Removing stray signal components from measured motor current signals to optimize motor diagnosis
By generating a reference current signal using a motor model and optimizing denoising parameters using a noise removal module, the problem of stray signal components in the motor current signal is solved, achieving accuracy and reliability in motor diagnosis and enabling the identification of problems such as motor misalignment and imbalance.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- SIEMENS AG
- Filing Date
- 2025-11-28
- Publication Date
- 2026-05-29
AI Technical Summary
Existing technologies struggle to effectively remove stray signal components from motor current signals, leading to inaccurate motor diagnostics, particularly in the inability to accurately identify motor misalignment and imbalance.
By generating reference current signals and modeling current signals using a motor model, and combining a noise removal module and a time series filter, the noise removal parameters are optimized to remove signal components caused by power grid interference, while retaining signal components specific to motor issues.
It achieves precise noise reduction of motor current signals, improves the reliability and accuracy of motor diagnosis, and can effectively identify problems such as motor misalignment and imbalance.
Smart Images
Figure CN122109808A_ABST
Abstract
Description
Technical Field
[0001] This disclosure relates to a computer-implemented method for removing stray signal components from a measured motor current signal to optimize motor diagnostics, a system for performing the method, and a corresponding computer program product. Background Technology
[0002] Electric motors play a crucial role in a wide range of industrial applications, such as pumps and conveyor belts. Therefore, motor failure can have significant consequences, causing critical components of production lines or water supply systems to malfunction. This, in turn, leads to substantial costs due to production stoppages. Given the relatively low costs associated with spare parts (such as bearings) or maintenance tasks (such as motor realignment), there is a high demand for monitoring systems for these critical motors.
[0003] Different condition monitoring systems are known to be used to monitor these critical production components. These systems utilize various motor signals to identify potential motor problems. A key signal is the motor current characteristic, which is recorded using sensors at the motor terminals. If the machine suffers different types of damage, this signal typically manifests as suspicious patterns and anomalies in the data. Machine learning algorithms can then be trained to detect these anomalies and provide early warnings to maintenance engineers.
[0004] However, the motor current is significantly affected by the supplied motor voltage signal. In real-world applications, this motor voltage signal does not have a perfect sine wave but is affected by various interference sources, such as artifacts from other electrical devices connected to the same power grid. These devices introduce voltage components into the grid, which in turn affect the motor current signal. Therefore, crucial indications in the motor's output signal can be obscured.
[0005] Noise reduction techniques such as low-pass or high-pass filtering are commonly used to improve the signal-to-noise ratio of a motor's input voltage. However, these techniques do not take into account the specific characteristics of the motor, which may result in the removal of signal components that are not caused by grid interference but may still be important for subsequent motor current analysis (e.g., motor current characteristic analysis (MCSA)).
[0006] Additionally, these techniques may unintentionally eliminate signal components caused by mechanical problems in the motor, such as misalignment or imbalance. It is crucial that these components are not discarded from the current signal, as they need to be analyzed in the MCSA to detect such problems.
[0007] US6199023 discloses a method for mitigating stray signals in the voltage input of a motor. This method involves using an electronic model of the motor to estimate the signal components generated by interference in the input voltage signal. The process involves transforming the motor voltage signal to the frequency domain, removing its fundamental components, and transforming it back to the time domain. The resulting signal is then used as input to the motor model to calculate the corresponding motor current. Importantly, this motor current signal retains the components originating from stray signal elements in the input signal. In the final step, the calculated current signal is subtracted from the current signal measured directly at the motor terminals.
[0008] However, this approach has some trade-offs. While it avoids using a high-fidelity finite element model of the motor (due to the complete removal of the fundamental frequency from the voltage signal), it relies on a low-fidelity simplified electrical model, providing less accurate motor current values corresponding to disturbances in the motor voltage signal. Furthermore, the signal subtraction assumes that the noise signal can be linearly decomposed into disturbance components and the actual motor signal, which is typically not the case in reality, thus failing to lead to optimal determination of the error components.
[0009] Therefore, the purpose of this invention is to reduce the signal component in the motor current caused by interference from the power grid, while retaining the signal component that indicates possible motor damage or misalignment. Summary of the Invention
[0010] This objective is achieved through the features of the independent claim. The dependent claims also include embodiments of the invention.
