Fault detection in a drive system based on fault-specific modulation patterns in the motion signal

Abstract

A method for determining a mechanical fault in a drive (AN) incorporating an electric machine (EM) involves capturing a signal (S) that represents the mechanical motion of the drive. The frequency components in the signal (S) are then determined. A specific mechanical fault (F) is identified based on a modulation pattern (M1 - M3) within these frequency components. Furthermore, a corresponding computer program product and a fault detection device (FD) are described.

Description

[0001] It is known to equip a drive system, in particular a vehicle drive system (traction drive), with an electric machine as a drive component. The electric machine can be connected to an output shaft via a transmission, with several bearings rotatably supporting the relevant shafts. Particularly when the drive system is designed as a hybrid drive, i.e., with an internal combustion engine as a further drive component, a mechanical transmission path results, which, together with the drive components, comprises numerous parts.

[0002] To monitor these components, acoustic analysis can be used, as faults or impending increased wear in a drive system become audibly apparent. German patent application DE 10 2017 204 941 A1 describes how, to monitor the function of vehicle components, the relevant noise is recorded and assigned to a corresponding fault condition. A corresponding noise database serves as the basis for this assignment. The document describes how the recorded noise is compared with the noise database, for example, using a spectral comparison, i.e., in the frequency domain.

[0003] The publication JP 2011 - 257 362 A describes how to identify the condition of a bearing based on a group of peaks in the spectral representation of a motor current signal.

[0004] The publication DE 11 2022 007 789 T5 describes a condition diagnosis for an electric motor, wherein peak intensities of sideband waves due to a belt drive frequency are extracted from peak intensities of sideband waves and a comparison compares the extracted peak intensities with a peak intensity threshold.

[0005] Representing and comparing signals in the frequency domain requires highly precise signal representation to avoid misattributions, which leads to high computational effort and large memory consumption. The challenge is to demonstrate a method for precisely detecting errors in a signal with less computational effort and memory consumption.

[0006] This problem is solved by the subject matter of claim 1. Further properties, features, embodiments and advantages become apparent with the dependent claims, the description and the figures.

[0007] It is proposed to use modulation patterns in a signal error for specific identification. Modulation patterns can be readily identified and specifically classified through spectral analysis, as their characteristics (such as maxima / minima of certain frequency components, especially when grouped) offer a simple and specific method of identification. Here, the modulation patterns and their groupings are defined by the ratios of the signal strengths of spurious frequency components to each other and / or to the main harmonic component, as well as (alternatively or in combination) by the frequency or order intervals between the spurious frequency component and the main harmonic component, or between the spurious frequency components themselves. Specific modulation patterns, or...Groupings can also be such that the signal strength of the main harmonic component is smaller than at least one of the associated spurious frequency components, or is not present due to displacement to the position of at least one spurious frequency component.

[0008] On the other hand, modulation patterns originate from specific mechanical defects that are characteristic of the modulation pattern. This allows for the selective identification of specific mechanical defects that lead to modulation patterns, even using simple methods (with low data resolution, etc.). Modulation patterns are characterized in particular by groupings of at least two (relative) optima (maxima or minima) of specific frequency components. Specifically, modulation patterns are characterized by a main frequency, i.e., a signal strength that is strongly frequency-dependent, and at least one sideband generated by the modulation of the main component. In particular, modulation creates sidelobes or secondary frequency components at specific intervals (left and right in the spectrum). In addition to the main component, modulation creates secondary components that are shifted relative to the main component by the modulation frequency (or multiples thereof).The main component here is, in particular, a harmonic component, such as a harmonic component of a fundamental frequency, which can, for example, correspond to a rotational speed of the drive (e.g., the electric machine). The main component can therefore be, in particular, a principal harmonic or principal frequency component with a specific principal frequency.

[0009] This section focuses in particular on the modulation of a harmonic component, which can be referred to as the principal harmonic component. Furthermore, it considers modulation patterns related to a fundamental frequency, i.e., modulation patterns characterized by (at least) one secondary frequency component resulting from the modulation of the principal component (related to a fundamental frequency). Therefore, modulation patterns are considered to consist of at least two frequency components, comprising either a secondary frequency component and another secondary frequency component (related to the same principal frequency component) or a principal frequency component. In particular, the signal strength of the components, the ratio of the signal strengths of the components, and the spectral separation (i.e., the difference in signal strength between the components) are considered.the frequency difference) between the components, the width of the components (especially if it is a peak) and / or other spectral properties of the components.

