Method for condition monitoring of an electrical machine

Abstract

The invention relates to a method (100) for monitoring the condition of an electrical machine, comprising: providing or receiving (101) a measurement signal that has been detected as a phase current of the electrical machine; determining (102), based on the measurement signal, a transformed measurement signal using a transformation of the phase current into a torque-generating current; determining (104) a deviation of the transformed measurement signal from a reference measurement signal (403); and determining (106) when the deviation exceeds a predetermined threshold that an anomaly of the electrical machine is present.

Description

The present invention relates to a method for monitoring the state of an electric machine, to a method for determining a reference measurement signal within the scope of a training phase for monitoring the state of an electric machine, and to a computing unit and to a computer program for carrying it out.BACKGROUND OF THE INVENTIONBy data-assisted monitoring of an electric motor with regard to anomalies that occur, it is possible to detect changes in its state at an early stage, which changes can occur, for example, as a result of wear. This anomaly detection can reduce or even avoid cost-intensive shutdowns in particular.The document CN 1 04 459 537 B describes a method and a device for monitoring and diagnosing the wear state of a drive motor for electric vehicles. It includes monitoring the feedback torque, rotor torque, and three-phase open-circuit currents to detect faults such as phase failure or demagnetization and shut down the system when needed.Laid-open specification KR 10 2018 0 038 723 A discloses a method and a device for fault detection in motor systems, in particular for assessing the presence of rotor anomalies. Here, d-axis current and q-axis current and an excitation current are used to determine the q-axis voltage and compare it with a reference voltage. Deviations indicate rotor anomalies.Disclosure of the InventionAccording to the invention, a method for monitoring the state of an electric machine, a method for determining a reference measurement signal, and a computing unit and a computer program for carrying them out are proposed, having the features of the independent patent claims. Advantageous embodiments are the subject matter of the dependent claims and of the following description.The invention relates to a method for monitoring the state of an electric machine, i.e. for example an electric motor, and specifically for abnormality detection. Abnormality detection can be based, for example, on an abnormality detection model which can be initially adapted in a training phase to the previously collected data-also referred to below as training measurement signals-of the electric motor to be monitored and which allows the assessment of newly arriving monitoring data in a subsequent or later monitoring phase. In this case, changes in the distribution or characteristic of the monitoring data of the motor can in particular give an indication of an abnormality or a change in the state of the motor.Anomaly detection models can be based in particular on complex models from the field of deep learning, wherein external storage and computing resources, for example cloud applications, are required in particular for creating and / or adapting the respective model, since a large amount of data has to be collected and stored for creating and / or adapting the model and the training itself is very computation-intensive.Against this background, a possibility is proposed within the scope of the present invention, which makes do with fewer memory and computing resources. In this case, a measurement signal is provided or received, wherein a phase current of the electric machine has been detected as the measurement signal, for example by means of a current sensor which is generally present in any case in the electric machine. The measurement signal can be received by a computer-assisted computing unit, for example, or provided by the latter. Based on the measurement signal, a transformed measurement signal is determined, wherein a transformation of the phase current into a torque-forming current is carried out. In a next method step, a deviation of the transformed measurement signal from a reference measurement signal is determined. If this deviation exceeds a predetermined threshold value, it can be determined by this that an abnormality of the electric machine is present.By transforming the measurement signal, it is possible in particular to obtain a measurement signal which is at least almost independent of the rotational speed of the electric machine. Therefore, the method for monitoring the state can reliably determine the presence of an anomaly of the electric machine even under variable operating conditions, in particular at a variable rotational speed of the electric machine, and the applicability of the method for monitoring the state of the electric machine does not have to be limited to a specific rotational speed range. In addition, separate anomaly detection models are not required for different operating conditions. This allows a reduction in the storage and / or the computational effort when performing the proposed method both during the training and the monitoring phase.A further advantage of the proposed method is that a phase current of the electric machine is determined as the measurement signal, which phase current can be detected in particular by means of a current sensor connected to a drive control device. The detection of the phase current is required in particular for the operation of the electric machine, so that no additional sensors, such as acceleration sensors for detecting vibrations or the like, are required for carrying out the method for monitoring the state. This allows, on the one hand, a cost-effective implementation of a device required for the method, and a simple and extensive applicability of the method, in particular in the case of different electric machines or in the case of electric machines of different construction.In one embodiment, the reference measurement signal is determined within the scope of a training phase. In this case, a plurality of training measurement signals, for example a computer-assisted arithmetic logic unit, are provided, wherein a phase current of the electric machine is in