for the operating environment of pump E1, E2 Operating environment F[n] Transfer function with n elements FFT1, FFT2 Fast Fourier transform L a Operating environment of pump lh Permissible operating environment for pump MD1, MD2 Measuring means P Prediction p(t) Model for pump P1, P2 Pump P Computing device with processor and memory S Sampling S1-S3 System SD1, SD2 Sensor data T Training of the AI ​​model xh<(t) Input signal y1[k], y2[k] Time-discrete sensor signal with k elements Y1[n], Y2[n] Transformed sensor signal in the frequency domain with n elements y1[n], y2[n], y3[n] Elements of the time-discrete output signal y1h<(t), y2h< (t), y 3 h< (t) Output signal, continuous-time y 1 h< [k], y 2 h< [k], y 3 h< [k] Output signal, discrete-time

Claims

1. Computer-implemented method for operating a technical device (P2) with a model based on artificial intelligence, comprising the following steps: a) Detecting a first operating parameter of a reference apparatus (P1), which is constructed identically to the technical device (P2), with the aid of a first measuring means (MD1) in the form of a time-discrete first sensor signal in a first operating environment (E1) and a first permissible operating mode, b) Detecting a second operating parameter of the reference apparatus (P1) with the aid of a second measuring means (MD2) in the form of a time-discrete second sensor signal in a second operating environment (E2) and a second permissible operating mode, c) Performing a first Fourier transform (FFT1) for the first sensor signal and a second Fourier transform (FFT2) for the second sensor signal by means of a computing apparatus (PD), d) Determining the quotient from the first Fourier transform and the second Fourier transform as a transfer function (F) by means of the computing apparatus (PD), e) Generating and training a first model on the basis of artificial intelligence for the reference apparatus in the first operating environment (E1) with permissible and impermissible operating environments by means of the computing apparatus (PD), f) Generating a second model on the basis of artificial intelligence for the technical device (P2) in the second operating environment (E2) using the first model and applying the transfer function (F) by means of the computing apparatus (PD), g) Operating the technical device (P2) with the second model by means of the computing apparatus (PD), wherein the respective operating parameters in steps a) and b) are detected in a permissible operating environment in each case and the respective operating environments (E1, E2) are mapped by their own operating environment models in each case.

2. Method according to the preceding claim, wherein the first permissible operating mode corresponds to the second permissible operating mode.

3. Method according to one of the preceding claims, wherein the application of the transfer function comprises a multiplication in the frequency range.

4. System for operating a technical device with a model based on artificial intelligence, comprising • a reference apparatus, which is constructed identically to the technical device, • a first measuring means for detecting a first operating parameter of the reference apparatus in a first operating environment and a first permissible operating mode, • a second measuring means for detecting a second operating parameter of the reference apparatus in a second operating environment and a second permissible operating mode, • a computing apparatus with a processor and a memory, wherein the computing apparatus is also designed to operate the technical device and the system is configured to implement the method according to one of the preceding claims.

5. Computer program product having machine-readable instructions stored therein, which, when executed by a processing unit, trigger this to execute the method according to one of claims 1 to 3.