[0010] Not only the absolute position of the main lobe is specific to errors, but also, and especially, the properties of the secondary lobes in relation to it, i.e., their relative strength to each other, their relative strength to the main lobe, the distance (i.e., frequency difference) between the secondary lobes and the main lobes, etc. The position of the main lobe can often be derived from normal operation and the associated movements of the drive. The position of the secondary lobes, on the other hand, indicates deviations from these movements, whereby the deviation is considered a modulation or modulation pattern.The term "tip" here refers to a frequency-specific maximum (or an inverted version thereof) and its spectral profile, analogous to the angular dependence of antenna gains, where instead of local, strongly angular-dependent gain maxima in antennas, the local maxima (or minima) of the strongly frequency-dependent signal strengths (for certain frequencies) are meant, which arise from modulation (i.e., frequency shifting) or resonance.

[0011] The procedure described here involves spectrally analyzing a signal generated by the movement in a drive system for modulation patterns. To perform this spectral analysis, the frequency components of the signal are determined, i.e., the signal strength (amplitude, power, etc.) for different frequencies or frequency bands. Modulation patterns are then identified in the spectral representation of the signal (i.e., based on the frequency components), allowing conclusions to be drawn about the corresponding, specific fault. This approach exploits the fact that various mechanical faults lead to modulation. A clear example is bearing damage, which, when a shaft rotates, causes wobble, resulting in a specific modulation pattern as a modulation of the rotational motion. Numerous other examples and different types of faults or modulation patterns are discussed below.

[0012] A corresponding method for determining a mechanical fault in a drive is proposed, wherein the drive is equipped with an electric machine. A signal is acquired that represents the mechanical motion of the drive, in particular a rotational motion within the drive. The mechanical motion is specifically the result of the active operation of the drive. The mechanical motion originates from movement in the mechanical power path, which leads from a drive component (electric machine, possibly an internal combustion engine) to an output of the drive. The signal can be acquired as an acoustic signal and converted into an electrical signal by an electroacoustic transducer, or it can be acquired as an electrical signal, for example by induction, where the motion leads to an electrical signal in a conductor. The signal is acquired as a time-domain signal.The next step in the spectral analysis of the signal is the conversion of the signal into a representation in the frequency domain.

[0013] Based on the acquired signal, frequency components present in the signal are determined. For this purpose, a representation of the signal in the frequency domain is generated. The conversion of the acquired signal (time domain) into frequency components can be achieved using a Fourier transform, such as an FFT, a wavelet transform, or similar. Determining the frequency components can lead to a representation of the signal in which the signal strengths (amplitudes) of the individual frequencies are displayed discretely.

[0014] At least one specific mechanical error is determined based on a modulation pattern in the frequency components thus obtained. For this purpose, the frequency components (or the signal represented in the frequency domain) can be examined with regard to predefined modulation patterns. Each modulation pattern is associated with a specific error that generates the respective modulation pattern. A mapping (data collection, such as a lookup table) can be provided that assigns characteristics of different modulation patterns to specific errors. These characteristics relate to the distance between secondary and primary components, the distance (i.e., frequency difference) between secondary components, the ratio of the signal strengths of secondary components (of the same primary component) to each other, or even the ratio of the signal strength of a secondary component to the signal strength of the corresponding primary component. The secondary components are the result of a modulation of the primary component.The specific modulation pattern is determined by the mechanical fault and is therefore specifically linked to that fault. Different mechanical faults generate different modulation patterns, allowing the triggering fault to be identified based on these specific patterns. Different mechanical faults are detected based on different modulation patterns.

