Online monitoring method and system for a pitch hub slip ring

Abstract

The embodiment of the application provides an online monitoring method and system for a variable pitch hub slip ring, which first acquires original main signals, EMI reference signals and PWM synchronous pulses synchronously, constructs an enhanced reference vector containing noise waveforms and accurate timing information. In order to solve the problem of nonlinear coupling between signals and noise under strong interference and the failure of traditional linear models, a nonlinear modulation link based on signal feedback is introduced. It uses the purified signal at the last moment to dynamically calculate a state-dependent coupling factor, and uses this factor to real-time modulate the linear noise base predicted according to the enhanced reference vector. Finally, the effective noise after nonlinear compensation is subtracted from the original main signal, so that a high-fidelity purified signal is obtained under strong electromagnetic interference background, and the noise residual defect caused by model mismatch of traditional linear cancellation technology is overcome.

Description

Technical Field

[0001] This invention relates to the field of wind power technology, specifically to an online monitoring method and system for a pitch hub slip ring. Background Technology

[0002] The pitch control system of a wind turbine generator is the core unit for achieving power control and safe shutdown. The pitch hub slip ring, as a crucial link connecting the nacelle control system and the pitch drive system inside the rotating hub, is responsible for transmitting power, control signals, and data between the relatively rotating components. Because the slip ring operates in a harsh environment of high altitude, large temperature differences, and high humidity and vibration, any failures such as poor contact, decreased insulation, or wear will directly lead to distorted pitch command transmission, sluggish drive, or even system malfunction, posing a serious threat to the safe operation and power generation efficiency of the wind turbine.

[0003] However, high-precision online monitoring of slip rings faces a severe technical challenge: the harsh electromagnetic environment inside the wheel hub. The pitch drive motor and its inverter within the hub generate high-intensity, wide-bandwidth electromagnetic interference during operation. The monitoring system aims to capture precisely the weak voltage signals, ranging from microvolts to millivolts, that characterize early, subtle faults in the slip ring. These weak fault characteristic signals are easily drowned out by electromagnetic noise several orders of magnitude stronger. Traditional hardware shielding, passive filtering, and optimized grounding designs have limited effectiveness against extremely strong EMI sources, and filters may introduce phase shifts or filter out valuable high-frequency fault information. Even instrumentation amplifiers with high common-mode rejection ratios cannot solve the conversion of common-mode noise to differential-mode noise caused by line impedance imbalance in practical applications; strong interference can even saturate the amplifier front-end, leading to complete information loss. While some adaptive noise cancellation schemes have made progress, they are mostly based on a fundamental physical assumption: that the real signal and electromagnetic noise have a simple linear superposition relationship. However, in real physical systems, the coupling path of noise may be affected by the instantaneous state of the real signal, exhibiting complex state-dependent coupling nonlinear effects. Existing linear adaptive cancellation methods cannot model this nonlinear relationship, resulting in incomplete noise cancellation, low signal fidelity after purification, and ultimately, the inability to accurately extract subtle early fault characteristics, affecting the sensitivity and diagnostic accuracy of the monitoring system.

[0004] Therefore, an optimized online monitoring scheme for the pitch hub slip ring 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 an online monitoring method and system for pitch hub slip rings.

[0006] In a first aspect, embodiments of the present invention provide an online monitoring method for a pitch hub slip ring, comprising: Acquire the original master signal, electromagnetic interference reference signal, and PWM synchronization pulse; Reference signal preprocessing and feature construction are performed on electromagnetic interference reference signal and PWM synchronization pulse to obtain enhanced reference vector; Based on the enhanced reference vector, adaptive noise prediction and cancellation are performed on the original master signal to obtain the purified voltage signal; The purification signal quality is evaluated based on the purified voltage signal and the original main signal to obtain the signal-to-noise ratio improvement. Fault features are extracted and diagnosed from the purified voltage signal to determine its health status.