each case detected as the training measurement signal. Based on the training measurement signals, a transformed training measurement signal is determined in each case, wherein a transformation of the phase current into a torque-forming current is used for this purpose. A reference measurement signal is determined from the transformed training measurement signals.The provision and transformation of the training measurement signals can be effected in accordance with the provision and transformation of the measurement signal in accordance with the method for state monitoring described above. In particular, the transformation of the training measurement signals can in each case determine a transformed training measurement signal which is at least virtually independent of the rotational speed, resulting in the advantages corresponding to the above description.The transformation of the training measurement signals makes it possible in particular to reduce or minimize the speed dependence thereof, such that anomalies can be reliably detected even in the case of variable operating conditions of the electric machine. In addition, the data volume of required training measurement signals can thereby be reduced, since it is not necessary to collect and store a sufficiently large amount of training measurement signals for a plurality of different operating conditions or rotational speeds, which can contribute, for example, to a more efficient use of storage and / or computing resources. Furthermore, costs for a corresponding infrastructure can be reduced and, in particular, the provision of external memory and / or computing resources can be dispensed with, as a result of which the disadvantages associated therewith, such as, for example, possible data protection problems, are avoided. Thus, in particular the data integrity of the training measurement signals or of the transformed training measurement signals can be ensured.Due to the reduced data set of transformed training measurement signals or with the associated reduction of the computing effort of the determination of the reference measurement signal that is in any case less complex, it is furthermore possible to reduce the time duration of the training phase even more, as a result of which costs can be saved, for example. The training phase is consequently kept very short in comparison with the subsequent monitoring phase. At the same time, it is thereby possible, in the event of process changes or in particular after the electric machine has been put into operation, to provide an adapted reference measurement signal promptly.Furthermore, it is possible that the reference measurement signal determined during the training phase is determined for one, or only for a few, electric machines. This reference measurement signal can subsequently be provided to identically constructed electric machines and used for the corresponding detection of anomalies. As a result, the training phase does not have to be run through for each electric machine, which results in a time saving and cost saving. At the same time, it is possible to carry out the method for monitoring the state using an individual reference measurement signal suitable for the electric machine, as a result of which anomalies in the electric machine can be detected particularly well.In addition, the training measurement signals and the transformed training measurement signals can be discarded or deleted (directly) after the corresponding transformation or after determination of the reference measurement signal. This allows, in particular, a memory resource-optimized execution of the corresponding method.In particular, the transformation of the phase current is carried out by means of a Park and Clark transformation. In particular, the resulting currents can be independent, or almost independent, of the rotational speed of the electric machine. According to the above description, this transformation allows advantageous detection of anomalies of the electric machine over variable operating conditions.In a further embodiment, the transformed measurement signal is adapted by a downstream preprocessing. According to a further embodiment, this preprocessing can comprise smoothing and / or mean value cleaning.The smoothing can thereby improve the signal-to-noise ratio of the transformed measurement signal. The mean value correction allows the compensation of offsets of the transformed measurement signals, whereby a comparison of different measurement signals and thus in particular the generalization of the resulting reference measurement signal can be improved.It is furthermore possible that the transformed training measurement signal is also adapted by the described preprocessing, in particular by smoothing and / or mean value cleaning. As a result, the reference measurement signal determined within the framework of the training phase has an advantageous signal-to-noise ratio. The mean value cleaning is particularly advantageous because the deviation of the transformed measurement signal from the reference measurement signal is thereby hardly subject to the influence of any offsets and can thus be determined particularly accurately. This enables an advantageous detection of anomalies in the electric machine to be monitored.In this case, the transformed training measurement signals and the transformed measurement signal are expediently preprocessed in the same way by means of the preprocessing described above. Alternatively, neither the transformed training measurement signals nor the transformed measurement signal can be preprocessed.A similar preprocessing is advantageous in this case since the transformed training measurement signals, and thus also the reference measurement signal, and the transformed measurement signal can be compared particularly well with respect to characteristic properties, as a result of which the method for state monitoring can be carried out particularly advantageously.In a further embodiment, at least one measure is carried out if an abnormality in the electric machine has been detected. In this case, the at least one measure can comprise storing the associated measurement signal, storing information relating to the associated measurement signal, displaying a warning and / or triggering an alarm.By storing the measurement signal or information relating thereto, a corresponding anomaly event can be tracked and analyzed further, in