[0015] The modulation patterns can each be defined or stored in the form of a specific set of features (as mentioned herein). Determining the modulation pattern can be accomplished by comparing the features (of the modulation pattern of the detected signal with features of those modulation patterns associated with specific errors). A similarity measure can be calculated that reflects the similarity of the features, and the specific error is identified if the similarity measure exceeds a predefined threshold. Based on the examples in the Fig. This is explained in more detail in sections 2a-c. The frequency components of the measured signal represent (if errors are present) the modulation pattern of the recorded signal. Comparing the frequency components of the measured signal with the modulation patterns associated with specific mechanical faults serves to identify the mechanical fault or to determine that no mechanical fault is present (if no modulation pattern is present in the frequency components or if it is recognizable with sufficient similarity). The specific modulation patterns can be identified using pattern recognition or by means of analytical or statistical evaluation. Furthermore, the specific modulation patterns can be identified using machine learning, in particular using a neural network. This can be trained using faulty drives and their signals.Frequency components, where, within the framework of machine learning, the fault associated with the faulty drive is linked to the relevant modulation patterns. Input vectors of such a mapping are frequency components characterized by their amplitude (height) and frequency (or frequency spacing), where one vector represents two or more frequency components.

[0016] The modulation pattern can be determined from the ratio of the amplitudes of secondary frequency components to each other. This applies particularly to the secondary frequency components on either side of a primary harmonic component. The primary harmonic component is specific to an electrical machine. The modulation pattern can also be determined from the ratio of the signal amplitude of the primary harmonic component to the signal amplitude of a secondary frequency component. This ratio can also be zero, for example, if the primary harmonic component has a signal amplitude of essentially zero, perhaps because the modulation (caused by the specific error) shifted the primary harmonic component to the position of the secondary component, or split it and shifted it to the positions of the secondary components.

[0017] The main harmonic component can be p times the rotational speed, where p is the number of poles of the electric machine. The main harmonic component can also be an integer multiple of p times the rotational speed, where p is the number of poles of the electric machine. The two secondary frequency components on either side of a main harmonic component represent, in particular, a positive-sequence system (the secondary frequency component above the main harmonic component) and a negative-sequence system (the secondary frequency component below the main harmonic component). If the two secondary frequency components are unequal, i.e., if the magnitude of their difference exceeds a predetermined threshold (for example, if the strength of the positive-sequence system is greater than the strength of the negative-sequence system by a minimum value, or vice versa), then...This can indicate a rotation-direction-dependent fault, where a deflection or oscillation in the direction of rotation of the electric machine has different effects (different stiffness or different tendency to oscillate) than in the opposite direction of rotation. Furthermore, an increased primary harmonic component (compared to a fault-free drive) can indicate torsional resonance, which may be related to a reduced belt tension in a belt drive. Specifically, an increased primary harmonic component compared to the secondary component (i.e., a ratio of the signal strength of the primary harmonic component to the signal strength of the secondary component) can point to such a belt tension (where the aforementioned ratio is higher than that observed in the fault-free case).

[0018] In particular, the distance between the secondary frequency components, or between one of the secondary frequency components and the main harmonic component, allows conclusions to be drawn about the order of the secondary frequency components relative to the main harmonic component, with different orders (distances from the main harmonic component) corresponding to different faults. For example, secondary frequency components with a signal strength above a given threshold value, which are one order away from the main harmonic component, can be attributed to a non-fully axial (concentric) rotational movement of the rotor (relative to the stator) as a fault. In other words, this symptom (i.e., secondary frequency components shifted by one order relative to the main harmonic component and exceeding a given signal strength threshold) can be attributed to the fault "radial bearing fault" or "eccentric rotor rotation".

[0019] As a further example, secondary frequency components with a signal strength above a predefined threshold, lying two orders away from the main harmonic component, can be attributed to rotor deflection (insufficient axial rotor stiffness), which is considered a defect. In other words, this symptom (i.e., secondary frequency components shifted two orders from the main harmonic component and exceeding a given signal strength threshold) can be associated with the defect "insufficient rotor stiffness" or "reduced radial stiffness." The defect of insufficient rotor stiffness can correspond to the defect "rotor laminations not fully pressed" or the defect "rotor lamination stacks not fully pressed."

[0020] A radial bearing fault can therefore be identified as a defect if secondary frequency components are separated by no more than a first frequency difference, and in particular if they are separated by two orders (and thus each is one order away from the main harmonic component). A different fault can be identified as reduced radial stiffness of the electric machine rotor if secondary frequency components are separated by more than the first frequency difference, for example, if they are twice the first frequency difference, such as if they are four orders apart or each is two orders away from the main harmonic component. The first frequency difference can correspond to one order (by which the secondary frequency component is separated from the main harmonic component due to modulation by the fault).The first frequency difference can correspond to the level of the fundamental frequency.

