Method and system for diagnosing mechanical faults of a variable pitch system based on motor current characteristics

Abstract

The embodiment of the application provides a kind of mechanical fault diagnosis method and system of variable pitch system based on motor current characteristics, it obtains the current and rotating speed signal of motor first, and not directly analyze original time domain current, but utilize rotating speed signal to carry out instantaneous rotation angle calculation and equal angle increment resampling to q-axis time domain current, with the non-stationary current signal that changes with time is converted into stationary angle domain current signal that changes with rotation angle.This overcomes the problem of fault feature frequency drift and spectrum dispersion caused by motor speed fluctuation.To further separate weak fault impact feature from the angle domain signal, adaptive signal deconstruction is carried out using variational mode decomposition, and the mode component most sensitive to fault impact is selected using kurtosis maximization criterion, and then selective envelope demodulation is carried out on the optimal mode.Finally, the angle domain envelope signal obtained can generate characteristic clear, order fixed envelope order spectrum through Fourier transform, so as to realize the analysis and diagnosis of fault.

Description

Technical Field

[0001] This invention relates to the field of pitch system fault location technology, specifically to a method and system for diagnosing mechanical faults in pitch systems based on motor current characteristics. Background Technology

[0002] The pitch control system of a wind turbine generator set is a key actuator for power control and safe shutdown. This system operates under complex conditions such as variable speed, heavy load, and strong vibration for extended periods. Internal components like gearboxes and bearings are prone to wear, cracks, and broken teeth. Failure to detect and address these early-stage faults in a timely manner can lead to pitch control failure, potentially causing overspeeding or even more serious equipment accidents, resulting in significant economic losses.

[0003] In existing technologies, methods based on motor current characteristic analysis have attracted attention because they do not require additional sensor installations. The principle is that mechanical faults cause periodic fluctuations in the motor load, which in turn modulate the stator current; faults can be identified by analyzing the spectrum of the current signal. However, pitch systems are essentially servo control systems, and their operating state depends entirely on wind conditions and control commands, resulting in continuous dynamic changes in motor speed and load—a typical non-stationary operating condition. Traditional fault diagnosis methods, such as Fourier transform-based spectrum analysis, are theoretically based on the assumption of signal stationarity. When such methods are directly applied to non-stationary current signals, the constantly changing speed causes the fault characteristic frequencies to drift and broaden in the spectrum, producing a severe spectral dispersion effect. This effect causes weak early fault characteristic signals to be submerged in broadened fundamental frequency energy and strong background noise, reducing the sensitivity and accuracy of the diagnostic method and making it difficult to meet the needs of field applications.

[0004] Therefore, an optimized mechanical fault diagnosis scheme for pitch systems is desired. Summary of the Invention

[0005] The present invention aims to at least solve one of the technical problems existing in the prior art, and provides a method and system for diagnosing mechanical faults in a pitch system based on motor current characteristics.

[0006] In a first aspect, embodiments of the present invention provide a method for diagnosing mechanical faults in a pitch system based on motor current characteristics, comprising: Acquire the three-phase raw current signal and motor speed signal; The three-phase raw current signal and motor speed signal are transformed to obtain the filtered q-axis time-domain current and synchronous speed signals; Instantaneous rotation angle calculation and constant angle incremental resampling are performed on the filtered q-axis time-domain current and synchronous speed signal to obtain the q-axis angle-domain current; Angle domain envelope signal is extracted from the q-axis angle domain current to obtain the angle domain envelope signal; The envelope order spectrum is obtained by calculating the envelope order spectrum of the angle domain envelope signal; Order spectrum analysis of the envelope order spectrum is performed to obtain a fault diagnosis report.

[0007] Secondly, embodiments of the present invention provide a mechanical fault diagnosis system for a pitch system based on motor current characteristics, comprising: The signal acquisition module is used to acquire the three-phase raw current signal and the motor speed signal; The signal conversion module is used to perform signal conversion on the three-phase raw current signal and the motor speed signal to obtain the filtered q-axis time-domain current and synchronous speed signals; The instantaneous rotation angle calculation and equal angle incremental resampling module is used to perform instantaneous rotation angle calculation and equal angle incremental resampling on the filtered q-axis time-domain current and synchronous speed signal to obtain the q-axis angle-domain current; The envelope signal extraction module is used to extract the angular domain envelope signal from the q-axis angular domain current to obtain the angular domain envelope signal. The envelope order spectrum calculation module is used to calculate the envelope order spectrum of the angle domain envelope signal to obtain the envelope order spectrum; The order spectrum analysis module is used to perform order spectrum analysis on the envelope order spectrum to obtain a fault diagnosis report.

[0008] Compared with existing technologies, this invention first acquires the current and speed signals of the motor. Instead of directly analyzing the original time-domain current, it uses the speed signal to perform instantaneous angle calculation and constant-angle incremental resampling on the q-axis time-domain current, converting the time-varying non-stationary current signal into a stationary angle-domain current signal that varies with the rotation angle. This overcomes the problems of fault characteristic frequency drift and spectral dispersion caused by motor speed fluctuations. To further separate weak fault impact features from this angle-domain signal, variational mode decomposition is used for adaptive signal deconstruction. The kurtosis maximization criterion is used to select the mode component most sensitive to fault impact, and then selective envelope demodulation is performed on this optimal mode. The resulting angle-domain envelope signal, after Fourier transform, generates a clearly defined and fixed-order envelope order spectrum, thereby achieving accurate fault analysis and diagnosis. Attached Figure Description

[0009] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0010] Figure 1 This is a flowchart of a mechanical fault diagnosis method for a pitch system based on motor current characteristics according to an embodiment of the present invention; Figure 2 This is a data flow diagram of a pitch system mechanical fault diagnosis method based on motor current characteristics according to an embodiment of the present invention; Figure 3 This is a flowchart illustrating the process of transforming three-phase raw current signals and motor speed signals to obtain filtered q-axis time-domain current and synchronous speed signals in a pitch system mechanical fault diagnosis method based on motor current characteristics according to an embodiment of the present invention. Figure 4 This is a flowchart illustrating the extraction of the angle domain envelope signal from the q-axis angle domain current using a pitch system mechanical fault diagnosis method based on motor current characteristics, according to an embodiment of the present invention. Figure 5 This is a flowchart illustrating the process of performing order spectrum analysis on the envelope order spectrum to obtain a fault diagnosis report in a pitch system mechanical fault diagnosis method based on motor current characteristics according to an embodiment of the present invention. Figure 6 This is a block diagram of a pitch system mechanical fault diagnosis system based on motor current characteristics according to an embodiment of the present invention. Detailed Implementation

[0011] To enable those skilled in the art to better understand the technical solutions of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the described embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0012] Unless otherwise specifically stated, the technical or scientific terms used in the embodiments of this invention should be understood in their ordinary meaning as understood by one of ordinary skill in the art to which this invention pertains. The terms "comprising" or "including," as used in the embodiments of this invention, do not limit the shapes, numbers, steps, actions, operations, components, elements, and / or groups thereof mentioned, nor do they exclude the appearance or addition of one or more other different shapes, numbers, steps, actions, operations, components, elements, and / or groups thereof, or the inclusion of these.