[0007] Secondly, embodiments of the present invention provide an online monitoring system for a pitch hub slip ring, comprising: The data acquisition module is used to acquire the original main signal, electromagnetic interference reference signal, and PWM synchronization pulse; The reference signal preprocessing and feature construction module is used to perform reference signal preprocessing and feature construction on the electromagnetic interference reference signal and the PWM synchronization pulse to obtain an enhanced reference vector. An adaptive noise prediction and cancellation module is used to perform adaptive noise prediction and cancellation on the original master signal based on the enhancement reference vector to obtain a purified voltage signal. The purification signal quality assessment module is used to assess the purification signal quality based on the purified voltage signal and the original main signal in order to obtain the signal-to-noise ratio improvement. The fault feature extraction and diagnosis module is used to extract and diagnose fault features from the purified voltage signal to obtain its health status.

[0008] Compared with existing technologies, this invention proposes a nonlinear adaptive noise cancellation method. It first constructs an enhanced reference vector containing noise waveforms and precise timing information by synchronously acquiring the original master signal, EMI reference signal, and PWM synchronization pulse. To address the nonlinear coupling between signal and noise and the failure of traditional linear models under strong interference, this concept introduces a nonlinear modulation stage based on signal feedback. It uses the previously purified signal to dynamically calculate a state-dependent coupling factor and uses this factor to modulate the linear noise floor predicted based on the enhanced reference vector in real time. Finally, this nonlinearly compensated effective noise is subtracted from the original master signal, thereby obtaining a high-fidelity purified signal under strong electromagnetic interference, overcoming the noise residue defects caused by model mismatch in traditional linear cancellation techniques. 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 A flowchart illustrating an online monitoring method for a pitch hub slip ring according to an embodiment of the present invention; Figure 2 This is a schematic diagram of the data flow in the online monitoring method of the pitch hub slip ring according to an embodiment of the present invention; Figure 3 This is a flowchart illustrating the online monitoring method for the pitch hub slip ring according to an embodiment of the present invention, which involves reference signal preprocessing and feature construction of electromagnetic interference reference signals and PWM synchronization pulses to obtain an enhanced reference vector. Figure 4 This is a flowchart illustrating the online monitoring method for the pitch hub slip ring according to an embodiment of the present invention, which extracts the timing phase features of the PWM synchronization pulse based on the PWM period and sampling frequency to obtain the normalized PWM phase features. Figure 5 This is a flowchart illustrating the online monitoring method for the pitch hub slip ring according to an embodiment of the present invention, which involves adaptive noise prediction and cancellation of the original main signal based on an enhanced reference vector to obtain a purified voltage signal. Figure 6 This is a block diagram of an online monitoring system for a pitch hub slip ring 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 pitch hub slip ring monitoring technologies mostly employ adaptive noise cancellation techniques in environments with strong electromagnetic interference. However, these techniques are generally based on a flawed physical model assumption: that the actual signal being measured and the intruding electromagnetic noise follow a simple linear superposition relationship. This assumption has significant flaws in practical systems because it ignores the nonlinear effects of the noise coupling path itself being affected by the instantaneous characteristics of the actual signal. This state-dependent coupling leads to incomplete noise cancellation using traditional linear methods, resulting in severely compromised signal fidelity after purification. Therefore, this invention proposes an online monitoring method for pitch hub slip rings. This method aims to overcome the signal fidelity degradation problem caused by the neglect of nonlinear coupling relationships in traditional linear models. It achieves high-fidelity purification by constructing a nonlinear noise model that can sense the signal's own state. Specifically, this method first synchronously acquires the contaminated original main signal, the electromagnetic interference reference signal, and a PWM synchronization pulse that characterizes the noise source behavior. Next, it preprocesses and constructs features for the reference signal, including extracting the EMI delay vector from the electromagnetic interference reference signal and extracting normalized timing phase features based on the PWM synchronization pulse; the two are then concatenated to form an enhanced reference vector. Subsequently, in the adaptive noise cancellation stage, this scheme creatively introduces a nonlinear feedback mechanism. It first predicts a linear noise floor based on the enhanced reference vector; simultaneously, it uses the previously purified output signal as feedback to calculate a state-dependent coupling factor that dynamically reflects the coupling characteristics of the system through a nonlinear function. This coupling factor is used to modulate the linear noise floor, generating an effective noise that is closer to physical reality. Finally, this effective noise is subtracted from the original main signal to obtain a high-fidelity purified voltage signal, which is then used for signal quality assessment and fault diagnosis, thereby solving the failure problem of traditional linear models under strong nonlinear coupling interference.