[0021] The frequency of the main harmonic component is n times the fundamental frequency gf. The fundamental frequency is, in particular, the mechanical speed of the electrical machine. The main harmonic component is obtained by multiplying this by the number of poles p of the electrical machine or an integer, positive multiple thereof. This applies especially to synchronous machines. For asynchronous machines, the slip must also be taken into account, or the signal must be related to the rotor motion (i.e., the rotor speed, which differs by the slip from the electrical speed, which relates to the stator field, must be used as the basis). For the secondary frequency components, a distance of m orders from the main harmonic component corresponds to a frequency-related distance of ± m*gf. Preferably, the placeholder m is a natural number and can be 1 or greater.The main harmonic component is, in particular, the p-th harmonic; the secondary harmonic components are then the (pm)-th harmonic and the (p+m)-th harmonic. If the main harmonic component is, for example, the 48th harmonic (with pole number p = 48), then the two first-order (m-th order) secondary harmonic components are located at the (48-1)-th harmonic and the (48+1)-th harmonic (i.e., at the (48-m)-th and the (48+m)-th harmonic). Different intervals between the main harmonic component and secondary harmonic components (i.e., different orders of the secondary harmonic components) can be associated with different errors. A specific error exists, in particular, when the corresponding frequency component has a signal strength above a threshold value.

[0022] Since the modulation pattern can be determined based on the ratio of the amplitudes of secondary frequency components (located on either side of the main harmonic component), different errors are identified for corresponding main harmonic components of different frequencies. In particular, different errors are identified for secondary frequency components of different frequencies. Thus, not only can the ratio of the amplitudes of secondary frequency components be associated with a specific error, but also the absolute (frequency) position of the secondary frequency components can be specific to a particular error—that is, the position or frequency of the main harmonic component.Thus, if a specific ratio is established, different errors can be detected for different frequencies of the main harmonic component, to which the ratio of the associated secondary frequency components refers, especially even with the same ratio.

[0023] The acquired signal (in spectral terms) can be used unchanged to determine the frequency components. Preferably, however, the acquired signal is filtered. The filtering can be performed with one or more filter spectra, each adapted to at least one modulation pattern. If, for example, certain modulation patterns are characterized by components (or their signal strength) at specific frequencies, then the filter spectrum can be configured to allow these frequencies to pass through with less attenuation than other frequencies. The filtering can therefore be selective for frequencies corresponding to the minor or major / minor components described here, and which are associated with specific errors. Several filters can be provided, adapted to different modulation patterns and thus specific to different errors.In this way, multiple filtering operations can be performed, each adapted to a specific modulation pattern associated with a particular error. If a modulation pattern is characterized by certain frequencies or frequency bands, then the corresponding filtering is configured to allow these frequencies or frequency bands to pass through while blocking others (or applying a higher attenuation / lower gain than the aforementioned frequencies or frequency bands). The filtering can be implemented, in particular, using an electronic filter whose pass frequencies are specific to a particular modulation pattern. The filter can be parameterized differently for different modulation patterns to account for the different characteristic frequency components or frequencies.Different modulation patterns may be provided for different errors, whereby the detection of a specific error involves filtering differently during the prior determination of the frequency components or acquisition of the signal, the filtering being specific to the modulation pattern in question (and allowing the frequencies characterizing the modulation pattern to pass through as opposed to others).

