Procedure and system for evaluating a technical plant

Abstract

A method for evaluating a technical system is proposed. During an operational phase, several operational data sets are recorded by a computer. Each set comprises a plurality of measured variables, with each measured variable containing a plurality of measured values. The measured values ​​of each measured variable are recorded sequentially. The recorded operational data sets are checked on the computer by an anomaly model. Aggregated values ​​are calculated from the measured values ​​of operational data sets identified as normal by the anomaly model, and these aggregated values ​​are stored. The measured values ​​of the operational data sets identified as normal by the anomaly model are discarded. A system according to the invention for evaluating a technical system comprises a computer. The computer includes an anomaly model for checking recorded operational data sets.The system according to the invention is set up to carry out the method according to the invention.

Description

[0001] The invention relates to a method for evaluating a technical system, wherein several operational data sets are recorded by a computer during an operational phase, and the recorded operational data sets are checked by an anomaly model. The invention also relates to a system for evaluating a technical system, which is configured to carry out the method according to the invention.

[0002] Technical systems, such as drive systems, comprise a multitude of technical components, including an electric motor, a converter to generate three-phase alternating current for the electric motor, and a gearbox to reduce the motor's speed. Examples of such technical systems include rotary tables, conveyor belts, and stacker cranes. After prolonged operation, the components of these systems can malfunction due to wear and tear.

[0003] It is known to monitor such technical systems by recording and evaluating specific measurements at set times. If the recorded measurements deviate too significantly from predefined target values, a fault in the system is assumed, and a corresponding message is sent to the operator. For example, a current generated by the inverter that exceeds a defined limit could indicate stiffness in the gearbox due to wear.

[0004] Condition monitoring involves collecting a relatively large amount of data from a technical system for later analysis. This data consumes storage space and incurs costs. A proven strategy to reduce costs is to aggregate the raw data directly and store only the aggregated data. However, in certain cases, historical raw data is also relevant, for example, for training machine learning models or for visualizing the raw data to identify anomalies in the event of a fault.

[0005] From US patent 20170024649 A1, a method for anomaly detection is known in which measurement data is received, features learned from a learning platform are extracted, and the learned features are fed to a classifier.

[0006] From EP 2 477 086 A1 a method and a system for anomaly detection in a technical plant are known, wherein measurement data are recorded and it is detected whether an anomaly is present.

[0007] From EP 4 443 258 A1, a method for fault detection in technical systems is known, whereby faults are detected and classified using methods of artificial intelligence based on operational data of the technical system.

[0008] WO 2024 / 002725 A1 discloses a monitoring device for the condition monitoring of a machine, which includes a machine learning unit. The learning unit receives sensor data collected from the machine during operation and determines an anomaly result and a feature explaining the anomaly result from the sensor data.

[0009] The invention is based on the objective of further developing a method and a system for evaluating a technical plant.

[0010] The problem is solved by a method for evaluating a technical system with the features specified in claim 1. Advantageous embodiments and further developments are the subject of the dependent claims. The problem is also solved by a system for evaluating a technical system with the features specified in claim 9.

[0011] A method for evaluating a technical installation is proposed. The technical installation comprises an application with a drive system, and the drive system comprises an electric motor, a converter for generating a three-phase alternating voltage for the electric motor, and a gearbox for reducing the speed of the electric motor.

[0012] During an operational phase, a computer records multiple operational data sets, each comprising a plurality of measured variables. Each measured variable contains multiple measured values, and these values ​​are recorded sequentially. The recorded operational data sets are then analyzed by an anomaly model on the computer. Aggregated values ​​are calculated from the measured values ​​of operational data sets identified as normal by the anomaly model and stored. The measured values ​​of the operational data sets identified as normal by the anomaly model are discarded.

[0013] The recorded operating data sets each include the rotational speed and torque of the electric motor as measured variables. These variables are measured indirectly via the inverter by measuring the frequency and current of an output current to the electric motor.

[0014] The computer in question is a digital computer located within the technical system, in the field, and close to the application. Among other things, the computer is used to acquire and process measured values. Such a computer is often also referred to as an edge device. For example, the acquired operational data set comprises eight different measured variables, and each measured variable contains 2048 data points. The measured values ​​of each variable are acquired at equidistant intervals of 5 ms. Examples of measured variables include DC link voltage, current, frequency, rotational speed, and torque. The anomaly model was previously trained using reference data sets.