[0013] Unless otherwise specifically stated, the relative arrangement, numerical expressions, and values ​​of the components and steps described in these embodiments do not limit the scope of the invention. It should also be understood that, for ease of description, the dimensions of the various parts shown in the drawings are not drawn to actual scale, and techniques, methods, and apparatus known to those skilled in the art may not be discussed in detail; however, where appropriate, the illustrated techniques, methods, and apparatus should be considered part of the specification. In all the examples shown and discussed herein, any other specific example may have different values. It should be noted that similar symbols and letters in the following figures denote similar items; therefore, once an item is defined in one figure, it need not be further discussed in subsequent figures.

[0014] In the description of the embodiments of the present invention, the terms "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., refer to specific features, structures, materials, or characteristics described in connection with that embodiment or example, which are included in at least one embodiment or example of the present invention. In the embodiments of the present invention, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described may be combined in a suitable manner in any one or more embodiments or examples. Furthermore, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in the embodiments of the present invention, as well as the features of different embodiments or examples.

[0015] Hereinafter, exemplary embodiments according to the present invention will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of the present invention, and not all embodiments of the present invention. It should be understood that the present invention is not limited to the exemplary embodiments described herein.

[0016] Existing methods for diagnosing mechanical faults in pitch systems based on motor current characteristics face significant challenges in practical applications. This is primarily because pitch systems operate under typical non-stationary conditions, where drastic changes in motor speed and load cause continuous drift in the fault characteristic frequencies of the current signal. Traditional time-domain signal analysis methods suffer from severe spectral dispersion, causing early, weak fault characteristics to be masked by strong background noise, leading to diagnostic failure. Therefore, this invention proposes a method for diagnosing mechanical faults in pitch systems based on motor current characteristics. First, it acquires the three-phase raw current signal and motor speed signal, and transforms these signals to obtain filtered q-axis time-domain current and synchronous speed signals. Next, it uses the synchronous speed signal to perform instantaneous angle calculation and constant-angle incremental resampling on the filtered q-axis time-domain current. This step reconstructs the non-stationary time-domain signal, measured in seconds and affected by speed variations, into a q-axis angle-domain current with constant order characteristics, measured in angles. Subsequently, to accurately extract the impact component carrying fault information from the angular domain current, this invention further performs adaptive signal deconstruction based on variational mode decomposition on the q-axis angular domain current to obtain a set of intrinsic mode functions. Then, using the kurtosis maximization criterion, the optimal fault-sensitive mode that best characterizes the fault impact characteristics is intelligently selected from the multiple decomposed modes. Based on this, selective envelope demodulation is performed only on this optimal mode to obtain an angular domain envelope signal with a high signal-to-noise ratio. Finally, the envelope order spectrum of this angular domain envelope signal is calculated to obtain the envelope order spectrum. Adaptive noise baseline modeling, fault mode matching, and diagnostic rule fusion inference are then performed on this spectrum to ultimately generate an accurate fault diagnosis report.

[0017] Figure 1 This is a flowchart of a mechanical fault diagnosis method for a pitch system based on motor current characteristics according to an embodiment of the present invention. Figure 2 This is a data flow diagram illustrating a pitch system mechanical fault diagnosis method based on motor current characteristics according to an embodiment of the present invention. Figure 1 and Figure 2 As shown, the mechanical fault diagnosis method for a pitch system based on motor current characteristics according to an embodiment of the present invention includes the following steps: S100, acquiring three-phase raw current signals and motor speed signals; S200, performing signal transformation on the three-phase raw current signals and motor speed signals to obtain filtered q-axis time-domain current and synchronous speed signals; S300, performing instantaneous angle calculation and equal-angle incremental resampling on the filtered q-axis time-domain current and synchronous speed signals to obtain q-axis angle-domain current; S400, extracting the angle-domain envelope signal from the q-axis angle-domain current to obtain the angle-domain envelope signal; S500, calculating the envelope order spectrum of the angle-domain envelope signal to obtain the envelope order spectrum; S600, performing order spectrum analysis on the envelope order spectrum to obtain a fault diagnosis report.

[0018] Specifically, in step S100, the three-phase raw current signal and the motor speed signal are acquired. It is understood that mechanical faults in the pitch system drivetrain cause periodic, minute fluctuations in the motor load, which directly modulate the three-phase stator current of the motor. Simultaneously, the operating characteristics of the pitch system dictate that its motor speed is dynamically changing, and this speed information is a crucial reference for interpreting the current signal under non-stationary operating conditions. Therefore, in the technical solution of this invention, the three-phase raw current signal and the motor speed signal are acquired to simultaneously capture the electrical signal carrying fault characteristic information and the mechanical signal characterizing the instantaneous motion state of the motor. This provides a complete and necessary data source for subsequent coordinate transformation, angle domain resampling, and fault order analysis.

[0019] More specifically, in a concrete example of the present invention, the process of acquiring the three-phase raw current signal and motor speed signal is as follows. First, a current sensing device, such as three Hall current sensors, is configured on the three-phase power cable supplying the pitch motor to measure the instantaneous current values ​​of phases A, B, and C, respectively. Second, the motor speed signal is acquired. This signal can be directly derived from the rotary encoder built into the pitch servo drive, which outputs high-resolution rotor position or speed data in real time. Subsequently, the analog output signals of the three Hall current sensors and the speed signal from the encoder are connected to a multi-channel synchronous data acquisition device. A unified sampling clock is set in the acquisition device to ensure data time synchronization. For example, the sampling frequency of the current signal is set to 20 kHz to capture possible high-frequency fault impact information, and the sampling frequency of the speed signal is set to 1 kHz. Finally, the acquisition device stores the digitized three-phase current time series data and speed time series data to form the dataset required for subsequent analysis.

[0020] Specifically, in step S200, the three-phase original current signal and motor speed signal are transformed to obtain filtered q-axis time-domain current and synchronous speed signals. It is understood that the original three-phase AC current signal waveform is complex, and the load fluctuations carrying fault information are directly related to the electromagnetic torque of the motor, while the q-axis current is a direct representation of the electromagnetic torque in a rotating coordinate system. Simultaneously, the instantaneous rotor angle required to perform this rotating coordinate system transformation must rely on speed data that is strictly synchronized with the current signal, and the transformed q-axis current still contains low-frequency motion trends unrelated to the fault. Therefore, in the technical solution of this invention, the three-phase original current signal and motor speed signal are further transformed to obtain filtered q-axis time-domain current and synchronous speed signals. This decouples the difficult-to-analyze three-phase time-domain signal and extracts the q-axis component most sensitive to mechanical load disturbances. Simultaneously, high-pass filtering suppresses low-frequency operating condition fluctuations unrelated to fault diagnosis and ensures that the speed data used for transformation, after upsampling preprocessing, is completely consistent with the current data in terms of time reference. In this way, a filtered q-axis time-domain current with improved signal-to-noise ratio containing only dynamic load fluctuation information can be generated, along with a matching high-resolution synchronous speed signal. Together, these provide the necessary and clean data foundation for subsequent instantaneous angle calculation and high-precision angle domain resampling.