order to find the cause of the occurrence of the anomaly, for example. This can serve, for example, as a basis for improving the electric machine. The display of a warning or the triggering of an alarm can give an indication of an abnormal behavior of the electric machine. Furthermore or additionally thereto, the detection of an abnormality could also lead to a shutdown of the electric machine in order to prevent any further damage to the same.In a further embodiment, the training measurement signals are detected within the scope of an initial data collection phase, wherein this data collection phase is initiated by a trigger signal. According to a further embodiment, the trigger signal can be generated by an PLC program.The data acquisition can be triggered in a targeted manner by the trigger signal, so that in particular no additional data are acquired and stored which are not required for the determination of the reference measurement signal. As a result, the load on computing and storage resources can be reduced. PLC programs offer a reliable and cost-effective solution for generating a trigger signal.In a further embodiment, the reference measurement signal is determined based on a Euclidean mean value and / or dynamic time warping.The determination of the reference measurement signal from a plurality of transformed training measurement signals based on the Euclidean mean value represents a particularly simple and thus efficient possibility of combining a plurality of training measurement signals to form a reference measurement signal.Dynamic time warping or dynamic time normalization is particularly advantageous if the transformed training measurement signals to be combined are not directly comparable on account of distortions or displacements of the time axis. By dynamic time warping, an optimum assignment between the transformed training measurement signals to be merged can be established in this case by the respective measurement signals being distorted or stretched along the time axis. By taking time displacements into account, the reference measurement signal determined in this way can be determined particularly accurately or depicts the real behavior of the phase current of the electric machine particularly well and thus represents a particularly representative reference measurement signal.In a further embodiment, an anomaly value is determined from the deviation of the transformed measurement signal from the reference measurement signal, which anomaly value can assume values between 0 and 1.This allows, for example, a simplified interpretability of the aforementioned deviation. The anomaly value can assume, for example, a value of 0 if the transformed measurement signal is identical to the reference measurement signal, i.e. there is no deviation, and a value of 1 if the deviation is so large that in any case it is possible to assume that the electrical machine to be monitored is malfunctioning in the sense of an anomaly. Since the anomaly value can only assume values between 0 and 1, a corresponding threshold value can also be defined independently of the absolute value of the deviation. As a result, it is possible, for example, to use a uniform threshold value for different electric machines or electric machines of different types of construction.The invention is also concerned with a method for determining a reference measurement signal within the scope of a training phase for use in monitoring the state of an electric machine, as has already been explained in detail above, for example. In this case, a plurality of training measurement signals, for example a computer-assisted arithmetic logic unit, are provided, wherein a phase current of the electric machine is in each case detected as the training measurement signal. Based on the training measurement signals, a transformed training measurement signal is determined in each case, wherein a transformation of the phase current into a torque-forming current is used for this purpose. A reference measurement signal is then determined from the transformed training measurement signals.The reference measurement signal can be used in particular for carrying out the method described above for monitoring the state of an electric machine.According to the above description, the speed dependence thereof can be reduced by the transformation of the training measurement signals, which results in particular in the advantages according to the above description. In this case, in particular, the complexity of the reference measurement signal determined from the transformed training measurement signal can be reduced, as a result of which no external storage and / or computing resources are required and, in addition, the time duration of the training phase can be reduced.In the following, further embodiments of the method for determining a reference measurement signal will be discussed. Advantages of the respective embodiment can also be taken from the corresponding preceding description.The transformation of the training measurement signals can be carried out in particular by means of a Park and Clark transformation. In this case, a three-phase training measurement signal can be converted into an at least nearly rotational-speed-independent DC variable, as a result of which in particular processing and / or storage of the corresponding measurement signals required for the training is less resource-intensive. In addition, the reference measurement signal determined from the transformed training measurement signals thus likewise has at least virtually no rotational speed dependency, which in particular makes it possible to advantageously carry out the above method for state monitoring.Furthermore, preprocessing of the transformed training measurement signals can be carried out, wherein the preprocessing can comprise, in particular, smoothing and / or mean value purification. As a result, the training measurement signals can be improved, which in particular makes the method for state monitoring particularly accurate and reliable to carry out.Furthermore, the training measurement signals can be detected within the scope of an initial data collection phase, wherein the initial data collection phase is initiated by a trigger signal. This trigger signal can be advantageously generated, in particular, by an PLC program. The data collection phase can thus be triggered in a targeted manner when training