[0011] The first aspect relates to a computer-implemented method for removing stray signal components from measured motor current signals to optimize motor diagnostics, comprising the following steps: Receives the measured motor voltage signal and the measured motor current signal measured at the motor. The motor model generates a modeled current signal based on the measured motor voltage signal. The motor model generates a reference current signal based on an undisturbed sinusoidal motor voltage signal, characterized in that... The noise removal module is adapted by inputting the modeled current signal and the reference current signal into it to infer the optimal denoising parameters. The denoised motor current signal is determined by inputting the measured motor current signal into an adapted noise removal module. The denoised motor current signal is output for use in optimizing motor diagnostics.
[0012] The second aspect relates to an auxiliary device comprising at least one processor configured to perform a step of removing stray signal components from a measured motor current signal to optimize motor diagnostics, the step comprising the following steps: Receives the measured motor voltage signal and the measured motor current signal measured at the motor. The motor model generates a modeled current signal based on the measured motor voltage signal. The motor model generates a reference current signal based on an undisturbed sinusoidal motor voltage signal, characterized in that... The noise removal module is adapted by inputting the modeled current signal and the reference current signal into it to infer the optimal denoising parameters. The denoised motor current signal is determined by inputting the measured motor current signal into an adapted noise removal module. The denoised motor current signal is output for use in optimizing motor diagnostics.
[0013] A third aspect of the present invention relates to a computer program product that can be directly loaded into the internal memory of at least one digital computer, the computer program product including a software code portion for performing the steps of the above-described method when the product is run on the at least one digital computer.
[0014] Advantageously, the described method uses a motor model that outputs a highly accurate motor current for the motor under consideration. The precisely calculated output signal of this machine model allows for the adaptation of a noise removal module to estimate the accurate current noise signal and / or filter parameters. If the noise signal is not accurate enough, this can result in the removal of important signal components from the measured motor current, containing information about potential motor problems such as misalignment and imbalance. Compared to more general techniques that do not use a motor model at all, the proposed method ensures the removal of only signal components arising from interference in the power grid.
[0015] Implementation Example Description In an embodiment of the method, the denoised motor current signal is input into a motor current analyzer, which outputs motor diagnostic results.
[0016] Advantageously, the resulting motor diagnostics are based on optimized noise-reducing motor current, thus producing reliable and accurate motor diagnostic results.
[0017] In one embodiment, a sinusoidal motor voltage signal is generated, the frequency of which is equal to the frequency of the supplied motor voltage signal.
[0018] Advantageously, based on this perfect sinusoidal motor voltage signal, the motor model generates a perfectly undisturbed reference current signal for an ideal motor. Subsequently, by correlating the modeled current signal with the undisturbed reference current signal, an "undisturbed" noise signal is generated, which can be considered for generating a denoised motor current signal.
[0019] In the embodiments, the motor model is a deterministic simulation model of the motor (preferably a high-fidelity finite element model, a high-fidelity digital twin) or a trained data-based model or a combination of both.
[0020] This motor model generates reliable and accurate motor current signals.
[0021] In an embodiment, the motor current analyzer is a motor current feature analyzer (MCSA) or a machine learning model trained to analyze motor current signals, preferably a deep neural network, a self-organizing map, or a decision tree.
[0022] MCSA is widely used in industrial applications for condition-based monitoring and predictive maintenance of motors because it provides an effective way to continuously monitor motor health and plan maintenance accordingly. It is cost-effective and requires low adaptation effort. Deep neural networks can be trained to analyze different types of motors, making it highly flexible.
[0023] In this embodiment, the noise removal module determines the set of denoising parameters based on the difference between the generated modeled current signal and the generated reference current signal.
[0024] The differential signal exhibits improved noise immunity because common-mode noise affecting the two signals is removed when the receiver measures the difference between the two signals.
[0025] In one embodiment, the set of denoising parameters is determined by inputting the difference between the modeled current signal and the reference current signal into a time series data filter and optimizing the time series data filter to minimize the difference.
[0026] Preferably, the time series data filter is constructed as a linear autoregressive model, a Wiener filter, a kernel adaptive filter, or a recurrent neural network.
[0027] The denoising parameters are weights of an AI-based filter model that are continuously adapted to the measured motor voltage under consideration, thereby generating a continuously optimized denoised motor current signal.
[0028] In one embodiment, the measured motor current is input to a trained time-series data filter, which outputs a denoised motor current signal.
[0029] Therefore, the output denoised motor current signal is optimized relative to the measured voltage signal and the measured current signal, respectively.
[0030] In another embodiment, the noise removal module includes a noise cancellation unit that filters out unwanted signal components of the measured motor voltage signal based on a reference noise signal.