[0021] Figure 3 This is a flowchart illustrating the signal transformation of the three-phase raw current signal and motor speed signal to obtain filtered q-axis time-domain current and synchronous speed signals, according to an embodiment of the mechanical fault diagnosis method for a pitch system based on motor current characteristics. Figure 3 As shown, step S200 includes: S210, upsampling and preprocessing the motor speed signal to obtain a synchronous speed signal; S220, determining the instantaneous electrical angle of the motor rotor based on the synchronous speed signal; S230, performing a rotating coordinate system transformation based on the instantaneous rotation angle on the three-phase original current signal based on the instantaneous electrical angle of the motor rotor to obtain the unfiltered q-axis time-domain current; S240, performing a high-pass filter on the unfiltered q-axis time-domain current to obtain the filtered q-axis time-domain current.

[0022] In step S210, the motor speed signal is upsampled and preprocessed to obtain a synchronous speed signal. It is understood that during data acquisition, the sampling frequency of the three-phase raw current signal is much higher than that of the motor speed signal, resulting in a resolution mismatch between the two signals in terms of time reference. The subsequent rotating coordinate system transformation requires a precisely corresponding instantaneous speed or angle value at each sampling point of the current signal. Therefore, in the technical solution of this invention, the motor speed signal is further upsampled and preprocessed to obtain a synchronous speed signal, thereby interpolating and reconstructing the low-sampling-rate speed data into a time series with the same time resolution and data length as the high-sampling-rate current signal. This ensures that the instantaneous electrical angle used during coordinate transformation is strictly aligned with the current sampling point, providing a data foundation for subsequent high-precision signal transformation and instantaneous angle calculation.

[0023] More specifically, in a specific example of the present invention, the implementation process of upsampling preprocessing for the motor speed signal is as follows: First, the reference synchronization frequency is determined to be the sampling frequency of the three-phase original current signal, i.e., 20 kHz. Simultaneously, the original motor speed signal and its sampling frequency, i.e., 1 kHz, are acquired. Then, a target time vector is constructed, with each time point of this vector corresponding one-to-one with the sampling time of the 20 kHz current signal. Next, a high-fidelity interpolation algorithm, such as cubic spline interpolation, is used to process the original motor speed signal data sequence. This algorithm uses the speed values ​​at the original 1 kHz sampling points to fit a smooth speed change curve, and calculates the instantaneous speed estimate corresponding to each 20 kHz sampling time point on the target time vector based on this curve. Finally, these interpolated speed estimates are combined to form a new speed signal sequence, which is a synchronous speed signal strictly synchronized with the high-sampling-rate current signal. The data length and sampling rate of this signal are consistent with the original three-phase current signal.

[0024] In step S220, the instantaneous electrical angle of the motor rotor is determined based on the synchronous speed signal. It is understood that the subsequent coordinate system transformation, i.e., the Park transformation, aims to convert the current component in the stationary coordinate system to a coordinate system that rotates synchronously with the rotor magnetic field. The angle reference of the rotation matrix required for this transformation is not a constant value, but must reflect the instantaneous electrical angle of the rotor magnetic field in space in real time. Therefore, in the technical solution of this invention, the instantaneous electrical angle of the motor rotor is further determined based on the synchronous speed signal. This allows the use of the motor's mechanical speed signal, which is strictly synchronized with the current, to accurately reconstruct the instantaneous electrical angle of the rotor magnetic field at each sampling point through integration and pole pair conversion. This generates a time series of instantaneous electrical angles that corresponds one-to-one with the current data points, providing an accurate angle reference that dynamically changes with non-stationary operating conditions for subsequent coordinate transformations.

[0025] More specifically, in a specific example of the present invention, the process of determining the instantaneous electrical angle of the motor rotor is as follows. First, the number of pole pairs P of the pitch motor is obtained from the design parameters or nameplate information of the motor. The number of pole pairs is an inherent physical constant characterizing the electromagnetic characteristics of the motor; for example, an 8-pole motor has 4 pole pairs P. Subsequently, based on the synchronous speed signal obtained in the previous step with a sampling frequency of 20 kHz... The instantaneous electrical angle of the motor rotor is determined by the following formula. :

[0026] in, for The instantaneous electrical angle of the motor rotor at any given moment. For synchronous speed signal, This represents the number of pole pairs of the motor. This formula indicates that the instantaneous electrical angle is the motor's electrical angular velocity, i.e. The integration over time. In the numerical computation of data processing, this integration process is performed as a discrete cumulative summation. Specifically, at a sampling frequency of 20 kHz, the time step is... For 1 / 20000 of a second, the first Instantaneous electrical angle at each sampling point pass Perform iterative calculations and set an initial angle. The value is 0. For example, when a pitch motor with 4 pole pairs performs a 30-degree mechanical pitch angle movement, This will reflect the instantaneous mechanical rotational speed of this process, which is obtained through the above integral calculation. Changes will accumulate The electrical angle is measured in degrees. Finally, this step outputs an instantaneous electrical angle time series that has the same 20 kHz sampling rate, the same data length, and is strictly synchronized with the current signal. This is for use in subsequent Park transformations.

[0027] In step S230, based on the instantaneous electrical angle of the motor rotor, a rotating coordinate system transformation based on the instantaneous rotation angle is performed on the three-phase original current signal to obtain the unfiltered q-axis time-domain current. It is understood that since the originally acquired three-phase current signals are mutually coupled AC quantities, it is difficult to directly identify weak load disturbances caused by mechanical faults from them. In the dq rotating coordinate system of the motor, the q-axis current component is proportional to the electromagnetic torque of the motor and is the physical quantity most sensitive to load fluctuations. Therefore, in the technical solution of this invention, a rotating coordinate system transformation based on the instantaneous electrical angle of the motor rotor is further performed on the three-phase original current signal to obtain the unfiltered q-axis time-domain current. This projects the current signal in the three-phase stationary coordinate system onto a coordinate system that rotates synchronously with the rotor, and decouples and separates the q-axis current component directly corresponding to the electromagnetic torque. In this way, the mechanical fault characteristic information contained in the complex three-phase AC signal can be concentrated into a single, physically meaningful time-domain current signal. This lays the foundation for subsequent feature extraction.

[0028] More specifically, in a specific example of the present invention, the implementation process of this rotating coordinate system transformation includes two consecutive calculation steps: Clarke transformation and Park transformation. First, after acquiring the three-phase raw current signal with a sampling rate of 20 kHz... , , Then, the Clarke transformation was applied to transfer it from the three-phase stationary coordinate system. Transform to two-phase stationary coordinate system Two-phase orthogonal current components are obtained. and The formula for calculating this transformation is:

[0029] in, The resulting vector represents the three-phase raw current signal. Subsequently, the instantaneous electrical angle of the motor rotor, calculated in the previous step and synchronized with the current, is used... The Park transform is applied to convert the two-phase quadrature current components. and Transform to a two-phase coordinate system that rotates synchronously with the rotor Below, the direct-axis current is obtained. and cross-axis current The formula for calculating this transformation is:

[0030] In this transformation, The instantaneous electrical angle of the motor rotor. This represents the unfiltered q-axis time-domain current. The resulting vector represents the original three-phase current signals. This step performs the matrix operation once at each 20 kHz sampling point, ultimately selecting the quadrature-axis current component. As an output, during the operation of the pitch system, if a tooth surface of the transmission gearbox experiences partial spalling, a momentary load impact will occur when that tooth surface engages in meshing. This impact will cause a momentary fluctuation in the motor's electromagnetic torque, which, after transformation, manifests as... A tiny perturbation in the signal amplitude.