[0015] The method according to the invention is suitable for persisting relevant data for the evaluation of a technical system. Operating data sets and their measured values, recognized as normal, describe regular operating conditions of the technical system and are therefore redundant and thus less relevant. The method according to the invention allows redundant or less relevant data to be discarded. This results in a significant reduction in the amount of data to be stored, without any substantial loss of information. This saves storage space when persisting the data, thereby also reducing costs.

[0016] According to the invention, during a reference phase, several reference data sets are acquired by the computer, each comprising a plurality of measured variables, wherein each measured variable comprises a plurality of measured values, and wherein the measured values ​​of each measured variable are acquired sequentially. The anomaly model on the computer is trained with the reference data sets in order to verify subsequently acquired operational data sets.

[0017] For example, each recorded reference dataset comprises eight different measured variables, and each measured variable comprises 2048 measured values. For example, the measured values ​​of each measured variable are recorded at equidistant time intervals of 5 ms each. Measured variables include, for example, DC link voltage, current, frequency, speed, and torque. The recorded reference dataset can be assigned to exactly one application cycle.

[0018] According to the invention, operational data sets that are identified as anomalous by the anomaly model are treated as further reference data sets. The anomaly model is then trained on the computer using these additional reference data sets.

[0019] Operational data records identified as anomalous may originate from an application cycle for which no reference data records have yet been recorded. Therefore, the additional application cycle is added to the anomaly model.

[0020] According to the invention, the anomaly model is implemented as a single-class classifier or as a deep neural network in autoencoder architecture or as combinations thereof.

[0021] A one-class classifier is, for example, implemented as a one-class SVM. Possible combinations include, for example, feature reduction by an autoencoder and classification of the one-class classifier based on the features reduced by the autoencoder.

[0022] According to an advantageous embodiment of the invention, a further operating data record is recorded each time a defined trigger condition is met and when a defined period of time has elapsed since the recording of the previous operating data record.

[0023] One such trigger condition is, for example, the start of an application cycle of the technical system when the speed of an electric motor in the system exceeds a predefined minimum speed. This ensures that the recorded operating data record originates from an application cycle.

[0024] According to an advantageous embodiment of the invention, the aggregated values ​​are sent to a server and stored on the server.

[0025] The server is a remote digital computer. It typically has more storage and processing resources than the local computer. The server is located, for example, in a local data center or in the cloud. It is connected to the local computer via a digital network, such as a LAN, WLAN, or the internet. This saves processing power and storage space on the local computer.

[0026] According to an advantageous embodiment of the invention, the measured values ​​of operational data sets, which are recognized as abnormal by the anomaly model, are sent to the server and stored on the server.

[0027] Operational data records and their measured values, identified as abnormal, describe unusual operating conditions or malfunctions of the technical system and are therefore relevant. Thus, relevant measured values, which can be attributed to a malfunction of the technical system, for example, are stored and can be examined in detail.

[0028] According to an advantageous embodiment of the invention, aggregated values ​​are calculated from the measured values ​​of operational data sets that are recognized as abnormal by the anomaly model. The aggregated values ​​are additionally sent to the server and stored there.

[0029] According to another advantageous embodiment of the invention, the aggregated values ​​are stored on the computer.

[0030] This saves bandwidth when communicating with a server.

[0031] According to an advantageous embodiment of the invention, the measured values ​​of operational data sets, which are recognized as abnormal by the anomaly model, are stored on the computer.

[0032] Operational data records and their measured values, identified as abnormal, describe unusual operating conditions or malfunctions of the technical system and are therefore relevant. Thus, relevant measured values, which can be attributed to a malfunction of the technical system, for example, are stored and can be examined in detail.

[0033] According to an advantageous embodiment of the invention, aggregated values ​​are calculated from the measured values ​​of operational data sets that are recognized as abnormal by the anomaly model. The aggregated values ​​are additionally stored on the computer.

[0034] According to an advantageous embodiment of the invention, the anomaly model is implemented as a single-class classifier or as a deep neural network in autoencoder architecture or as combinations thereof.

[0035] A one-class classifier is, for example, implemented as a one-class SVM. Possible combinations include, for example, feature reduction by an autoencoder and classification of the one-class classifier based on the features reduced by the autoencoder.