[0024] Another aspect is the assignment of specific faults to associated, individual frequency differences between the fundamental harmonic component and its associated (at least one) secondary frequency component (or between two secondary frequency components of the same fundamental harmonic component). These frequency differences can be not only an integer multiple of the fundamental frequency, but can also be odd. This is particularly relevant when the method relates to a drive system that, in addition to the electric motor, includes a gearbox (belt drive, gear drive, etc.) in the transmission path. When motion is transmitted via the gearbox, faults in the gearbox or connected transmission paths or mechanical units can be detected by the resulting modulation pattern that characterizes the gearbox's transmission ratio.Especially when the turns ratio is not even, the resulting spurious frequency components can be readily identified, as these components have a spacing from each other or from the main harmonic component that does not correspond to any harmonic order (i.e., is not even-numbered), but rather reflects the odd turns ratio. The fault-specific modulation pattern can thus be determined from the frequency spacing of at least one spurious frequency component to a main harmonic component specific to the electrical machine (or from the spacing of two spurious frequency components that are symmetrical to the main harmonic component). The associated fault can be a fault in a mechanical unit (bearing failure of this unit or some other fault of this unit).Such a unit is connected to the electric machine via the transmission, so that a fault in this unit affects the electric machine via the transmission (with overdrive or underdrive). Such a mechanical unit can be, in particular, an internal combustion engine, for example, if the drive is a hybrid drive. The associated fault can also be a fault in the transmission, such as bearing damage, especially on the side facing away from the electric machine. Furthermore, it can be a fault in the transmission path between the electric machine and the mechanical unit, in particular a bearing failure in the mechanical transmission path.These errors can be associated with a modulation pattern characterized by a frequency difference (between secondary frequency components or between secondary frequency component and a main harmonic component specific to the electric machine) that characterizes the gear ratio of the transmission.

[0025] Further embodiments provide that a specific modulation pattern can be determined based on the frequency difference between at least one secondary frequency component and a primary harmonic component specific to the electric machine. In this case, the fault represents a fault in a power transmission component to which the electric machine is connected, if the frequency difference characterizes the circumferential speed of the power transmission component. The power transmission component is, in particular, a belt or a gear to which the electric machine is connected. The fault symptom is thus a frequency difference between the secondary frequency component and the primary harmonic component, which corresponds to the circumferential speed of a power transmission component to which the electric machine is connected for motion transmission.The fault can be, in particular, a mechanical fault at a point in the power transmission component, such as a defect in a belt or gear (as examples of power transmission components). The power transmission component can be part of the gearbox. The power transmission component is, in particular, located within a mechanical transmission path in the drive system.

[0026] Particularly when using a belt as a power transmission component, the fault can be indicative of insufficient mechanical tension in the power transmission component. This occurs when the frequency difference represents the circumferential speed of the power transmission component (especially the belt). This is especially true if, in addition, the off-frequency component in the acquired signal is larger than the off-frequency component in a signal acquired at an earlier time. This change compared to the signal acquired at an earlier time indicates belt aging. An aging parameter can be provided, which is higher the greater the difference between the signal strengths of the off-frequency components at the earlier and the current (acquisition) time.

[0027] Furthermore, it is conceivable that the modulation pattern (i.e., error symptom in the signal) is a pattern characteristic of a resonant mechanical oscillation with a displacement force that depends non-linearly (e.g., quadratically) on the displacement. Here, non-linear dependence is preferably defined as an exponential dependence with an exponent > 1. In particular, the modulation pattern can be a pattern that results from a magnetic displacement with the non-linear, and especially quadratic, ratio of the generated force to the distance characteristic of magnetic force. For example, the modulation pattern can be a pattern with a displacement force that depends non-linearly on the displacement and with a restoring force that depends linearly or proportionally on the displacement. If such a modulation pattern (which is characteristic of a displacement force with a non-linear (e.g., quadratic) dependence) is detected, the signal can be further analyzed.If the modulation pattern is determined to be characterized by a quadratic dependence on the displacement, then a bearing fault (radial bearing fault) in the electric machine is identified as the fault. If the modulation pattern exhibits a non-linear relationship between displacement and displacement force, in particular an exponential relationship with a power > 1, as is the case with a magnetic displacement force, then at least one of the secondary frequency components is increased compared to a linear or proportional relationship. Such a relationship exists, in particular, in an air gap between the rotor and stator (especially in synchronous motors), if a bearing fault allows radial displacements of the rotor, resulting in a magnetically induced displacement force that depends exponentially on the displacement. The air gap thus amplifies the fault-induced displacement. The displacement here is specifically a radial displacement.A modulation pattern characterized by mechanical (especially radial) vibration with a displacement force that depends non-linearly on the displacement indicates that a (defective) radial bearing in the electric machine allows greater radial movement than a faultless radial bearing (of the same design). This results in wobbling movements of the rotor relative to the stator, or an air gap between the stator and rotor that is not constant in size and therefore depends on the angle.