[0031] In step S240, the unfiltered q-axis time-domain current is high-pass filtered to obtain the filtered q-axis time-domain current. It is understood that the unfiltered q-axis time-domain current signal contains both high-frequency resonant components excited by faults such as gear cracks or bearing pitting, and low-frequency trend terms and DC bias caused by macroscopic movements of the pitch system, such as start-stop, acceleration, and deceleration. The amplitudes of these low-frequency components are much larger than the weak fault characteristic signal, which will interfere with it. Therefore, in the technical solution of this invention, the unfiltered q-axis time-domain current is further high-pass filtered to obtain the filtered q-axis time-domain current, thereby setting a cutoff frequency to filter out low-frequency operating condition fluctuation components and DC bias below this frequency. This effectively improves the signal-to-noise ratio of high-frequency dynamic fault characteristics in the signal, allowing subsequent angle-domain resampling and envelope analysis to focus on the key frequency band carrying fault information.

[0032] More specifically, in a specific example of the present invention, the process of high-pass filtering the unfiltered q-axis time-domain current is as follows: First, determine the type and parameters of a digital high-pass filter, for example, selecting a 4th-order Butterworth digital high-pass filter. Second, set the cutoff frequency of the filter. The selection of this cutoff frequency aims to isolate the macroscopic motion frequencies of the pitch system from the high-frequency resonant bands excited by fault impacts. For example, the main motion energy of the pitch system is concentrated below 5Hz, while early fault characteristics of gears or bearings may manifest in the resonant band of several hundred hertz; therefore, the cutoff frequency can be chosen accordingly. The frequency was set to 5Hz. Finally, this fourth-order Butterworth high-pass filter was applied to the unfiltered q-axis time-domain current sampled at a frequency of 20 kHz. Sequence. Through filtering algorithms, such as recursive computation of difference equations, the sequence... Each data point in the sequence is processed, and a filtered q-axis time-domain current is finally output. The output sequence effectively attenuates low-frequency trend terms and DC bias below 5Hz, while dynamic components above 5Hz that may contain fault information are preserved.

[0033] Specifically, in step S300, the filtered q-axis time-domain current and synchronous speed signal are subjected to instantaneous angle calculation and equal-angle incremental resampling to obtain the q-axis angle-domain current. It is understood that since the filtered q-axis time-domain current is a non-stationary signal based on time, the frequency representation of periodic impacts caused by mechanical faults in the time-domain signal will drift during the dynamic change of the pitch motor speed, leading to a dispersion effect in subsequent spectrum analysis. Therefore, in the technical solution of this invention, the filtered q-axis time-domain current and synchronous speed signal are further subjected to instantaneous angle calculation and equal-angle incremental resampling to obtain the q-axis angle-domain current. This allows the accurate instantaneous mechanical angle of the motor rotor at each sampling moment to be calculated using real-time speed data, and the original time-domain current signal is interpolated and reconstructed based on this angle. In this way, the signal can be converted from the time domain to the angle domain, eliminating the influence of speed fluctuations on the fault characteristic order, transforming the non-stationary frequency analysis problem into a stationary order analysis problem, and providing a stable q-axis angle domain current signal synchronized with the rotation angle for subsequent envelope analysis and order spectrum calculation.

[0034] More specifically, in a specific example of the present invention, the implementation process of performing instantaneous rotation angle calculation and equal-angle incremental resampling is as follows. First, based on the input synchronous rotation speed signal with a sampling frequency of 20 kHz... The instantaneous mechanical rotation sequence is calculated through discrete integration, i.e., cumulative summation. This calculation can be expressed as:

[0035] in The sampling period is 20 kHz, which is 50 microseconds. Let be the instantaneous mechanical rotation angle at the k-th sampling time. Secondly, set a fixed angular resolution, for example, set the resolution per revolution of the motor rotor. Each sampling point is used to determine the angular increment. Radius. Based on this increment, a target angle mesh sequence is generated. ,in Index the sampling points in the angle domain. Finally, perform resampling. This will generate the instantaneous mechanical rotation angle sequence. As a non-uniform independent variable, the filtered q-axis time-domain current is used. As the dependent variable, a cubic spline interpolation algorithm is used in the target angle grid sequence. At various angles, Interpolation calculations are performed to obtain a new signal sequence. The new sequence This refers to the q-axis angular domain current, whose independent variable is a constant angular increment and is no longer affected by fluctuations in motor speed.

[0036] Specifically, in step S400, the q-axis angular domain current is subjected to angular domain envelope signal extraction to obtain the angular domain envelope signal. It is understandable that directly using the Hilbert transform to extract the envelope of the angular domain current signal has a fundamental technical flaw stemming from the contradiction between the theoretical limitations of the transform method itself and the complexity of the signal in practical application scenarios. Mathematically, the Hilbert transform requires the signal to be a narrow-band, single-component analytic signal; only under this premise can the extracted envelope possess clear physical meaning. However, in the specific scenario of mechanical fault diagnosis in a pitch system, the acquired angular domain current signal is not an ideal single-component signal, but a typical multimodal composite signal. Specifically, this signal simultaneously contains at least three modes with distinct physical origins: first, the baseline transmission mode generated by transmission chain components such as normal gear meshing; second, the electrical noise mode penetrating from the motor control system and electromagnetic environment; and third, the high-frequency inherent resonance mode of the structure excited by the periodic impact of the fault point when an early local fault occurs, which is also the target mode carrying key diagnostic information. Therefore, applying the Hilbert transform directly to such a composite signal without differentiation inevitably leads to mutual interference and aliasing of instantaneous amplitudes and frequencies between different modes. This results in a severely distorted envelope signal that fails to reflect the true impact of the fault. Especially when the fault is in its early stages and the corresponding resonant mode energy is weak, this feature will be completely submerged in the interference of other dominant modes, leading to diagnostic failure. Therefore, in the technical solution of this invention, the q-axis angular domain current is further extracted to obtain the angular domain envelope signal. This step specifically includes adaptive signal deconstruction of the q-axis angular domain current based on variational mode decomposition to obtain a set of intrinsic mode functions (EMFs). Then, the EMF set is optimized for fault-sensitive modes based on kurtosis maximization to obtain the optimal fault-sensitive mode. Finally, the optimal fault-sensitive mode is selectively envelope-demodulated to obtain the angular domain envelope signal. This achieves accurate amplification and extraction of weak fault impact characteristics by actively deconstructing complex signals and intelligently selecting fault-sensitive modes. In this way, a high signal-to-noise ratio angle domain envelope signal can be separated and reconstructed from the original multimodal aliasing signal. Its amplitude fluctuation can clearly reflect the variation law of fault impact energy with rotor angle, providing a reliable input for subsequent order spectrum analysis.

[0037] Figure 4This is a flowchart illustrating the extraction of the angle domain envelope signal from the q-axis angle domain current using a mechanical fault diagnosis method for a pitch system based on motor current characteristics, according to an embodiment of the present invention. Figure 4 As shown, step S400 further includes: S410, performing adaptive signal deconstruction based on variational mode decomposition on the q-axis angle domain current to obtain an intrinsic mode function set; S420, performing fault-sensitive mode optimization based on kurtosis maximization on the intrinsic mode function set to obtain an optimal fault-sensitive mode; S430, performing selective envelope demodulation on the optimal fault-sensitive mode to obtain the angle domain envelope signal.