[0031] Preferably, the reference noise signal is generated by inputting the modeled current signal and the reference current signal into the noise extractor unit and estimating the reference noise signal caused only by voltage fluctuations.
[0032] The noise signal is estimated by calculating the difference between the simulated motor current and the reference current signal. This is a well-known method and requires low processing power for computation. Therefore, the noise removal module can be implemented using a processor with low processing power.
[0033] This method is most effective if the reference noise signal is not correlated with the signal of interest (i.e., the measured current signal), but is correlated with the noise to be reduced (i.e., the disturbance originating from the input voltage signal).
[0034] In this embodiment, the estimated noise signal is refined by filtering out any frequency components typically observed in the measured motor signal, such as the fundamental power supply frequency or winding harmonics.
[0035] Therefore, the reference noise signal is further reduced to the fluctuation of interest. Attached Figure Description
[0036] The invention will be explained in more detail with reference to the accompanying drawings. Similar objects will be labeled with the same reference numerals.
[0037] Figure 1 An embodiment of the method of the present invention is illustrated schematically as a flowchart.
[0038] Figure 2 An embodiment of the auxiliary device of the present invention is illustrated schematically.
[0039] Figure 3 A detailed embodiment of the auxiliary device of the present invention, which has a noise removal module utilizing a time series filter, is schematically illustrated.
[0040] Figure 4 A detailed embodiment of the auxiliary device of the present invention, which includes a noise removal module utilizing noise cancellation functionality, is schematically illustrated.
[0041] Figure 5 An exemplary motor current signal for measurement and noise reduction is schematically shown.
[0042] It should be noted that in the following detailed description of the embodiments, the drawings are merely illustrative, and the elements illustrated are not necessarily shown to scale. Rather, the drawings are intended to illustrate the function and cooperation of components. Here, it should be understood that any connection or coupling of functional units, modules, components, or other physical or functional elements may also be achieved by directly or indirectly connecting coupling elements (e.g., via one or more intermediate elements). Connections or couplings of entities or components may be achieved, for example, by wired, wireless, and / or a combination of wired and wireless connections. Functional modules may be implemented by dedicated hardware (e.g., processor, firmware) or by software and / or by a combination of dedicated hardware and firmware and software. It should also be noted that functional steps of the associated method can be performed for each functional module described for the apparatus, and vice versa. Detailed Implementation
[0043] It should be understood that the above description of the examples is intended to be illustrative, and the components shown are readily adaptable to various modifications. For example, the concepts shown can be applied to different technical systems, and especially to different subtypes of corresponding technical systems with only minor adaptations.
[0044] In industrial environments, motor failures can have a significant impact, causing critical components of production lines or water supply systems to malfunction. This, in turn, leads to substantial costs due to production stoppages. Given the relatively low costs associated with maintenance tasks such as motor realignment, there is a high demand for monitoring systems for these critical motors. A key signal is the motor current characteristic, which is recorded using sensors at the motor terminals. If the machine suffers different types of damage, this signal typically manifests as suspicious patterns and anomalies in the data.
[0045] Implementation examples of the proposed computer-based method are in Figure 1 The diagram in the image below illustrates and explains that this method is used to reduce the signal component in the motor current signal caused by interference in the power grid, while retaining the signal component that indicates possible motor damage or misalignment.
[0046] The motor is preferably deployed in industrial environments, such as stepper motors in production facilities, as drives used in conveyor systems, heavy machinery, vehicles, and so on.
[0047] In the first step S1, a measured motor voltage signal Um and a corresponding measured motor current signal Im, measured at the motor, are received. The measured motor voltage and current signals include data samples recorded via sensors located at the respective motor terminals. A modeled current signal Im is generated by the motor model based on the measured motor voltage signal Um, see S2. In parallel, a reference current signal Ips is generated by the same motor model based on an undisturbed sinusoidal motor voltage signal Up, see step S3. The undisturbed sinusoidal motor voltage signal Up is a perfect sinusoidal signal with a frequency equal to the frequency of the supplied motor voltage, such as 50Hz or 60Hz of the power network.
[0048] The modeled current signal Ims and the reference current signal Ips are input into the noise removal module. The noise removal module is adapted to infer the optimal denoising parameters, see step S4. Different adaptation methods are applicable, particularly training time-series data filters to generate optimized filter parameters, such as the weights of the corresponding machine learning model, or applying signal processing techniques for adaptive noise cancellation. The denoised motor current signal Id is determined by inputting the initially measured motor current signal Im into the adapted noise removal module, see S5. The resulting denoised motor current signal Id is output by the adapted noise removal module, see S6 for optimizing motor diagnostics.