[0028] The signal can be a structure-borne sound signal, an airborne sound signal, an electrical signal induced in the stator, a harmonic compensation signal from a motor control system of the electric machine, or a signal obtained through dynamic distance measurement (laser distance detection). Signal acquisition can thus involve using a microphone or a structure-borne sound transducer to record an airborne or structure-borne sound signal emanating from the drive or the electric machine. Alternatively, signal acquisition can involve scanning a part of the drive, such as the stator and / or the rotor, using a laser distance meter (as a device for dynamic distance measurement) to obtain the signal.Furthermore, the signal can be acquired by detecting an electrical signal at the stator windings or an excitation winding of the electric machine, such as a current signal at the phase terminals (especially if the stator windings are short-circuited) or a voltage signal at the phase terminals. Finally, an (internal) signal from a motor control system with active harmonic suppression can be used. Such a motor control system features harmonic detection and compensation signal generation. Within a motor control system, such as a vector control system, the harmonic detection system identifies harmonics of an actual value, manipulated variable, or controlled variable of the motor control system. The compensation signal generation system then creates a compensation signal that is added to (or subtracted from) the manipulated variable of the control system.The compensation signal is designed to reduce harmonics through at least partial compensation, similar to an active noise cancellation (ANC) method. The signal used here can be derived from the output signal of the harmonic detection system or from signals derived from it, such as a signal from the compensation signal generation (which also reproduces harmonics). Therefore, the signal can be derived from the (active) harmonic suppression of a corresponding motor control system for the electric machine.

[0029] The method described here can be used in particular for determining aging; for example, the strength of the relevant frequency component, at least one of which is decisive for identifying the fault, can be used as a measure of aging. In this way, in addition to the fault itself, the aging of the relevant component exhibiting the detected fault can be determined. It may be possible to output the aging and / or the detected fault as a single output signal.

[0030] The procedure described here can be implemented, in particular, using computer code (or simply code) that runs on a processor. Therefore, a computer program product is described that includes code suitable for executing the procedure described here when executed on a (programmable) processor. The code may include an interface for transmitting the signal or electronically processable signals that represent the acquired signal. Furthermore, the code may include a section configured to calculate a spectral representation from the signal (representation in the time domain), for example, using an FFT. This allows the frequency components to be determined, since the spectral representation represents the signal in the frequency domain. The code may also include a further section configured to determine a modulation pattern in the spectral representation (or in the frequency components).Pattern recognition or analytical / statistical evaluation can be used for this purpose. The code also includes a section that assigns individual, specific errors to different modulation patterns. The errors can be identified by the faulty or aged components exhibiting the errors, and alternatively or additionally by a type of movement enabled by the error, which mechanically characterizes the error. An interface for transmitting the detected error, or a signal indicating the error, can be used to transfer the detected error, for example, to another entity or to another code (such as application software or a harness).

[0031] The procedure described here can also be implemented using a fault detection device designed to execute the described method. The fault detection device has an input configured to perform the signal detection step itself, or configured to receive a data signal that reproduces the detected signal, for example, when a unit not belonging to the device detects the signal and outputs it (in a processed form) to the input. The fault detection device also has a data processing module. This module is connected downstream of the input. The data processing module is configured to identify the fault in question based on the modulation pattern. In particular, the data processing module is configured to generate a spectral representation of the signal or to determine the frequency components present in the signal. The device also has an output.This is connected downstream of the data processing module. The output is configured to output the specific error associated with the modulation pattern.

[0032] The Fig. Figure 1 shows a drive to illustrate the approach described here.

[0033] The Fig. 2a - 2c serve as an example to explain modulation patterns, their properties, and the associated errors.

[0034] The Fig. Figure 1 shows a drive unit AN comprising an electric machine EM, a gearbox G, and a mechanical unit designed as an internal combustion engine VM. The electric machine and the internal combustion engine are connected to each other via the gearbox G for the transmission of motion. This can serve to jointly generate motion or torque for an output (not shown). Three bearings L1, L2, and L3 are shown as examples, which rotatably support the respective shafts. The electric machine EM is connected to the gearbox G, and the drive unit G is connected to the internal combustion engine rotor VM. The gearbox G can, in particular, be a belt drive.