[0038] In step S410, the q-axis angle domain current is subjected to adaptive signal deconstruction based on variational mode decomposition to obtain the set of intrinsic mode functions (EMFs). It is understood that the q-axis angle domain current is a multimodal composite signal, mixing normal baseline transmission modes, electrical noise modes, and fault impulse resonance modes carrying critical diagnostic information. Without separation, the mutual interference between these modes will cause weak early fault characteristics to be submerged. Therefore, in the technical solution of this invention, the q-axis angle domain current is further subjected to adaptive signal deconstruction based on variational mode decomposition to obtain the set of EMFs, thereby performing the deconstruction of the original multimodal composite signal and eliminating mode aliasing interference in subsequent envelope analysis from the root. This process uses a variational mode decomposition algorithm to regard the original angle domain current signal as a signal composed of the superposition of K finite bandwidth EMFs, and finds this optimal set of modal components by solving a constrained variational problem. In this way, an optimal set of decomposition bases can be adaptively found based on the data characteristics of the signal itself, achieving accurate sorting of the original signal in a physical sense. This effectively separates the aforementioned basic transmission mode, electrical noise mode, and fault impact resonance mode, producing a set of narrowband signals with relatively singular physical sources. This lays a solid foundation for accurate location of subsequent fault characteristics.

[0039] More specifically, in a specific example of the present invention, the implementation process of the adaptive signal decomposition is as follows. First, the key parameters of the variational mode decomposition algorithm are set, including the preset number of mode decompositions K and the quadratic penalty factor. For example, based on prior analysis of the signal complexity of the pitch system, K=5 is set to separate signal components from different physical sources. Secondly, a constrained variational problem is constructed, the core optimization objective of which is to minimize the sum of the estimated bandwidths of all modal components, while ensuring that the sum of all modal components can accurately reconstruct the original signal. Subsequently, the constrained variational problem is solved iteratively using the alternating direction multiplier method. During the iteration process, the modal components are updated alternately. , center angular frequency And Lagrange multipliers, until a convergence condition is met, such as the update magnitude of the modal components being less than a preset tolerance threshold. Finally, after iterative convergence, the algorithm outputs K eigenmode functions with finite bandwidth, that is, it produces a set of narrowband signals with relatively singular physical sources. This set is the set of intrinsic mode functions. For example, in an actual operating segment of a pitch system, let K=5. After processing by the VMD algorithm, the five modal components may correspond to: It manifests as low-order components (baseline transmission modes) related to the planetary carrier frequency. and It manifests as a narrowband signal (baseline transmission mode) related to the gear meshing order and its harmonics. It manifests as a broadband electrical noise component (electrical noise mode); while This manifests as an impact resonance signal with a high central angular frequency and an amplitude that fluctuates periodically with the angle. This signal is the potential fault impact resonance mode that needs to be locked in the subsequent steps.

[0040] In step S420, the intrinsic mode function set is optimized for fault-sensitive modes based on kurtosis maximization to obtain the optimal fault-sensitive mode. It is understood that after the adaptive deconstruction of the signal, an intrinsic mode function set containing the baseline mode, noise mode, and potential fault mode is obtained. The key next step is to establish an effective evaluation criterion to intelligently identify and filter out the mode dominated by fault impact from the multiple decomposed modal components. Early mechanical faults physically manifest as periodic impact sequences, and kurtosis, as a fourth-order statistic, is highly sensitive to the impact characteristics of the signal. Therefore, in the technical solution of this invention, the intrinsic mode function set is further optimized for fault-sensitive modes based on kurtosis maximization to obtain the optimal fault-sensitive mode, thereby improving the fault sensitivity of the intrinsic mode function set. Each modal component in the set The kurtosis value is calculated, and by comparing the kurtosis values ​​of all modes, the mode with the largest kurtosis value is selected as the optimal fault-sensitive mode. In this way, the physical characteristics of fault impact can be mapped onto kurtosis in a quantitative and automated manner, and the fault resonance mode that best characterizes the fault impact can be identified from many background modes, thus realizing the transformation from blind analysis to intelligent focusing.

[0041] More specifically, in a concrete example of this application, the preferred implementation process for the fault-sensitive mode is as follows. First, obtain the set of intrinsic mode functions obtained from the previous VMD decomposition step. ,in In this scenario, these five modal components physically correspond to: (Planetary carrier frequency mode) and (Gear meshing mode) (Electrical noise modes) and (High-frequency impact resonance mode). Next, iterate through each modal component in this set. And calculate its kurtosis value according to the following formula. :

[0042] in, Represents the expectation operation. It is the k-th modal component. It is the mean of that modal component. It is its standard deviation. In numerical calculations, and Through calculation respectively The sample mean and sample standard deviation of the sequence are obtained. This is obtained by calculating the fourth sample moment of the sequence's central moment. Corresponding to the above physical mode, the calculated kurtosis value might be: The stationary frequency transfer mode is close to a Gaussian distribution. and The meshing modes are close to a Gaussian distribution and have a stable meshing mode. (Broadband noise mode), and The resonant modes exhibiting significant impact characteristics were then compared, and the mode with the maximum kurtosis value was selected using the following formula:

[0043] The argmax operation here represents selecting the value that makes the kurtosis value... The largest modal component In the example above, because The kurtosis value of 18.5 is much larger than the kurtosis values ​​of all other modes (all close to 3), which clearly indicates that... The sequence contains strong non-Gaussian impact components, perfectly matching the statistical characteristics of the physical phenomenon of periodic impacts generated by bearing roller crushing cracks or tooth surface spalling points. Therefore, this mode... Automatically selected as the optimal fault-sensitive mode The output is then sent to the next step for envelope demodulation.

[0044] In step S430, selective envelope demodulation is performed on the optimal fault-sensitive mode to obtain the angle-domain envelope signal. It is understood that since the preceding steps have already obtained an approximately single-component optimal mode rich in fault impact information, the application prerequisites of the Hilbert transform are met, enabling accurate envelope extraction. This aims to overcome the inherent defects of traditional Hilbert envelope analysis methods when processing multimodal composite signals. Therefore, in the technical solution of this invention, selective envelope demodulation is further performed on the optimal fault-sensitive mode to obtain the angle-domain envelope signal, thereby ensuring that the powerful demodulation tool of the Hilbert transform is applied to its most suitable scenario, namely, processing a purified and filtered quasi-single-component signal. In this way, the weak fault impact characteristics in the original current signal, submerged in strong background noise, can be transformed into a significantly enhanced angle-domain envelope signal with a very high signal-to-noise ratio. The amplitude fluctuation of this signal can accurately and reliably reflect the variation law of fault impact energy with rotor angle, providing high-quality input for subsequent clear envelope order spectrum analysis and accurate fault diagnosis.