[0017] Figure 1 This is a flowchart of an online monitoring method for a pitch hub slip ring according to an embodiment of the present invention. Figure 2 This is a schematic diagram of the data flow in the online monitoring method for the pitch hub slip ring according to an embodiment of the present invention. Figure 1 and Figure 2As shown, the online monitoring method for the slip ring of the pitch hub according to an embodiment of the present invention includes the following steps: S100, acquiring the original main signal, electromagnetic interference reference signal, and PWM synchronization pulse; S200, performing reference signal preprocessing and feature construction on the electromagnetic interference reference signal and PWM synchronization pulse to obtain an enhanced reference vector; S300, performing adaptive noise prediction and cancellation on the original main signal based on the enhanced reference vector to obtain a purified voltage signal; S400, performing purified signal quality assessment based on the purified voltage signal and the original main signal to obtain the signal-to-noise ratio improvement; S500, performing fault feature extraction and diagnosis on the purified voltage signal to obtain the health status.

[0018] Specifically, in step S100, the original master signal, electromagnetic interference reference signal, and PWM synchronization pulse are acquired. It is understood that the weak fault signal of the slip ring under test is overwhelmed by high-intensity, periodic electromagnetic interference from the inverter, and this interference is strongly correlated with the PWM switching behavior of the inverter in terms of timing, while the waveform of the interference is affected by the coupling path. Therefore, in the technical solution of this invention, the original master signal, electromagnetic interference reference signal, and PWM synchronization pulse are acquired to provide three key, mutually orthogonal information dimensions for the subsequent adaptive noise cancellation algorithm: namely, the contaminated mixed signal, the waveform characteristic proxy of the noise source, and the precise timing reference of the noise source. This lays the data foundation for the subsequent construction of the enhanced reference vector and the execution of nonlinear predictive cancellation, enabling the algorithm to know not only the shape of the noise but also the precise timing of its occurrence.

[0019] More specifically, in a particular example of the invention, the original main signal is acquired by connecting a high-input-impedance instrumentation amplifier across the slip ring contact to be monitored, in order to measure the differential voltage drop generated when carrying the business current. Next, the electromagnetic interference reference signal is acquired by placing a noise probe, such as a small induction coil, near the main interference source, i.e., the power cable of the VFD (Variable Frequency Drive), without electrical connection to the main signal circuit; this probe is only used to pick up the radiated noise field in that area. Simultaneously, the PWM synchronization pulse is acquired by safely extracting a digital square wave signal consistent with the PWM switching cycle from the IGBT gate drive logic signal of the VFD controller using a high-speed optocoupler. Finally, these three analog or digital signals are connected to a multi-channel synchronous data acquisition device, and a unified sampling clock is set to ensure instantaneous sampling of the three channels under the same time base, obtaining a time-aligned raw data stream. In a specific scenario, the data acquisition card has a 16-bit resolution, 3 synchronous channels, and a sampling rate Fs of 100 kHz. The primary signal channel uses a high input impedance instrumentation amplifier (CMRR≥100 dB) and is equipped with an anti-aliasing analog low-pass filter (cutoff frequency≥0.45Fs). The EMI reference channel uses a small induction coil probe (bandwidth covering the PWM fundamental frequency to 10th harmonic) and is electrically isolated from the main circuit. The PWM synchronization pulse channel extracts the IGBT gate drive pulse through high-speed optocoupler isolation. All three channels share the same hardware clock to ensure sampling alignment.