measurement signals are to be detected. It can therefore be avoided in particular that the training measurement signals contain data which are not required for the determination of the reference measurement signal, whereby the utilization of computing and / or storage resources is optimized.Furthermore, the reference measurement signal can be determined based on a Euclidean mean value and / or dynamic time warping. This makes it possible to combine the individual transformed training measurement signals advantageously to form a reference measurement signal, which can be used in particular for the above method for state monitoring.A computing unit according to the invention, e.g. a control device of a motor vehicle, is configured, in particular by programming, to carry out a method according to the invention.The implementation of the method according to the invention in the form of a computer program or computer program product with program code for carrying out all method steps is also advantageous since this causes particularly low costs, in particular if an executing control device is also used for further tasks and is therefore present in any case. Finally, a machine-readable storage medium is provided with a computer program stored thereon, as described above. Suitable storage media or data carriers for providing the computer program are, in particular, magnetic, optical and electrical memories, such as hard disks, flash memories, EEPROMs, DVDs, among others. Download of a program via computer networks (Internet, intranet, etc.) is also possible. Such a download can be effected in a wired or wired or wireless manner (e.g. via a WLAN network, a 3G, 4G, 5G or 6G connection, etc.).Further advantages and embodiments of the invention will become apparent from the description and the accompanying drawing.The invention is schematically illustrated in the drawing on the basis of exemplary embodiments and is described below with reference to the drawing.Brief Description of the DrawingsFIG. 1 shows a flow diagram of an embodiment of the method according to the invention for monitoring the state of an electric machine. FIG. 2 shows a flow diagram of an embodiment of the method according to the invention for determining a reference measurement signal. FIG. 3 shows a flow chart of an embodiment in which the reference measurement signal is provided. FIG. 4 ashows schematically the transformation of the acquired training measurement signals into a torque-forming current in each case. FIG. 4 bshows schematically the determination of a reference measurement signal from the training measurement signals illustrated in FIG. 4 a.Embodiment(s) of the InventionFIG. 1 shows a flow diagram 100 which shows an embodiment of the method according to the invention for monitoring the state of an electric machine.In a method step 101, a measurement signal is provided to a computer-assisted computing unit, for example, or received by it. The measurement signal is a phase current of the electric machine, which can be detected, for example, by means of a corresponding current sensor.In a step 102, a transformed measurement signal is determined based on the measurement signal, wherein the corresponding phase current is converted into a torque-forming current by the transformation. This transformation can be based in particular on a Park and Clark transformation.Furthermore, in step 103, the transformed measurement signal is further processed or improved by preprocessing. This preprocessing can comprise in particular a smoothing and / or a mean value correction of the transformed measurement signal.In a further method step 104, a deviation of the transformed measurement signal from a reference measurement signal is determined.In addition, in method step 105, a check is made as to whether this deviation exceeds a predetermined threshold value. If this is not the case, the method is carried out again, for example starting from step 101, in order thereby to enable continuous state monitoring of the electric machine.If the deviation exceeds the predetermined threshold value, it is determined in a further step 106 that an abnormality of the electric machine is present. In step 107, an anomaly value is determined from the deviation of the transformed measurement signal from the reference measurement signal, which anomaly value can assume values between 0 and 1. This makes it possible to facilitate the interpretability of this deviation.Subsequently, in step 108, at least one measure is carried out, which can comprise in particular a storage of the associated measurement signal, a storage of information relating to the associated measurement signal, a display of a warning and / or a triggering of an alarm.FIG. 2 shows a flow diagram 200 of an embodiment of the method according to the invention for determining a reference measurement signal within the scope of a training phase for monitoring the state of an electric machine.In a step 201, a plurality of training measurement signals are provided, wherein a phase current of the electric machine has been detected as the training measurement signal in each case.Furthermore, in a step 202, a transformed training measurement signal is determined in each case, wherein the respective phase current is converted into a torque-forming current by the transformation. This method step 202 corresponds in particular to step 102 of the method illustrated in FIG. 1. The transformation can also be based in particular on a Park and Clark transformation.In step 203, preprocessing of the transformed training measurement signals is carried out. This preprocessing may include smoothing and / or averaging. The preprocessing of the transformed training measurement signals is carried out in particular in the same way as the preprocessing of the transformed measurement signal, according to method step 103 in FIG. 1.In a further method step 204, a reference measurement signal is determined on the basis of the transformed training measurement signals. The determination of the reference measurement signal can be based in particular on a Euclidean mean value and / or dynamic time warping. This reference measurement signal can be used in particular in the method for state monitoring.FIG. 3 shows a flow chart 300 in which a method is shown in which first of all the embodiment 200 of the method for determining a reference measurement signal shown in FIG. 2 is carried out