[0036] A system according to the invention for evaluating a technical plant comprises a computer. The computer has an anomaly model for checking recorded operational data sets. The system according to the invention is configured to carry out the method according to the invention.

[0037] The system according to the invention is suitable for persisting relevant data for the evaluation of a technical plant. The method according to the invention allows redundant or less relevant data to be discarded. This results in a significant reduction in the amount of data to be stored, without any substantial loss of information. This saves storage space when persisting the data, thereby also reducing costs.

[0038] The invention is not limited to the combination of features stated in the claims. For a person skilled in the art, further meaningful combinations of claims and / or individual claim features and / or features of the description and / or the figures will become apparent, in particular from the problem statement and / or the problem arising from a comparison with the prior art.

[0039] The invention will now be explained in more detail with reference to the illustrations. The invention is not limited to the embodiments shown in the illustrations. The illustrations only depict the subject matter of the invention schematically. They show: Fig. 1: A flowchart of the recording of a reference dataset during a reference phase, Fig. 2: A flowchart of the processing of recorded reference datasets during the reference phase, Fig. 3: a flowchart of the processing of recorded operational data records during an operational phase and Fig. 4: A diagram of an example operational data set.

[0040] Fig. Figure 1 shows a flowchart of the acquisition of a reference dataset 10 in a technical plant during a reference phase. The reference phase is carried out as a preparatory step to train an anomaly model 26, which will later be needed for the evaluation of the technical plant.

[0041] The technical system comprises at least one application, for example, a rotary table with a drive system. The drive system includes an electric motor, a converter for generating a three-phase alternating voltage for the electric motor, and a gearbox for reducing the speed of the electric motor.

[0042] During the reference phase, a reference data set 10 is acquired by a computer 22 in step 101. Computer 22 is a digital computer located in the technical system, in the field, and close to the application. Computer 22 serves, among other things, to acquire and process measured values. Such a computer 22 is often also referred to as an edge device.

[0043] The recorded reference data set 10 comprises a number of measured variables. These measured variables include, in particular, the rotational speed and torque of the electric motor. These measured variables, rotational speed and torque, are, for example, measured indirectly via the inverter by measuring the frequency and current of the output current to the electric motor.

[0044] Each measurand comprises a plurality of measured values, for example, 2048 measured values. The measured values ​​of each measurand are recorded sequentially and synchronously. For example, the measured values ​​of each measurand are recorded at equidistant time intervals of 5 ms each. The recorded reference dataset 10 thus comprises, for example, two different measurands, each with 2048 measured values.

[0045] The recorded reference data set 10 is sent from computer 22 to server 24 in step 102 and stored on server 24. Server 24 is a remote digital computer. Server 24 typically has more storage and computing resources than the local computer 22. Server 24 is located, for example, in a local data center or in a cloud. Server 24 is connected to computer 22 via a digital network, such as a LAN, WLAN, or the internet.

[0046] In this process, all measured values ​​of the recorded reference data set 10 are sent from the computer 22 to the server 24 and stored on the server 24.

[0047] During the reference phase, several more reference data sets 10 are recorded by computer 22 and sent to server 24, where they are stored. Steps 101 and 102 are repeated, and the recorded reference data sets 10 are stored on server 24. All measured values ​​from the recorded reference data sets 10 are stored on server 24.

[0048] Another reference data record 10 is recorded each time a defined trigger condition is met. Such a trigger condition is, for example, the start of an application cycle of the technical system when, for instance, the speed of an electric motor in the technical system exceeds a predefined minimum speed.

[0049] Alternatively, the recorded reference data set 10 is processed on computer 22. In this case, all measured values ​​of the recorded reference data set 10 are stored on computer 22. Server 24 is not required for this.

[0050] Fig. Figure 2 shows a flowchart of the processing of recorded reference data sets 10 during the reference phase. The processing of the previously recorded reference data sets 10 according to Fig. 2 will initially take place only once.

[0051] In step 103, an anomaly model 26 is trained on server 24 using the reference data records 10, which were previously sent to server 24. The anomaly model 26 is trained to check future operational data records 12.

[0052] After all reference data records 10 have been fed to the anomaly model 26, the anomaly model 26 is transferred from the server 24 to the computer 22 in step 104.

[0053] Alternatively, an anomaly model 26 is trained on computer 22 using the reference data sets 10. The anomaly model 26 is trained to check future recorded operational data sets 12. Server 24 is not required for this.