[0020] Specifically, in step S200, the electromagnetic interference reference signal and PWM synchronization pulse are preprocessed and feature-constructed to obtain an enhanced reference vector. It is understood that the electromagnetic interference reference signal and PWM synchronization pulse obtained in the previous step are two independent, raw data streams, each carrying information about the noise waveform and noise timing, respectively. However, this information has not yet been structured into a feature form that can be directly used by the subsequent adaptive filter. The adaptive algorithm requires a unified, high-dimensional input vector that can simultaneously characterize the dynamic waveform characteristics of the noise and its precise phase position within the PWM period. Therefore, in the technical solution of this invention, the electromagnetic interference reference signal and PWM synchronization pulse are further preprocessed and feature-constructed to obtain an enhanced reference vector. This converts the independent waveform data and timing pulse data into an EMI delay vector containing historical waveform dynamics and a normalized PWM phase feature representing the position within the period, respectively. Finally, these two features of different dimensions are concatenated and fused. In this way, a unified feature vector with enhanced information dimensions can be generated, which simultaneously contains the dynamic waveform information of the noise and the accurate temporal phase information, providing structured input data that is highly correlated with the real noise for the subsequent adaptive noise prediction steps.

[0021] Figure 3 This is a flowchart illustrating the online monitoring method for the pitch hub slip ring according to an embodiment of the present invention, which involves reference signal preprocessing and feature construction of electromagnetic interference reference signals and PWM synchronization pulses to obtain an enhanced reference vector. For example... Figure 3 As shown, step S200 includes: S210, extracting the timing phase features of the PWM synchronization pulse based on the PWM period and sampling frequency to obtain normalized PWM phase features; S220, extracting the EMI delay vector from the electromagnetic interference reference signal based on the delay line order; S230, concatenating the normalized PWM phase features and the EMI delay vector to obtain an enhanced reference vector.

[0022] In step S210, based on the PWM period and sampling frequency, the timing phase features of the PWM synchronization pulse are extracted to obtain normalized PWM phase features. It is understood that since the original PWM synchronization pulse signal is a discrete binary sequence, it can only identify the start time of the PWM period, but cannot provide a continuous numerical feature representing any position within the ZE period for the subsequent adaptive filter. Furthermore, the waveform characteristics of electromagnetic interference noise are strongly correlated with the relative position of the PWM switching behavior within the period. Therefore, in the technical solution of this invention, the timing phase features of the PWM synchronization pulse are further extracted based on the PWM period and sampling frequency to obtain normalized PWM phase features. This converts the discrete pulse events into a continuous floating-point sequence that linearly increases and periodically resets within the [0,1] interval. This sequence is obtained through sample counter state updates and normalization processing, accurately describing the relative time position of each sampling point within its respective PWM period. This provides a numerical timing phase feature that is strongly correlated with the periodicity of the noise for the subsequent enhancement reference vector, enabling the adaptive filter to learn and predict the precise shape of the noise at different phase points within the PWM period.

[0023] Figure 4 This is a flowchart illustrating the online monitoring method for the pitch hub slip ring according to an embodiment of the present invention, which involves extracting the timing phase features of the PWM synchronization pulse based on the PWM period and sampling frequency to obtain normalized PWM phase features. For example... Figure 4 As shown, step S210 includes: S211, extracting the previous PWM pulse sample based on the PWM period and sampling frequency; S212, updating the sample counter state based on the PWM synchronization pulse and the previous PWM pulse sample to obtain the current sample count value; S213, normalizing the current sample count value to obtain the normalized PWM phase feature.