[0045] Specifically, in step S500, the envelope order spectrum of the angle domain envelope signal is calculated to obtain the envelope order spectrum. It is understood that since the angle domain envelope signal obtained in the previous step characterizes the fluctuation of fault impact energy in the angle domain, although its morphology reflects periodic impact, its precise periodic component, i.e., the fault characteristic order, is still implicit and cannot be directly used for quantitative diagnosis. Different mechanical faults, such as cracks in the inner and outer rings of a bearing or broken teeth in a gear, each correspond to a unique characteristic order, measured in the number of occurrences per revolution. Therefore, in the technical solution of this invention, the envelope order spectrum of the angle domain envelope signal is further calculated to obtain the envelope order spectrum, thereby transforming the signal from the angle domain to the order domain. In this way, the periodic impact component hidden in the angle domain signal can be resolved into spectral peaks clearly visible on the envelope order spectrum, located at specific fault characteristic order positions, thereby achieving accurate quantitative identification of the fault source.

[0046] More specifically, in a specific example of the present invention, the process of calculating the envelope order spectrum is as follows. First, the angle-domain envelope signal generated in the previous step, which is a real number sequence, is obtained. This signal sequence is based on equal-angle increments. Sampling, for example, per revolution contains The data consists of several data points, and the total data length covers multiple cycles of motor rotation. Subsequently, a Fast Fourier Transform (FFT) algorithm is applied to this angle-domain envelope signal sequence. This transform converts the signal from the angle domain... Transform to order domain Finally, the magnitude of the FFT transform result is calculated to obtain the envelope order spectrum. The horizontal axis of the spectrum represents the order. Its physical meaning is the number of cycles per revolution. For example, if the theoretical outer ring fault characteristic order of a bearing in a pitch system is 5.43, and the bearing has already experienced outer ring spalling, then the angle domain envelope signal obtained in the previous step... The impact pulses will exhibit periodicity in terms of angle. After this step of FFT calculation, the envelope order spectrum is obtained. On the x-axis The location and its harmonic order , At positions such as [position name], energy spectral peaks significantly higher than those in the background noise are observed.

[0047] Specifically, in step S600, order spectrum analysis is performed on the envelope order spectrum to obtain a fault diagnosis report. It is understandable that the envelope order spectrum calculated in the previous step is raw data containing background noise, order spectrum lines generated by normal mechanical operation, and order spectrum peaks representing potential fault characteristics. Without analysis, it is impossible to automatically and accurately determine whether a system fault has occurred and what type of fault has occurred. Therefore, in the technical solution of this invention, further order spectrum analysis is performed on the envelope order spectrum to obtain a fault diagnosis report. First, adaptive noise baseline modeling and significant order extraction are performed on the envelope order spectrum, using dynamic thresholds to distinguish statistically significant abnormal order spectrum peaks, forming a significant order set. Next, based on a pre-established theoretical fault order database containing theoretical fault orders of various components of the pitch system (such as bearing outer ring fault order, gear meshing order, etc.), fault mode matching is performed on this significant order set to generate a candidate fault set containing all possible fault types. Finally, diagnostic rules are applied to this candidate fault set for fusion reasoning, such as verifying the presence of supporting evidence like harmonics or sidebands, to make a final decision. In this way, complex numerical spectra can be intelligently translated into a clear and explicit fault diagnosis report, which directly identifies the faulty component and fault type, providing precise decision support for maintenance personnel.

[0048] Figure 5 This is a flowchart illustrating the process of performing order spectrum analysis on the envelope order spectrum to obtain a fault diagnosis report in a pitch system mechanical fault diagnosis method based on motor current characteristics, according to an embodiment of the present invention. Figure 5 As shown, step S600 further includes: S610, performing adaptive noise baseline modeling and significant order extraction on the envelope order spectrum to obtain a significant order set; S620, based on the theoretical fault order database, performing fault mode matching and candidate fault set generation on the significant order set to obtain a candidate fault set; S630, performing diagnostic rule fusion reasoning and final report generation on the significant order set and candidate fault set to obtain a fault diagnosis report.

[0049] In step S610, adaptive noise baseline modeling and significant order extraction are performed on the envelope order spectrum to obtain a significant order set. It is understood that the acquired envelope order spectrum contains complex background noise components, and the amplitude of this noise baseline is not constant across different order intervals. If a fixed global threshold is used to detect spectral peaks, it is difficult to simultaneously address false alarm suppression in high-noise regions and the sensitivity of detecting weak fault features in low-noise regions. Therefore, in the technical solution of this invention, adaptive noise baseline modeling and significant order extraction are further performed on the envelope order spectrum to obtain a significant order set. This first establishes a noise baseline model that dynamically changes with the order of the envelope order spectrum, and an adaptive significance threshold is calculated based on this model. Finally, only those spectral peaks whose energy amplitude exceeds their local dynamic threshold are extracted. This effectively overcomes the interference of non-uniform background noise distribution, accurately separates statistically significant anomalous order components from the spectrum, and forms a reliable significant order set, providing high-quality input data for subsequent accurate fault mode matching.

[0050] More specifically, in a concrete example of this application, the implementation process is as follows. First, the input envelope order spectrum is obtained. To establish a noise baseline model A moving median filter is used. The window width of this filter is set. There are 51 order points. Apply this filter to... For each order point Calculate its The median value within the width neighborhood is assigned as . This median filter, because it is insensitive to spectral peaks, can robustly estimate the background noise level of the spectrum. Subsequently, the significance threshold, which dynamically varies with order, is calculated. This threshold is defined as the noise baseline. Add an increment related to the level of local noise fluctuation. This increment is the standard deviation of the local noise. of times, It can be calculated In the same The standard deviation within the window is used to obtain the sensitivity coefficient. It can be set to 6. Therefore, the dynamic threshold... Finally, significant order extraction is performed. This involves traversing the envelope order spectrum. First, identify all local peak points. For each local peak point... and its amplitude Its amplitude is compared with the dynamic threshold corresponding to that point. Compare. If If so, the spectral peak is determined to be a significant spectral peak, and its order is determined. and amplitude Stored as a tuple in a set of significant orders For example, in order If its amplitude It is 0.8, while the noise baseline at that location is... Only 0.1, standard deviation If the value is 0.05, then the dynamic threshold is... .because The spectral peak was identified as a significant peak and extracted.

[0051] In step S620, based on the theoretical fault order database, fault mode matching and candidate fault set generation are performed on the significant order set to obtain a candidate fault set. It is understood that the significant order set extracted in the previous step is merely a set of statistically anomalous order values ​​and their amplitudes, and does not carry any diagnostic information about physical faults. However, each specific component in the pitch system drivetrain generates a unique theoretical fault characteristic order that can be pre-calculated using a mechanical dynamics model. Therefore, in the technical solution of this invention, further fault mode matching and candidate fault set generation are performed on the significant order set based on the theoretical fault order database to obtain a candidate fault set. This allows for the comparison and correlation of abnormal phenomena extracted from the data with fault causes derived from the mechanism model. In this way, a series of physically meaningless numerical spectral peaks can be translated into one or more specific candidate fault lists with physical component orientation, providing clear targets for subsequent accurate diagnostic decisions.