[0054] Fig. Figure 3 shows a flowchart of the processing of recorded operational data records 12 during an operational phase. The evaluation of the technical system is carried out during the operational phase.

[0055] During the operating phase, several operating data records 12 are recorded by the computer 22 in step 121. Each recorded operating data record 12 comprises a plurality of measured variables. These measured variables include, in particular, the rotational speed and torque of the electric motor. These measured variables, rotational speed and torque, are measured indirectly, for example, via the inverter by measuring the frequency and current of the output current to the electric motor.

[0056] Each measured quantity comprises a plurality of measured values, for example, 2048 measured values. The measured values ​​of each measured quantity are recorded sequentially and synchronously. For example, the measured values ​​of each measured quantity are recorded at equidistant time intervals of 5 ms each. The recorded operational data sets 12 thus comprise, for example, two different measured quantities, each with 2048 measured values.

[0057] In step 122, the recorded operational data records 12 are fed to the anomaly model 26 on the computer 22 and checked by the anomaly model 26. The operational data records 12 are then identified by the anomaly model 26 as normal or abnormal.

[0058] From the measured values ​​of the operational data sets 12, which are recognized as normal by the anomaly model 26, 123 aggregated values ​​are calculated in one step.

[0059] The aggregated values ​​calculated in step 123 are sent to server 24 in step 124 and stored on server 24. The measured values ​​of the operational data sets 12, which are recognized as normal by the anomaly model 26, are discarded.

[0060] The measured values ​​from the operational data sets 12, which are recognized as abnormal by the anomaly model 26, are sent to server 24 in step 125 and stored on server 24.

[0061] Optionally, aggregated values ​​are calculated from the measured values ​​of the operational data sets 12, which are recognized as abnormal by the anomaly model 26, and the calculated aggregated values ​​are sent to server 24 and stored on server 24.

[0062] Another operational data record 12 is recorded and processed each time a defined trigger condition is met, and when a defined period of time has passed since the recording of the previous operational data record 12.

[0063] One such trigger condition is, for example, the start of an application cycle of the technical system when, for example, the speed of an electric motor of the technical system exceeds a predetermined minimum speed.

[0064] The time period is defined, for example, as six hours. Therefore, a maximum of four operational data records (12) are recorded and processed per day.

[0065] Alternatively, the aggregated values ​​of the operational data records 12, which are identified as normal by the anomaly model 26, are stored on computer 22. The measured values ​​from the operational data records 12, which are identified as abnormal by the anomaly model 26, are also stored on computer 22. Optionally, aggregated values ​​are additionally calculated from the measured values ​​of the operational data records 12, which are identified as abnormal by the anomaly model 26, and stored on computer 22. Server 24 is not required for this.

[0066] Optionally, operational data records 12, which are identified as abnormal by the anomaly model 26, are treated as further reference data records 10. The measured values ​​of these further reference data records 10 are then sent from computer 22 to server 24, as in step 102, and stored on server 24.

[0067] The anomaly model 26 on server 24 is supplied with the additional reference data sets 10, and the anomaly model 26 is trained, as in step 103, with the additional reference data sets 10 sent to server 24.

[0068] After all further reference data sets 10 have been added to the anomaly model 26, the anomaly model 26 is transferred from the server 24 to the computer 22, as in step 104.

[0069] Alternatively, the additional reference data sets 10 are processed on computer 22. The measured values ​​of the additional reference data sets 10 are stored on computer 22. The additional reference data sets 10 are then fed to the anomaly model 26 on computer 22, and the anomaly model 26 is trained using these additional reference data sets. Server 24 is not required for this.

[0070] It is also conceivable that the technical system comprises several similarly designed applications, for example, rotary tables, each with a drive system. In this case, it is conceivable to send the reference data sets 10 of several similar applications, which are monitored by one or more computers 22, to the server 24 and merge them there. A common anomaly model 26 is then trained on the server 24 using the reference data sets 10 of the applications. After the reference phase, the common anomaly model 26 is then transferred from the server 24 to all computers 22.

[0071] For example, there are two computers 22. One computer 22 has a rotary table connected to it, and the other computer 22 has two additional rotary tables connected to it. The reference data sets 10 are recorded on the computers 22 and sent to the server 24. During the reference phase, a common anomaly model 26 is trained with all reference data sets 10 and then transferred from the server 24 to all computers 22. If anomalies arise due to a new operating point, the common anomaly model 26 is retrained and again transferred from the server 24 to all computers 22.