[0024] In step S211, the previous PWM pulse sample is extracted based on the PWM period and sampling frequency. It is understood that subsequent sample counter state updates rely on the accurate identification of the start point of the period in the PWM synchronization pulse sequence, which is determined by detecting the rising edge transition event from 0 to 1. To determine the occurrence of a rising edge, the pulse value at the current sampling moment must be compared with the pulse value at the previous sampling moment. Therefore, in the technical solution of this invention, the previous PWM pulse sample is further extracted based on the PWM period and sampling frequency to provide the necessary historical data representing the state at the previous moment for the edge detection logic. In this way, by comparing the state of the current sample with the extracted previous sample, the start transition event of the PWM period can be accurately captured, ensuring that the sample counter is reset at the correct time, which is the basis for accurate calculation of subsequent phase features.

[0025] More specifically, in a concrete example of the invention, the extraction process is implemented in a digital processing unit operating at the system sampling frequency Fs. A memory unit with a single-cycle delay, such as a register, is configured. At each sampling time t, the processing flow first reads the currently stored value from the register, which is the PWM synchronization pulse value stored at time t-1, i.e., the required previous PWM pulse sample. After using this historical value for subsequent edge detection logic, the processing flow writes the newly acquired PWM synchronization pulse value at the current time t back into the register, overwriting its original content. Through this read-write update operation performed in each sampling cycle, the register always maintains a pulse sample value that is delayed by one sampling cycle from the current data stream.

[0026] In step S212, the sample counter state of the PWM synchronization pulse and the previous PWM pulse sample is updated based on the synchronization pulse to obtain the current sample count value. It is understood that since subsequent normalization processing requires an original value that accurately reflects the relative time position of the current sampling point within the PWM cycle, relying solely on the previous PWM pulse sample extracted at the previous moment is insufficient for cumulative time tracking. A state mechanism must be established that is cleared at the rising edge of the PWM cycle and increments with each sampling point for the remainder of the cycle. Therefore, in the technical solution of this invention, the sample counter state of the PWM synchronization pulse and the previous PWM pulse sample is further updated based on the synchronization pulse to obtain the current sample count value. This allows for the detection of rising edge events by comparing the current and previous pulse sample states, and a conditional operation of resetting or incrementing an internal sample counter is performed based on the detection result. This generates an original sample count value that grows linearly with time and precisely returns to zero at the beginning of each PWM cycle, providing accurate, unscaled phase information for subsequent normalized phase feature calculations.

[0027] More specifically, in a concrete example of the invention, the state update process is executed in a digital logic unit during each sampling clock cycle. First, an edge-detection logic circuit receives the current PWM synchronization pulse and a sample of the previous PWM pulse from the previous moment. By performing a logical check to determine if the previous sample is 0 and the current sample is 1, a single-cycle reset flag signal is generated. Next, the reset flag signal is sent to the control terminal of a sample counter. This sample counter is a resettable adder register. When the reset flag signal is valid, the counter is forced to load a zero value; when the reset flag signal is invalid, the counter increments its currently stored count value and stores the result back to itself. The value output by the counter register at the end of the clock cycle is the required current sample count value, which is then passed to the subsequent normalization processing unit.

[0028] In step S213, the current sample count value is normalized to obtain a normalized PWM phase feature. It is understood that since the current sample count value obtained in the previous step is a raw integer, its range (maximum value) is directly related to the specific sampling frequency and PWM period setting, lacking consistency and hindering the subsequent adaptive algorithm's learning and convergence across operating conditions. Therefore, in the technical solution of this invention, the current sample count value is further normalized to obtain a normalized PWM phase feature, thereby linearly mapping this raw count value, which varies with the sampling rate, to a fixed floating-point interval of [0,1] independent of specific hardware parameters. This generates a standardized phase feature that represents only the relative time position within the period. This feature can be directly used for subsequent enhancement reference vector construction, improving the robustness and versatility of the feature.

[0029] More specifically, in a concrete example of the invention, the normalization process is performed in a digital computing unit. First, based on a preset PWM period and sampling frequency, a floating-point normalized divisor is pre-calculated. This divisor is equal to the product of the PWM period and the sampling frequency, representing the total number of sampling points within a complete period. Next, at each sampling moment, the computing unit receives the current sample count value from the sample counter state update step. Subsequently, the computing unit performs a floating-point division operation, using the current sample count value as the numerator and the pre-calculated normalized divisor as the denominator. The result of this division operation is the desired normalized PWM phase feature, which is output to the subsequent vector concatenation process.