[0052] More specifically, in a concrete example of the present invention, the implementation process is as follows. First, a theoretical fault order database is constructed. This database pre-calculates and stores the fault characteristic orders of all key components based on the mechanical drawings and bearing model parameters of the pitch system gearbox. For example, the database stores entries such as: {Fault Name: Outer ring fault of the first-stage planetary carrier output bearing, Fault Order: 5.43}, {Fault Name: Crack on the sun gear tooth surface, Meshing Order: 24.0, Modulation Order: 0.25}. Second, a matching tolerance window is set. To account for computational model errors, manufacturing tolerances, and potential minor deviations in signal processing, a tolerance is set, for example... This means that during matching, a fault with a theoretical order of 5.43 will cause the search to fail. The salient order is determined within the range. Finally, matching and candidate set generation are performed. The salient order set obtained in the previous step is iterated over. Simultaneously, iterate through the theoretical fault order database. When in A significant order of 5.42 was found, which falls within the theoretical order of 5.43. Tolerance window If the condition is met, the match is considered successful. The fault mode is determined to be: primary planetary carrier output bearing outer ring fault, with a significant correlation order of: } is added to the candidate fault set as a diagnostic unit. Similarly, when significant orders 23.76 and 24.24 are found, they correspond to the theoretical meshing order 24.0. The tolerance is inconsistent, but the matching logic further checks whether it conforms to the sideband mode. Calculation reveals... These two theoretical sideband orders fall within the significant orders of 23.76 and 24.24, respectively. Within the tolerance window, the match is considered successful, and the fault mode is determined to be: sun gear tooth surface crack, with a significant correlation order of: Add to candidate fault set .

[0053] In step S630, diagnostic rule fusion reasoning and final report generation are performed on the significant order set and the candidate fault set to obtain a fault diagnosis report. It is understood that the candidate fault set generated in the previous step is only a preliminary result based on the main feature order or simple pattern matching, and may contain misjudgments due to harmonic interference or spectral leakage. A specific mechanical fault, in terms of its behavior on the order spectrum, such as the integrity of harmonic families and sideband clusters, has a specific mechanistic pattern and requires more complex logic for confirmation. Therefore, in the technical solution of this invention, diagnostic rule fusion reasoning and final report generation are further performed on the significant order set and the candidate fault set to obtain a fault diagnosis report. This allows for the searching of broader supporting evidence in the significant order set for each candidate fault in the candidate fault set, based on fault mechanism knowledge, such as the associated rules of harmonics and sidebands, and the quantification of diagnostic confidence by fusing this evidence. This enables refined identification of the preliminary matching results, eliminates false candidate faults, and ultimately outputs a fault diagnosis report with high reliability and clear conclusions.

[0054] More specifically, in a specific example of the present invention, the implementation process is as follows. First, obtain the following: {failure mode: first-stage planetary carrier output bearing outer ring failure, related significant order:} } and {failure mode: sun gear tooth surface crack, relevant significance order:} The candidate fault set was obtained. The set of significant orders also included spectral peaks. Next, diagnostic rules are applied to each candidate fault for fusion reasoning. For a fault in the outer ring of the output bearing of the first-stage planetary carrier, the diagnostic rule base requires checking its harmonic order. The system searches for its second harmonic (theoretical value) in the set of significant orders. The system successfully matched a significant order of 10.83 (within a 1% tolerance), providing strong supporting evidence. For the sun gear tooth surface crack, the diagnostic rule base required checking the meshing order 24.0 itself and multiple sidebands. The system did not find a spectral peak corresponding to 24.0 in the significant order set, and only had one pair of sidebands (23.76, 24.24), indicating insufficient supporting evidence. Subsequently, the confidence score for each candidate fault was calculated. The bearing outer ring fault, due to matching both the fundamental order and the second harmonic, had a confidence score of 0.85. The sun gear crack, lacking both the main meshing order and higher-order sidebands, had a confidence score of 0.35. Finally, a diagnostic decision threshold of 0.5 was set. Since the bearing outer ring fault's score of 0.85 was higher than the threshold, while the sun gear crack's score of 0.35 was lower, the system made a final decision and generated a fault diagnosis report. The report states: "Diagnostic conclusion: Fault in the outer ring of the first-stage planetary carrier output bearing. Evidence: Significant spectral peaks were found at the fault characteristic order 5.42 (amplitude 0.8) and its second harmonic order 10.83 (amplitude 0.4). Confidence level: High." In summary, the mechanical fault diagnosis method for pitch systems based on motor current characteristics according to embodiments of the present invention is explained. It first acquires the motor's current and speed signals, but does not directly analyze the original time-domain current. Instead, it uses the speed signal to perform instantaneous angle calculation and constant-angle incremental resampling on the q-axis time-domain current, converting the time-varying non-stationary current signal into a stationary angle-domain current signal that varies with the angle. This overcomes the problems of fault characteristic frequency drift and spectral dispersion caused by motor speed fluctuations. To further separate weak fault impact features from this angle-domain signal, variational mode decomposition is used for adaptive signal deconstruction, and the kurtosis maximization criterion is used to select the mode component most sensitive to fault impact. Then, selective envelope demodulation is performed on this optimal mode. The final angle-domain envelope signal, after Fourier transform, generates a clearly defined and fixed-order envelope order spectrum, thereby achieving accurate fault analysis and diagnosis.

[0055] Furthermore, a mechanical fault diagnosis system for pitch systems based on motor current characteristics is also provided.

[0056] Figure 6 This is a block diagram of a pitch system mechanical fault diagnosis system based on motor current characteristics according to an embodiment of the present invention. Figure 6As shown, the mechanical fault diagnosis system 100 for a pitch system based on motor current characteristics according to an embodiment of the present invention includes: a signal acquisition module 110 for acquiring three-phase raw current signals and motor speed signals; a signal transformation module 120 for performing signal transformation on the three-phase raw current signals and motor speed signals to obtain filtered q-axis time-domain current and synchronous speed signals; an instantaneous angle calculation and equal-angle incremental resampling module 130 for performing instantaneous angle calculation and equal-angle incremental resampling on the filtered q-axis time-domain current and synchronous speed signals to obtain q-axis angle-domain current; an envelope signal extraction module 140 for extracting the angle-domain envelope signal from the q-axis angle-domain current to obtain an angle-domain envelope signal; an envelope order spectrum calculation module 150 for calculating the envelope order spectrum of the angle-domain envelope signal to obtain an envelope order spectrum; and an order spectrum analysis module 160 for performing order spectrum analysis on the envelope order spectrum to obtain a fault diagnosis report.

[0057] Furthermore, the signal conversion module 120 includes: The upsampling preprocessing unit is used to perform upsampling preprocessing on the motor speed signal to obtain the synchronous speed signal; The rotor instantaneous electrical angle determination unit is used to determine the instantaneous electrical angle of the motor rotor based on the synchronous speed signal; The rotating coordinate system transformation unit is used to perform a rotating coordinate system transformation on the three-phase original current signal based on the instantaneous electrical angle of the motor rotor to obtain the unfiltered q-axis time-domain current. The high-pass filter unit is used to perform high-pass filtering on the unfiltered q-axis time-domain current to obtain the filtered q-axis time-domain current.

[0058] Furthermore, the envelope signal extraction module 140 includes: An adaptive signal deconstruction unit is used to perform adaptive signal deconstruction based on variational mode decomposition on the q-axis angular domain current to obtain the set of intrinsic mode functions; The fault-sensitive mode selection unit is used to perform fault-sensitive mode selection based on kurtosis maximization on the intrinsic mode function set to obtain the optimal fault-sensitive mode; A selective envelope demodulation unit is used to selectively demodulate the optimal fault-sensitive mode to obtain the angle domain envelope signal.