[0049] Preferably, the denoised motor current signal is input to a motor current analyzer that outputs motor diagnostic results optimized for reliability and focusing on motor current components caused by motor-specific problems rather than motor current components originating from fluctuations in the power grid. The motor current analyzer is a motor current feature analyzer (MCSA) or a machine learning model trained to analyze motor current signals, preferably a deep neural network, self-organizing map, or decision tree.
[0050] MCSA is a non-invasive technology that does not require direct access to the motor's interior. It involves measuring the motor's stator current and analyzing its spectrum for any abnormal patterns or deviations. MCSA is widely used in industrial applications for condition-based monitoring and predictive maintenance of motors because it provides an effective way to continuously monitor motor health and plan maintenance accordingly.
[0051] The applied motor model can be implemented using a deterministic simulation model, preferably a high-fidelity finite element model or a high-fidelity digital twin. The applied motor model can also be implemented using a trained data-based model or a combination of a deterministic simulation model of the motor and a data-driven model.
[0052] Electronic motor models are commonly used for circuit analysis, control system design, and predicting the overall electrical performance of motors. An electronic model typically refers to a circuit-based representation of the electrical components and behavior of a motor. This type of model focuses on the electrical characteristics and performance of the motor, such as voltage, current, resistance, inductance, and torque output. Electronic motor models require less processing power, but their outputs are often less accurate.
[0053] In contrast, the finite element model of an electric motor is a more comprehensive, physical-based representation of its structure and operation. The finite element model breaks down the motor into small, discrete components and applies numerical analysis to simulate the complex physical interactions within the motor, including electromagnetic fields, structural mechanics, and thermal effects. Therefore, the finite element model provides a detailed multi-physics simulation of the motor, allowing for more accurate predictions of performance, efficiency, and reliability compared to simpler electronic models.
[0054] Figure 2 An auxiliary device 10 of the present invention is shown, comprising related objects as inputs and outputs to its data source. The measured motor voltage signal Um and the corresponding measured motor current signal Im are sampled by sensors 12 and 13 located at the motor 11 (e.g., at the respective motor terminals). A voltage source 14 generating the reference current signal Up is preferably located separately from the motor 10 and its power grid. In an embodiment, the voltage source 14 is located within the auxiliary device. Furthermore, the output reference voltage signal Up exhibits a perfect sinusoidal signal path with a frequency equal to the motor voltage input frequency, for example, 50 Hz.
[0055] The auxiliary device 10 includes a data input interface (not explicitly depicted) configured to receive voltage and current signals. The auxiliary device 10 includes a motor model unit 15, which includes at least one processor configured to execute a motor model. Furthermore, the auxiliary device 10 includes a noise removal unit 16, which includes at least one processor configured to perform the functions of a noise removal module.
[0056] The noise removal unit 16 receives the modeled current signal Ims and the reference current signal Ips from the motor model unit 15. Additionally, the noise removal unit 16 receives the measured motor current signal Im from the motor 10. The noise removal unit 16 is configured to execute the noise removal module. The noise removal unit 16 determines the set of denoising parameters based on the differential signal between the generated modeled current signal Ims and the generated reference current signal Ips.
[0057] The auxiliary device 10 includes a data output interface (not explicitly depicted) configured to output the denoised motor signal Id received from the noise removal unit 16 to the motor current analyzer 17.
[0058] Figure 3 The illustration shows an auxiliary device 30 having a noise removal unit 22 configured as a time-series filter. The noise removal unit 22 includes a time-series filter 24 to be trained and a trained time-series filter 26.
[0059] The difference 25 between the modeled current signal Ims and the reference current signal Ips represents the difference in current caused by signal components other than the basic motor voltage frequency. The difference 25 between the two current signals Ims and Ips refers to the difference or mismatch between two corresponding current waveforms or values in the motor system.
[0060] The time-series filter 24, applied to the current signal Im for modeling, is trained with the difference 25 as input, aiming to reduce the difference 25 to near zero. Training the time-series filter produces denoising parameters w, which are weights of the underlying model. The trained time-series filter 26 is configured with the denoising parameters w. In the final step, the trained time-series filter 26 is applied to the measured current signal Im recorded at motor terminal 11 to remove noise components, thereby generating a denoised current signal Id. The denoised current signal Id can be used for motor current characteristic analysis (MCSA) to detect motor problems. The time-series data filter is constructed as a linear autoregressive model, a Wiener filter, a kernel adaptive filter, or a recurrent neural network.