[0030] In step S220, an EMI delay vector is extracted from the electromagnetic interference reference signal based on the delay line order. It is understood that, due to the existence of an unknown transfer function between the electromagnetic interference reference signal at its pickup point and the noise coupled at the main signal acquisition point, determined by the physical propagation path and coupling method, this function introduces time delay, filtering, and phase distortion. This means that the noise intruding into the main signal is related not only to the reference noise at the current moment but also to the reference noise waveform at a series of past moments. Therefore, in the technical solution of this invention, an EMI delay vector is further extracted from the electromagnetic interference reference signal based on the delay line order to capture the short-time dynamic waveform history of the reference noise signal, providing an input structure similar to an FIR filter for the adaptive filter. This provides a multi-dimensional feature containing the time correlation of the noise signal for the subsequent construction of the enhanced reference vector, enabling the adaptive algorithm to learn and simulate this unknown transfer function, thereby more accurately predicting the actual noise pattern intruding into the main signal.

[0031] More specifically, in a concrete example of the present invention, the extraction process is implemented through a digital tapped delay line structure. First, according to a preset delay line order, for example, order M, a shift register or circular buffer of length M is configured in the processing unit, and all M storage cells are initialized to zero. At each sampling clock cycle t, when a new electromagnetic interference reference signal sample arrives, the new sample is written to the first storage cell of the register. Simultaneously, all existing samples in the register are shifted one storage cell to the right. The earliest historical sample located in the Mth storage cell, i.e., the sample acquired at time tM, is removed and discarded. After the shift operation is completed, all M samples stored in the shift register, i.e., the consecutive M reference signal samples from the current time t to time t-M+1, are output as an M-dimensional vector. This output vector is the required EMI delay vector and is sent to subsequent vector concatenation processing.

[0032] In step S230, the normalized PWM phase feature and the EMI delay vector are concatenated to obtain an enhanced reference vector. It is understood that the preceding steps independently generated the EMI delay vector characterizing the noise waveform dynamics and the normalized PWM phase feature characterizing the periodic timing of the noise. Subsequent adaptive noise prediction requires information in both dimensions—the noise's shape and its precise location within the period—to accurately predict the noise actually intruding into the main signal. Therefore, in this invention, the normalized PWM phase feature and the EMI delay vector are further concatenated to obtain an enhanced reference vector. This integrates these two independent features, which differ in both physical meaning and data dimension, into a single, higher-dimensional, and fully informative feature vector. This provides a structured, unified input to the subsequent adaptive filter, containing both the noise's dynamic waveform and the precise timing reference, enabling it to learn and establish the complex mapping relationship between the noise waveform and the PWM phase.

[0033] More specifically, in a concrete example of the present invention, the vector concatenation process is performed at each sampling time t. At time t, a data fusion processing unit simultaneously receives an M-dimensional EMI delay vector from the EMI delay vector extraction step and a 1-dimensional normalized PWM phase feature scalar from the timing phase feature extraction step. The fusion processing unit constructs an M+1-dimensional vector storage space in memory as a carrier for the enhancement reference vector. Subsequently, a data combination operation is performed: all elements of the M-dimensional EMI delay vector are copied in a fixed order to the first M positions of the M+1-dimensional vector; then, the 1-dimensional normalized PWM phase feature scalar is copied to the M+1-th position of the M+1-dimensional vector. After this operation, the M+1-dimensional vector is constructed as the enhancement reference vector for the current time t and is immediately transmitted to the adaptive noise prediction and cancellation step as input data for its linear noise basis component prediction.