[0059] As described above, the pitch system mechanical fault diagnosis system 100 based on motor current characteristics according to embodiments of the present invention can be implemented in various types of computing devices or control units. For example, it can be deployed in the main controller of a wind turbine generator, a dedicated controller for the pitch system, or an industrial computer for wind farm-level monitoring. In one possible implementation, the pitch system mechanical fault diagnosis system 100 based on motor current characteristics according to embodiments of the present invention can be integrated into the computing device as a software module and / or a hardware module. For example, the pitch system mechanical fault diagnosis system 100 based on motor current characteristics can be a software module in the control firmware of the computing device or control unit, configured to perform equal-angle incremental resampling, variational mode decomposition, and envelope order spectrum calculation, or it can be a dedicated diagnostic algorithm program developed for the computing device or control unit. Of course, the pitch system mechanical fault diagnosis system 100 based on motor current characteristics can also be one of the many hardware modules of the computing device or control unit, for example, implemented as dedicated digital signal processor logic to efficiently perform the interpolation operations and fast Fourier transforms required for the resampling, or embedded in a field-programmable gate array circuit to process the iterative operations of variational mode decomposition in parallel, or as an application-specific integrated circuit.

[0060] It is understood that the above embodiments are merely exemplary implementations used to illustrate the principles of the present invention, and the present invention is not limited thereto. For those skilled in the art, various modifications and improvements can be made without departing from the spirit and essence of the present invention, and these modifications and improvements are also considered to be within the scope of protection of the present invention.

Claims

1. A method for diagnosing mechanical failure of a variable pitch system based on motor current signatures, the method comprising: include: Acquire the three-phase raw current signal and motor speed signal; The three-phase raw current signal and motor speed signal are transformed to obtain the filtered q-axis time-domain current and synchronous speed signals; Instantaneous rotation angle calculation and constant angle incremental resampling are performed on the filtered q-axis time-domain current and synchronous speed signal to obtain the q-axis angle-domain current; Angle domain envelope signal is extracted from the q-axis angle domain current to obtain the angle domain envelope signal; The envelope order spectrum is obtained by calculating the envelope order spectrum of the angle domain envelope signal; Order spectrum analysis of the envelope order spectrum is performed to obtain a fault diagnosis report.

2. The method of claim 1, wherein, Signal transformation is performed on the three-phase raw current signal and the motor speed signal to obtain the filtered q-axis time-domain current and synchronous speed signals, including: The motor speed signal is upsampled and preprocessed to obtain the synchronous speed signal; The instantaneous electrical angle of the motor rotor is determined based on the synchronous speed signal; Based on the instantaneous electrical angle of the motor rotor, the three-phase original current signal is transformed by a rotating coordinate system based on the instantaneous rotation angle to obtain the unfiltered q-axis time-domain current; The unfiltered q-axis time-domain current is high-pass filtered to obtain the filtered q-axis time-domain current.

3. The method of claim 2, wherein, Based on the synchronous speed signal, the instantaneous electrical angle of the motor rotor is determined, including: determining the instantaneous electrical angle of the motor rotor using the following formula, wherein the formula is: wherein is a motor speed signal, is a motor pole pair number.

4. The method of claim 3, wherein, Based on the instantaneous electrical angle of the motor rotor, a rotating coordinate system transformation based on the instantaneous rotation angle is performed on the three-phase raw current signal to obtain the unfiltered q-axis time-domain current. This includes performing a rotating coordinate system transformation on the three-phase raw current signal based on the instantaneous rotation angle using the following formula: wherein, is the instantaneous electrical angle of the motor rotor, is the unfiltered q-axis time domain current, is the three-phase raw current signal.

5. The method of claim 1, wherein, Angle domain envelope signal extraction is performed on the q-axis angle domain current to obtain the angle domain envelope signal, including: Adaptive signal deconstruction based on variational mode decomposition is performed on the q-axis angular domain current to obtain the set of intrinsic mode functions; The optimal fault-sensitive mode is obtained by performing fault-sensitive mode selection based on kurtosis maximization on the intrinsic mode function set; Selective envelope demodulation of the optimal fault-sensitive mode is performed to obtain the angle domain envelope signal.

6. The method of claim 5, wherein, To obtain the optimal fault-sensitive modes, fault-sensitive mode selection based on kurtosis maximization is performed on the intrinsic mode function set, including: performing fault-sensitive mode selection based on kurtosis maximization on the intrinsic mode function set using the following formula: in, Represents the expectation operation. It is the k-th modal component. It is the mean of that modal component. It is its standard deviation, and the argmax operation represents the selection of kurtosis values. The largest modal component, This is the optimal fault-sensitive mode.

7. The method of claim 1, wherein, Order spectrum analysis of the envelope order spectrum is performed to obtain a fault diagnosis report, including: Adaptive noise baseline modeling and significant order extraction are performed on the envelope order spectrum to obtain a significant order set; Based on the theoretical fault order database, fault mode matching and candidate fault set generation are performed on the significant order set to obtain the candidate fault set. A fault diagnosis report is obtained by performing diagnostic rule fusion reasoning and final report generation on the significant order set and candidate fault set.

8. A mechanical fault diagnostic system for a pitch system based on motor current signatures, characterized by, include: The signal acquisition module is used to acquire the three-phase raw current signal and the motor speed signal; The signal conversion module is used to perform signal conversion on the three-phase raw current signal and the motor speed signal to obtain the filtered q-axis time-domain current and synchronous speed signals; The instantaneous rotation angle calculation and equal angle incremental resampling module is used to perform instantaneous rotation angle calculation and equal angle incremental resampling on the filtered q-axis time-domain current and synchronous speed signal to obtain the q-axis angle-domain current; The envelope signal extraction module is used to extract the angular domain envelope signal from the q-axis angular domain current to obtain the angular domain envelope signal. The envelope order spectrum calculation module is used to calculate the envelope order spectrum of the angle domain envelope signal to obtain the envelope order spectrum; The order spectrum analysis module is used to perform order spectrum analysis on the envelope order spectrum to obtain a fault diagnosis report.

9. The motor current signature based system mechanical fault diagnostic system of claim 8, wherein, The signal conversion module includes: The upsampling preprocessing unit is used to perform upsampling preprocessing on the motor speed signal to obtain the synchronous speed signal; The rotor instantaneous electrical angle determination unit is used to determine the instantaneous electrical angle of the motor rotor based on the synchronous speed signal; The rotating coordinate system transformation unit is used to perform a rotating coordinate system transformation on the three-phase original current signal based on the instantaneous electrical angle of the motor rotor to obtain the unfiltered q-axis time-domain current. The high-pass filter unit is used to perform high-pass filtering on the unfiltered q-axis time-domain current to obtain the filtered q-axis time-domain current.

10. The mechanical fault diagnosis system for a pitch system based on motor current characteristics according to claim 8, characterized in that, The envelope signal extraction module includes: An adaptive signal deconstruction unit is used to perform adaptive signal deconstruction based on variational mode decomposition on the q-axis angular domain current to obtain the set of intrinsic mode functions; The fault-sensitive mode selection unit is used to perform fault-sensitive mode selection based on kurtosis maximization on the intrinsic mode function set to obtain the optimal fault-sensitive mode; A selective envelope demodulation unit is used to selectively demodulate the optimal fault-sensitive mode to obtain the angle domain envelope signal.

Images