[0061] Motor model 21 is a model of motor 10 in a "perfect" state, which means that no misalignment or imbalance is parameterized to ensure that the signal components indicating motor problems are not removed.
[0062] Compared to existing solutions, the described method allows for the calculation of motor current using both low-fidelity and high-fidelity finite element motor models. The high-fidelity finite element motor model typically represents a more accurate model of the motor. This is important because the motor model is used to estimate precise discrepancies, i.e., the current noise signal. If this noise term is not accurate enough, it can lead to the removal of important signal components from the measured motor current Im, which contain information about possible motor problems such as misalignment and imbalance. Compared to more general techniques that do not use a motor model at all, this process ensures that only signal components arising from interference in the power grid are removed.
[0063] Figure 4 A detailed embodiment of the auxiliary device 30 of the present invention is schematically illustrated, the auxiliary device 30 having a noise removal unit 32 utilizing a noise cancellation function.
[0064] The auxiliary device 30 includes a motor model unit 31 constructed similarly to the motor model unit 21 that performs the motor model. The noise removal unit 32 of the auxiliary device 30 is configured to perform the functions of a noise removal module. The noise removal unit 32 incorporates two novel key features.
[0065] First, the noise extractor unit 34 infers the optimal denoising parameters In and configuration based on the modeled current signal Ims generated by the motor model in the motor model unit 31 and the reference current signal Ips. Second, the adaptive noise cancellation unit 35 performs signal denoising on the measured motor current signal Im based on the reference noise signal In provided by the noise extractor unit 34. In this way, the quality of the measured motor current signal Im is enhanced, which facilitates subsequent fault detection using MCSA or a suitable machine learning model located in the motor current analyzer 33.
[0066] The measured motor current signal Im is denoised using adaptive noise cancellation (a signal processing technique that filters out unwanted signal components given an appropriate reference noise signal). An example adaptive noise cancellation method is described in B. Widrow et al., "Adaptive noise cancelling: Principles and applications," in Proceedings of the IEEE, vol. 63, no. 12, pp. 1692-1716, Dec. 1975, doi: 10.1109 / PROC.1975.10036. It is assumed that the reference noise signal In is expected to be uncorrelated with the signal of interest (i.e., the undisturbed motor current Io), but correlated with the noise to be reduced originating from a disturbance on the measured input voltage Um.
[0067] To obtain this suitable reference noise signal In, the noise extractor unit 34 receives the modeled current signal Ims and the reference current signal Ips as input from the motor model unit 31, and estimates the noise caused purely by voltage fluctuations. The motor model unit 31 is configured to execute a motor model, which is constructed as a high-fidelity digital twin or low-fidelity model simulating the motor 11, a high-fidelity finite element model, or a trained data-based model, or a combination of both.
[0068] The reference noise signal In is determined by calculating the difference between the modeled motor current signal Ims and the reference current signal Ips. Optionally, this difference, i.e., the reference noise signal In, is further refined by filtering out any frequency components typically observed in real motor signals, and therefore this difference should not be considered noise because it is assumed to be uncorrelated with the signal of interest. Such typically observed frequency components are, for example, odd harmonics of the fundamental power supply frequency, primarily 50 Hz, also known as winding harmonics.
[0069] exist Figure 5 Exemplary results of the proposed method and auxiliary device 30 are presented in the paper.
[0070] The measured motor current signal Im is depicted as a solid line, with its time-domain curve shown in the top plot illustrating current versus time, and its frequency representation shown in the lower plot illustrating signal height versus frequency. The spectrum reveals suspicious peaks at approximately 20, 30, 40, 60, 70, and 80 Hz caused by voltage fluctuations triggered by other machines operating on the same power grid. If the motor is operating in isolation, these peaks are not present in the signal Io, see the signal depicted by the dashed line. An enhanced version of this signal (i.e., the denoised signal Id obtained by the described method) is shown as a dotted line. It mitigates artifacts caused by voltage fluctuations because the corresponding peaks in the spectrum disappear after filtering, see the marked portion in the lower plot.
[0071] The presented method allows for denoising of signals that cannot be described as additive signal components, resulting in significantly more accurate results. Furthermore, when multiple similar motors are connected to the same power supply network and exposed to the same interference, a filter trained on one motor can be applied to the remaining motors to remove the signal for subsequent Motor Current Characteristic Analysis (MCSA).