[0034] Specifically, in step S300, based on the enhanced reference vector, adaptive noise prediction and cancellation are performed on the original master signal to obtain the purified voltage signal. It is understandable that the acquired original master signal is not only contaminated by noise, but this contamination is not a simple linear superposition. This assumption has significant technical flaws in actual physical systems. The fundamental reason is that the state of the noise coupling path itself is affected by the instantaneous characteristics of the real signal, forming a state-dependent coupling nonlinear effect. For example, when the real signal voltage or current carried by the slip ring contact changes, its microscopic contact impedance is not constant. Simultaneously, when the amplifier at the acquisition front end faces the combined effect of a large-amplitude real signal and strong EMI noise, its operating point may enter the nonlinear region or even saturate. Therefore, the technical solution that attempts to directly subtract a predicted noise component unrelated to the signal state from the original mixed signal inevitably leads to incomplete noise cancellation. Therefore, in the technical solution of this invention, based on the enhanced reference vector, adaptive noise prediction and cancellation are performed on the original main signal to obtain the purified voltage signal. This involves implementing a nonlinear noise cancellation mechanism based on adaptive modulation of the signal state. By introducing a nonlinear feedback loop, the noise prediction process can dynamically sense and compensate for the influence of the signal's own state on the noise coupling path. Specifically, this process uses the purified signal from the previous moment to calculate a state-dependent coupling factor, and uses this factor to modulate the linear noise base predicted based on the enhanced reference vector. Finally, nonlinear modulation noise cancellation is performed. In this way, by modeling this state-dependent nonlinear coupling relationship, more accurate noise cancellation can be achieved, avoiding the problem of large residual errors and severely compromised signal fidelity in purified signals under conditions of large signal dynamic range or extremely strong noise. This results in a high-fidelity purified voltage signal that can be used for accurate fault diagnosis.

[0035] Figure 5 This is a flowchart illustrating the online monitoring method for the pitch hub slip ring according to an embodiment of the present invention, which involves adaptive noise prediction and cancellation of the original main signal based on an enhanced reference vector to obtain a purified voltage signal. Figure 5 As shown, step S300 includes: S310, predicting the linear noise basis component of the enhancement reference vector based on the filter weight vector to obtain the linear noise basis; S320, calculating the state-dependent coupling factor based on the purified signal at the previous moment; S330, performing modulation noise cancellation on the original main signal based on the state-dependent coupling factor and the linear noise basis to obtain the purified voltage signal.

[0036] In step S310, based on the filter weight vector, linear noise basis component prediction is performed on the enhanced reference vector to obtain the linear noise basis. It is understood that since the goal of the entire noise cancellation is to solve the complex nonlinear coupling problem caused by the influence of the instantaneous characteristics of the real signal on the state of the noise coupling path itself in complex application scenarios with strong electromagnetic interference, directly modeling this nonlinear problem in an integrated manner is difficult. Therefore, in the technical solution of this invention, linear noise basis component prediction is further performed on the enhanced reference vector based on the filter weight vector to obtain the linear noise basis, thereby performing the prediction of the linear noise basis component first. Specifically, an adaptive filter is used, with the acquired enhanced reference vector highly correlated with the noise source as input, and the linear basis component of the noise signal is calculated in real time through linear filtering operations. This effectively decomposes the complex nonlinear problem into linear and nonlinear parts for processing, equivalent to simulating the process of electromagnetic noise propagating through an ideal, fixed linear channel and coupling to the main signal channel, thus providing a stable and reliable noise basis waveform, providing raw material for subsequent nonlinear modulation.

[0037] More specifically, in a particular example of the invention, the prediction process performs a linear combination operation once at each sampling time t. At time t... A computing unit receives the augmented reference vector at the current time. The vector is a A column vector of dimension . The first dimensional column vector of this vector. Weiyou The EMI delay vector is composed of electromagnetic interference reference signal samples from consecutive sampling times, i.e., the EMI delay vector. The dimension represents the normalized PWM phase characteristic at the current moment. Simultaneously, this calculation unit reads a... dimensional filter weight vector The vector The weight values ​​were determined by the adaptive algorithm at the previous time step. Update confirmed. Next, perform the vector dot product operation, i.e. of Each weight value is multiplied by middle The first historical noise sample, and at the same time the first Each weight value is multiplied by the current PWM phase characteristic value, and then all Summing the product terms. This operation is expressed by the following formula:

