DATA-REDUCED EDGE-TO-CLOUD TRANSMISSION BASED ON PREDICTION MODELS

Abstract

A method for providing process data of a device in an industrial automation environment to a computer system. In one embodiment, the method includes the following steps: executing a process data model on the device for generating estimated process data; determining that the estimated process data deviates from the real process data by more than a threshold value; and only if the estimated process data deviates from the real process data by more than the threshold value: transmitting information representing the real process data from the device to the computer system.

Claims

1. A method for providing process data of a device in an industrial automation environment to a computer system, the method comprising: executing a process data model on the device to generate estimated process data, the computer system also executing the process data model to generate corresponding estimated process data; determining whether the estimated process data deviates from real process data by more than a threshold value; and transmitting, only if the estimated process data deviates from the real process data by more than the threshold value, information representing the real process data from the device to the computer system.

2. The method according to claim 1, wherein the information representing the real process data comprises the real process data.

3. The method according to claim 1, wherein the information representing the real process data comprises reconstruction data, the reconstruction data allowing the computer system to reconstruct the real process data from the estimated process data.

4. The method according to claim 3, wherein the reconstruction data comprise difference values.

5. The method according to claim 1, wherein the determination is carried out on the device.

6. The method according to claim 1, further comprising: generating the process data model via a machine learning system based on real process data, the machine learning system being arranged in the computer system or in a second computer system.

7. The method according to claim 6, further comprising: transmitting the process data model to the device and, if the process data model was generated on the second computer system, transmitting the process data model to the computer system.

8. The method according to claim 1, further comprising: retraining, if the estimated process data deviates from the real process data by more than the threshold value, the process data model and updating the process data model on the device and on the computer system.

9. The method according to claim 1, wherein the device comprises an edge device, a field device, a control device and/or a programmable logic controller; and/or wherein the computer system comprises a cloud system and/or a server; and/or wherein the second computer system comprises a cloud system and/or a server; and/or wherein the computer system and/or the second computer system is or are located at a distance from the device.

10. A computer program including instructions to implement the method according to claim 1.

11. A computer readable medium device containing program instructions to implement the method according to claim 1.

12. A computer system adapted to perform the method according to claim 1.

13. A system comprising a device and a computer system, wherein the system is configured to carry out the method claim 1.

Description

BRIEF DESCRIPTION OF THE DRAWINGS

[0022] The present invention will become more fully understood from the detailed description given hereinbelow and the accompanying drawings which are given by way of illustration only, and thus, are not limitive of the present invention, and wherein:

[0023] FIGS. 1a to 1c show an example of the invention, in which the machine learning system is integrated into the remote computer system;

[0024] FIGS. 2a to 2c show an example including an external machine learning system; and

[0025] FIG. 3 shows a flowchart of a method according to an example of the invention.

DETAILED DESCRIPTION

[0026] The currently preferred exemplary embodiments of a method according to the invention for the reduced data transmission between a device and a computer system are explained in greater detail below with reference to FIG. 3.

[0027] In a preparation phase, which is illustrated in FIG. 1a, real process data D100, which are generated, for example, in the context of an automated system, are first collected and, on this basis, a prediction model 400 (also referred to as a process data model) is generated (cf. step S5 in FIG. 3), which is then able to predict the (future) process data. Process data model 400 is preferably a (mathematical or statistical) model for simulating real process data. Prediction model 400 is preferably generated within a cloud system 200, based on a machine learning system 300. Cloud systems are particularly suitable for this purpose, since they may provide the necessary IT resources as needed.

[0028] Machine learning system 300 may be based on different algorithms, each of which is used depending on the problem. In one preferred implementation, a structurally predefined statistical model is trained on the basis of input/output data pairs. A portion of the data points iteratively runs through the model. The learning algorithm may adapt the model step by step, based on the deviation between the calculated results and the expected results. Suitable techniques from machine learning and artificial intelligence may be used for this purpose. One objective of a monitored learning of this type is that the network is trained to be able to establish associations after multiple computing passes using different inputs and outputs.

[0029] Prediction model 400 is subsequently distributed to edge device 100 in step S10, as illustrated in FIG. 1b.

[0030] At runtime (cf. FIG. 1c), process data model 400 is preferably executed in step S15 in parallel in both cloud system 200 and on edge device 100, and prediction data D200 are generated for the corresponding process data. Data D200 are thus estimated process data.

[0031] In step S20, a (continuous) comparison between actual process data D100 of the real process and estimated process data D200, which prediction model 400 generates, additionally takes place on edge device 100.

[0032] If a deviation between real process data D100 and process data D200 of prediction model 400 is determined on edge device 100 (it being possible to set a certain deviation tolerance here using a threshold value), this means that predicted process data D200 are no longer correct within cloud system 200. To have correct process data present within cloud system 200 at any time, either real process data D100 must be transmitted to cloud system 200 in a case of this type (cf. step S25), or so-called reconstruction data D300 must be sent, which are used to reconstruct the process data from prediction model 400 (e.g., based on differences), so that these data subsequently correspond to real process data D100.

[0033] It is understood that the preparation phase illustrated in FIGS. 1a and 1b (cf. steps S5 and S10 in FIG. 3) is functionally independent of the runtime phase illustrated in FIG. 1c (cf. steps S15 through S25 in FIG. 3), i.e., process data model 400 may also originate in another source or have been generated in a different way.

[0034] Prediction model 400 may possibly also be retrained in step S30 and be updated on edge device 100 as well as in cloud system 200 in step S35 for the purpose of keeping its prediction accuracy as high as possible, which may decrease due to seasonalities or changed environmental conditions.

[0035] As illustrated in FIGS. 2a through c, the operation of machine learning system 300 and the generation of prediction model 400 may also take place on a separate (cloud) system 500. Cloud system 500 may be used from the outside as well as be operated internally. A cloud system 200 or 500 may possibly also be represented by a conventional server.

[0036] The invention being thus described, it will be obvious that the same may be varied in many ways. Such variations are not to be regarded as a departure from the spirit and scope of the invention, and all such modifications as would be obvious to one skilled in the art are to be included within the scope of the following claims.