[0035] The Fig. Figure 1 symbolically shows that a signal S is generated by the drive AN, specifically an acoustic signal S. This signal S is detected by a microphone M. The microphone converts the acoustic signal es into a data signal DS. This data signal DS is output to an input E of a fault detection device FE. The fault detection device FE has a data processing module DV. This module converts the data signal DS, and thus also the signal S, into frequency components, determines the frequency components in the respective signal (S or DS), and identifies a specific mechanical fault F based on a modulation pattern in the frequency components. An output A of the fault detection device FE outputs this specific fault (in the form of an output signal). Thus, the fault detection device FE is configured to perform the steps of the procedure described here.

[0036] The Fig. Figure 2a shows an example of frequency components in the signal S. In addition to the main harmonic component of the 24th harmonic, there is another frequency component (with a corresponding minimum signal strength) at the 24.8 harmonic. The two relevant signal maxima are readily identifiable and labeled as modulation pattern M1. The distance between the two maxima does not correspond to an integer multiple of the fundamental frequency on which the 24th harmonic is based. Rather, this distance corresponds to the gear ratio of the gearbox G. Therefore, from this distance, which corresponds to an essential characteristic of the gearbox G, it can be concluded that the observed fault has an effect on the mechanical unit VM, which is connected to the electric machine EM via the gearbox G.In the illustrated case, an uneven rotational motion of the rotor of the electric machine EM is transmitted to the internal combustion engine VM via the gearbox G due to the eccentric rotation of the motor. This uneven rotational motion results in an uneven load torque in the internal combustion engine VM, which the (inactive) internal combustion engine exerts and which is transmitted back to the electric machine EM via the gearbox G. In the illustrated case, the uneven rotational motion leads to an uneven drag torque in the internal combustion engine and thus to a gas pressure which, due to the valve position in the internal combustion engine, generates the uneven load torque. This gas pressure is transmitted via the gearbox and leads to the secondary component (24,8 th) in the signal S. Based on the modulation pattern M1 of the... Fig. In 2a, which has an odd-numbered secondary component (24.8th) in addition to the main component (24th), a radial bearing failure in bearing L1 (between gearbox G and electric machine EM) can thus be identified as a defect. In an alternative embodiment, the pattern extends not only over the main component (24th) and the first (odd-numbered) secondary component (24.8th), but also over the second (odd-numbered) secondary component (25.7th), which is twice as far from the main component as the first secondary component, in the sense of a second secondary lobe.

[0037] Even the Fig. Figure 2b shows a main harmonic component of the 24th harmonic (24th) and further frequency components (23.3th, 25.7th) that are separated from the main harmonic component by an odd multiple of the fundamental frequency. Therefore, the error in the pattern can also be deduced from these components. Fig. 2a is the basis. However, it can also be seen that the aforementioned components are two-part, or each exhibits two peaks, separated by a small frequency (0.15th). This feature originates from a power transmission component of the gearbox G, such as a belt, whose rotational speed corresponds to the small frequency. This power transmission component becomes visible in the signal due to a radial bearing fault in the electric machine EM, whereby the radial bearing fault is transmitted via the gearbox and thus via the power transmission component of gearbox G. The eccentric rotation of the rotor of the electric machine (caused by the radial bearing fault in bearing L1, i.e., the rotor bearing of the electric machine) generates an additional movement (vibration) in the power transmission component of gearbox G, to which the electric machine EM is connected, which can be recognized as modulation pattern M2 or M2'.The fact that two adjacent frequency components are separated by a factor of 0.15 of the fundamental frequency indicates the presence of a fault affecting the belt's operation. Furthermore, the presence of a secondary component (23.3th) in addition to the primary harmonic (24th), which is not separated by a factor of 0.15 but by another odd multiple, leads to the conclusion that the operation of the mechanical unit VM is also affected by the fault. This suggests an eccentric rotation of the electric machine's rotor, caused by a radial bearing fault in bearing L1.

[0038] The Fig. Figure 2c shows a modulation pattern M3 with a main harmonic component (48th) and two (even-order-separated) secondary components to the left (46th, 47th) and right (49th, 50th) of the main harmonic component (48th). The two secondary components (47th, 49th), separated by only one order, define a modulation pattern that can be attributed to eccentric motor operation and thus to a radial bearing fault in bearing L1, which rotatably supports the rotor. If this is recognizable from the modulation pattern, either a bearing fault of the electrical machine can be reported as an error (if the signal strength of the secondary component exceeds a critical threshold), or (if the signal strength of the secondary component does not exceed the critical threshold but is greater than a threshold value) instead of an error, a warning can be issued that the bearing of the electrical machine should be checked ("predictive maintenance").