Claims
1. A computer-implemented method for optimizing motor diagnostics by removing stray signal components from a measured motor current signal of a motor (11), comprising the following steps: Receive (S1) the measured motor voltage signal (Um) and the measured motor current signal (Im) measured at the motor (11), The motor model generates a modeled current signal (Ims) based on the measured motor voltage signal (Um) (S2). The motor model generates a reference current signal (Ips) (S3) based on the undisturbed sinusoidal motor voltage signal (Up). Its features The noise removal module is adapted (S4) by inputting the modeled current signal (Ims) and the reference current signal (Ips) into the noise removal module to infer the optimal denoising parameters. The measured motor current signal (Im) is input into the adapted noise removal module to determine (S5) the denoised motor current signal (Id), and The noise-reduced motor current signal (Id) output (S6) is used to optimize motor diagnostics.
2. The computer-implemented method according to claim 1, wherein the denoised motor current signal (Id) is input to a motor current analyzer (17), and the motor current analyzer (17) outputs motor diagnostic results.
3. The computer-implemented method according to any one of the preceding claims, wherein the sinusoidal motor voltage signal (Up) is generated to have a frequency equal to the frequency of the supplied motor voltage signal.
4. The computer-implemented method according to any one of the preceding claims, wherein, The motor model is a deterministic simulation model that simulates the motor (11), preferably a high-fidelity finite element model, a high-fidelity digital twin, or a trained data-based model, or a combination of both.
5. The computer-implemented method according to claim 2, wherein the motor current analyzer (17) is a motor current feature analyzer (MCSA) or a machine learning model trained to analyze the motor current signal, preferably a deep neural network, self-organizing map, or decision tree.
6. The computer-implemented method according to any one of the preceding claims, wherein the denoising parameter set (w, In) is determined by the noise removal module based on the differential signal between the generated modeled current signal (Ims) and the generated reference current signal (Ips).
7. The computer-implemented method of claim 6, wherein the set of denoising parameters (w) is determined by optimizing the time series data filter to minimize the difference (25) by inputting the difference (25) between the modeled current signal (Ims) and the reference current signal (Ips) into the time series data filter (24).
8. The computer-implemented method according to claim 7, wherein the measured motor current (Im) is input to a trained time series data filter (26), the trained time series data filter (26) outputting the denoised motor current signal (Id).
9. The computer-implemented method according to any one of claims 7-8, wherein the time series data filter (24, 26) is constructed as a linear autoregressive model, a Wiener filter, a kernel adaptive filter, or a recurrent neural network.
10. The computer-implemented method according to any one of claims 1-6, wherein the noise removal module includes a noise cancellation unit (35) that filters out unwanted stray signal components of the measured motor current signal (Im) based on a reference noise signal (In).
11. The computer-implemented method of claim 10, wherein the reference noise signal (In) is generated by inputting the modeled current signal (Ims) and the reference current signal (Ips) into a noise extractor unit (34) and estimating the reference noise signal (In) caused only by voltage fluctuations.
12. The computer-implemented method of claim 11, wherein the reference noise signal (In) is estimated by calculating the difference between the simulated motor current (Ims) and the reference current signal (Ips).
13. The computer-implemented method according to claim 12, wherein, The estimated reference noise signal (In) is refined by filtering out frequency components typically observed in the measured motor signal, such as the fundamental power supply frequency or winding harmonics.
14. An auxiliary device including at least one processor, said at least one processor being configured to perform a step for removing stray signal components from a measured motor current signal of the motor (11) to optimize motor diagnostics, comprising the following steps: Receive the measured motor voltage signal (Um) and the measured motor current signal (Im) measured at the motor (11). The motor model generates a modeled current signal (Ims) based on the measured motor voltage signal (Um). The motor model generates a reference current signal (Ips) based on the undisturbed sinusoidal motor voltage signal (Up). Its features The noise removal module is adapted by inputting the modeled current signal (Ims) and the reference current signal (Ips) into the noise removal module to infer the optimal denoising parameters. The denoised motor current signal (Id) is determined by inputting the measured motor current signal (Im) into an adapted noise removal module. The noise-reduced motor current signal (Id) is output to optimize motor diagnostics.
15. A computer program product, directly loaded into the internal memory of at least one digital computer, including a software code portion for performing the steps of claim 1 when the product is run on the at least one digital computer.