and then the method 100 for monitoring the state of the electric machine shown in FIG. 1 is shown. In this case, the reference measurement signal provided by the method for determining a reference measurement signal is used accordingly in the method for state monitoring.It is possible in particular for the method for determining a reference measurement signal to be carried out in a first step and then, using the reference measurement signal determined therefrom, for the method for state monitoring.FIG. 4 ashows schematically the transformation of the acquired training measurement signals 401, 402, for example within the scope of the method 200 illustrated in FIG. 2, into a respective torque-forming current 401', 402'. The transformation can be based in particular on a Park and Clark transformation. The horizontal axis 410 in FIG. 4 ais a time axis, while the corresponding current value of the respective training measurement signal 401, 402 or of the respective transformed training measurement signal 401', 402' is plotted on the vertical axis 420.The transformed training measurement signals 401', 402' have a virtually negligible rotational speed dependence, in particular in comparison with the non-transformed training measurement signals 401, 402.The transformed measurement signals 401', 402' can be further adjusted by preprocessing, which can comprise smoothing and / or mean value cleaning.Analogously to the transformation of the training measurement signals 401', 402' shown in FIG. 4 a, the measurement signals which are detected within the scope of the method for state determination can also be transformed in the same way.FIG. 4 bshows schematically the determination of a reference measurement signal 403 from the transformed training measurement signals 401', 402' shown in FIG. 4 a. In particular, the two transformed training measurement signals 401', 402' are combined to form a reference measurement signal 403, which can be based, for example, by calculating a Euclidean mean value and / or on dynamic time warping.The training measurement signals 401, 402 and the transformed training measurement signals 401', 402' can be discarded or deleted after the corresponding transformation or after determination of the reference measurement signal 403, so that a memory resource-optimized execution of the corresponding method is thereby made possible.

Claims

A method (100) for monitoring the state of an electric machine, comprising: providing or receiving (101) a measurement signal that has been detected as a phase current of the electric machine; determining (102), based on the measurement signal, a transformed measurement signal using a transformation of the phase current into a torque-forming current; determining (104) a deviation of the transformed measurement signal from a reference measurement signal (403); and determining (106), if the deviation exceeds a predetermined threshold value, that an anomaly of the electric machine is present.The method of claim 1, wherein the transformation of the phase current is performed by means of a Park and Clark transformation.Method according to any of the preceding claims, wherein the determining, based on the measurement signal, the transformed measurement signal comprises a subsequent preprocessing (103) of the measurement signal.Method according to claim 3, wherein the preprocessing (103) comprises smoothing and / or mean value cleaning.The method of any preceding claim, further comprising, when an anomaly has been determined: performing (108) at least one action.Method according to claim 5, wherein the at least one measure comprises storing the associated measurement signal, storing information relating to the associated measurement signal, displaying a warning and / or triggering an alarm.Method according to one of the preceding claims, wherein an anomaly value is determined (107) from the deviation of the transformed measurement signal from the reference measurement signal (403), wherein the anomaly value has a value between 0 and 1.Method according to one of the preceding claims, wherein the reference measurement signal (403) has been determined within the scope of a training phase, wherein the training phase comprises: providing (201) a plurality of training measurement signals (401, 402), wherein a phase current of the electric machine is or has been detected in each case as the training measurement signal (401, 402); determining (202), based on the training measurement signals (401, 402), a transformed training measurement signal (401', 402'), in each case, using a transformation of the phase current into a torque-forming current; determining (204), based on the transformed training measurement signals (401', 402'), the reference measurement signal (403).Method (200) for determining a reference measurement signal (403) within the scope of a training phase for the use of state monitoring of an electric machine, comprising: providing (201) a plurality of training measurement signals (401, 402), wherein a respective phase current of the electric machine is or has been detected as the training measurement signal (401, 402); determining (202), based on the training measurement signals (401, 402), a respective transformed training measurement signal (401', 402'), using a transformation of the phase current into a torque-forming current; determining (204), based on the transformed training measurement signals (401', 402'), the reference measurement signal (403).Method according to Claim 8 or 9, wherein the training measurement signals (401, 402) are detected within the scope of an initial data collection phase, wherein the initial data collection phase is initiated by a trigger signal.The method of claim 10, wherein the trigger signal is generated by an PLC program.Method according to one of Claims 8 to 11, wherein the determination of the reference measurement signal is based on a Euclidean mean value and / or dynamic time warping.Computing unit which is configured to carry out the method according to one of Claims 1 to 8 and / or the method according to one of Claims 9 to 12.A computer program comprising instructions which, when the program is executed by a computer, cause the computer to carry out the method according to any one of claims 1 to 8 and / or the method according to any one of claims 9 to 12.Computer-readable data medium on which the computer program according to Claim 14 is stored.

Images