[0039] The two minor components (46th, 50th), separated by more than one order, define a modulation pattern that can be attributed to reduced (and therefore faulty) rotor stiffness or axial rotor deflection. If these two minor components exceed a threshold, a rotor fault can be generated, indicating incompletely (and therefore faultily) pressed rotor laminations.

[0040] Finally, it should be mentioned that Fig. 2c represents modulation patterns around the 48th harmonic, while the Fig. 2a and Fig. 2b. Represent modulation patterns around the 24th harmonic. Due to the position of the modulation pattern, a fault in the rotor of the electric machine ( Fig. 2c) are detected, or a fault in a bearing of the drive or in the electric machine ( Fig. 2a, b).

Claims

[1] Method for determining a mechanical fault of a drive (AN) with an electric machine (EM), comprising the steps: Capturing a signal (S) that represents the mechanical movement of the drive; Determining the frequency components in the signal (S); Determining a specific mechanical fault (F) based on a modulation pattern (M1 - M3) in the frequency components, whereby the modulation pattern is determined - based on the ratio of the strengths of secondary frequency components to each other, - by the ratios of the signal strengths of side frequency components to the main harmonic component or - by the frequency or order intervals between the spurious frequency components. [2] Method according to claim 1, wherein different mechanical defects (F) are detected for different modulation patterns. [3] Method according to claim 1 or 2, wherein the modulation pattern is determined based on the ratio of the strengths of secondary frequency components to each other, wherein the secondary frequency components are located on both sides of a main harmonic component specific to the electrical machine. [4] Method according to claim 3, wherein the specific mechanical error is determined based on the frequency position of the main harmonic component. [5] Method according to claim 3 or 4, wherein the specific mechanical error is determined based on the frequency difference between the secondary frequency components on the one hand and the main harmonic component on the other. [6] Method according to claim 3, 4 or 5, wherein a radial bearing fault is determined as a fault if secondary frequency components are not more than a first frequency difference apart, and a reduced radial stiffness of the rotor of the electric machine is determined as a fault if secondary frequency components are more than the first frequency difference apart. [7] Method according to one of the preceding claims, wherein the modulation pattern is determined based on the frequency difference of at least one secondary frequency component to a main harmonic component specific to the electric machine, wherein the fault is determined to be a fault in a mechanical unit (VM) connected to the electric machine (EM) via a gearbox (G), a fault in the gearbox (G), or a bearing fault in the mechanical transmission path between the electric machine and the mechanical unit, where the frequency difference characterizes the gear ratio of the gearbox. [8] Method according to one of the preceding claims, wherein the modulation pattern is determined based on the frequency difference of at least one secondary frequency component to a main harmonic component specific to the electrical machine, wherein the fault represents a fault in a power transmission part to which the electrical machine is connected, if the frequency difference characterizes the circumferential speed of the power transmission part. [9] Method according to claim 8, wherein the fault represents insufficient mechanical stress in the power transmission part when the frequency difference characterizes the circumferential speed of the power transmission part (belt) and when the off-frequency component in the detected signal is greater than the off-frequency component in a signal detected at an earlier time. [10] Method according to one of the preceding claims, wherein the fault is a bearing damage in the bearing of the rotor of the electric machine, if the modulation pattern is characteristic of a resonant mechanical vibration with a displacement force that depends non-linearly on the displacement. [11] Method according to one of the preceding claims, wherein the signal is a structure-borne sound signal, an airborne sound signal, an electrical signal induced in the stator, a harmonic compensation signal of a motor control of the electric machine or a signal obtained by dynamic distance measurement (laser). [12] Computer program product comprising code capable of performing the method according to any of the preceding claims when executed on a programmable processor. [13] Fault detection device (FE) configured to perform the method according to any one of claims 1-11, wherein the fault detection device (FE) has an input (E) configured to perform the detection step or configured to receive a data signal (DS) that reproduces the detected signal, a data processing module (DV) connected downstream of the input (E) configured to perform the detection step, and an output (A) connected downstream of the data processing module configured to output the specific fault (F).

Images