[0072] This has the advantage, among others, that if a new operating point is first detected on one rotary table and the anomaly model 26 is retrained accordingly, then when the operating point appears on the other rotary tables, it is already known and not considered anomalous. This saves computing power and memory, as well as manual effort.

[0073] Fig. Figure 4 shows a diagram of an exemplary operating data set 12. This exemplary operating data set 12 comprises two measured variables: the rotational speed and the torque of the electric motor. These measured variables, rotational speed and torque, are measured indirectly, for example, via the inverter by measuring the frequency and current of the output current to the electric motor.

[0074] Each measured quantity comprises 2048 individual measurements. These measurements are recorded sequentially and synchronously. Specifically, the measurements are taken at equidistant intervals of 5 ms each. Reference symbol list 10 Reference data set 12 Operational data set 22 computers 24 servers 26 Anomaly model 101-104 steps during the reference phase 121-125 steps during the operational phase

Claims

[1] Method for evaluating a technical installation, wherein the technical installation comprises an application with a drive system, wherein the drive system comprises an electric motor, a converter for generating a three-phase alternating voltage for the electric motor, and a gearbox for reducing the speed of the electric motor, and wherein during an operational phase several operational data records (12) are recorded by a computer (22), each comprising a plurality of measured quantities, whereby Each measurement quantity comprises a plurality of measured values, and wherein The measured values ​​of each quantity are recorded sequentially over time; and the recorded operational data records (12) on the computer (22) are checked by an anomaly model (26); and aggregated values ​​are calculated from the measured values ​​of operational data sets (12) which are recognized as normal by the anomaly model (26); and the aggregated values ​​are stored; and the measured values ​​of the operational data sets (12) which are recognized as normal by the anomaly model (26) are discarded, and wherein the recorded operating data sets (12) each include a rotational speed and a torque of the electric motor as measured variables, and wherein the measured quantities speed and torque are measured indirectly via the inverter by measuring the frequency and current of an output current to the electric motor; and wherein during a reference phase several reference data sets (10) are recorded by the computer (22), each comprising a plurality of measured quantities, whereby Each measurement quantity comprises a plurality of measured values, and wherein The measured values ​​of each quantity are recorded sequentially over time; and on the computer (22) the anomaly model (26) is trained with the reference data sets (10) in order to check future recorded operating data sets (12), wherein the recorded reference data sets (10) each include a rotational speed and a torque of the electric motor as measured variables, and wherein The measured quantities speed and torque are measured indirectly via the inverter by measuring the frequency and current of an output current to the electric motor, and wherein Operational data records (12) which are identified as abnormal by the anomaly model (26) are treated as further reference data records (10); and on the computer (22) the anomaly model (26) is trained with the further reference data sets (10), and wherein the anomaly model (26) is implemented as a single-class classifier or as a deep neural network in autoencoder architecture or as combinations thereof. [2] Method according to any of the preceding claims, characterized by , that each additional operational data record (12) is recorded, when a defined trigger condition is met, and when a defined period of time has passed since the recording of the previous operational data record (12). [3] Method according to any of the preceding claims, characterized by , that the aggregated values ​​are sent to a server (24) and stored on the server (24). [4] Method according to claim 3, characterized by , that the measured values ​​of operational data sets (12) which are identified as abnormal by the anomaly model (26) are sent to the server (24) and stored on the server (24). [5] Method according to claim 4, characterized by , that aggregated values ​​are calculated from the measured values ​​of operational data sets (12) which are identified as abnormal by the anomaly model (26); and The aggregated values ​​are additionally sent to server (24) and stored on server (24). [6] Method according to any of the preceding claims, characterized by , that the aggregated values ​​are stored on the computer (22). [7] Method according to claim 6, characterized by , that the measured values ​​of operational data sets (12) which are identified as abnormal by the anomaly model (26) are stored on the computer (22). [8] Method according to claim 7, characterized by , that aggregated values ​​are calculated from the measured values ​​of operational data sets (12) which are identified as abnormal by the anomaly model (26); and the aggregated values ​​are additionally stored on the computer (22). [9] System for the evaluation of a technical installation, comprising a computer (22) which has an anomaly model (26) for checking recorded operational data sets (12), wherein the system is set up to carry out the method according to one of the preceding claims.

Images