[0038] in, is a linear noise basis, which represents the linear part of the noise predicted at time t; W(t) is the filter weight vector, which is the weight state vector of the adaptive filter at time t. To enhance the reference vector, which is a feature vector composed of the noisy reference source signal and synchronization timing information, the result of the dot product operation is a scalar value, which is determined as the linear noise basis at the current time t and output to subsequent state-dependent coupling factor calculation and modulation noise cancellation steps.

[0039] In step S320, the state-dependent coupling factor is calculated based on the purified signal from the previous time step. It is understood that the linear noise basis predicted in the previous step did not consider the nonlinear effect of state-dependent coupling caused by the influence of the actual signal state on the noise coupling path. Therefore, in the technical solution of this invention, the state-dependent coupling factor is further calculated based on the purified signal from the previous time step, thus creatively introducing a calculation step for the state-dependent coupling factor based on signal feedback to solve the core technical problem of state-dependent coupling. This is used to convert the purified output signal from the previous sampling time step into a state-dependent coupling factor. Feedback is sent to the input, where a pre-defined nonlinear function is used for calculation, dynamically generating a state-dependent coupling factor. This allows for a sophisticated simulation of the nonlinear response in a physical system using a hyperbolic tangent square function: when the instantaneous amplitude of the real signal approaches zero, the system is most vulnerable, with the coupling factor close to 1, allowing most predicted noise to pass through; as the signal amplitude increases, the system tends to stabilize, and the coupling factor smoothly decays to a pre-defined lower limit, thus suppressing noise coupling. Ultimately, a modulation coefficient is generated that dynamically reflects the nonlinear coupling effect in real time, making noise prediction no longer static but adaptive to changes in the signal itself, improving the accuracy of noise modeling.

[0040] More specifically, in a particular example of the invention, this calculation step is performed at each sampling time t. A computing unit first retrieves the purified voltage signal calculated and stored at the previous time t-1 from a memory delayed by one sampling period. The computing unit also reads two preset floating-point constants from the parameter memory: the minimum coupling coefficient. and sensitivity parameters Next, a series of nonlinear function operations are performed: First, obtain... absolute value Then, multiply the absolute value by the sensitivity parameter. Subsequently, a hyperbolic tangent function is applied to the product. And calculate its square value; then multiply the square value by... Finally, subtract the product from the previous step from 1 to obtain the state dependency coupling factor at the current time t. This step can be expressed by the formula:

[0041] in, is the state-dependent coupling factor, which is a calculated coefficient used to modulate the noise. This is the purified signal from the previous moment, where α is a sensitivity parameter representing the response speed of the coupling factor to changes in signal amplitude. is the minimum coupling coefficient, which is a preset lower limit for coupling strength; tanh is the hyperbolic tangent function. For example, suppose... Set it to 0.1, and α to 0.5. When the purification signal from the previous moment... When the amplitude is very small (e.g., 0.01V), The value is close to 0. The calculation results will be close to On the contrary, when When the amplitude is large (e.g., 5V), The value is close to 1. The calculation results will be close to This means that the set minimum coupling coefficient has been achieved. Calculated It is immediately output to the subsequent modulation noise cancellation step.

[0042] In step S330, based on the state-dependent coupling factor and the linear noise basis, modulation noise cancellation is performed on the original main signal to obtain the purified voltage signal. It is understood that since the preceding steps have already generated a linear noise basis to characterize the linear propagation path of noise, and a state-dependent coupling factor to characterize the nonlinear coupling effect, the technical solution of this invention further performs modulation noise cancellation on the original main signal based on the state-dependent coupling factor and the linear noise basis to obtain the purified voltage signal, i.e., modulation noise cancellation is performed. This integrates the linear noise basis and nonlinear coupling factor generated in the aforementioned steps to complete the final signal purification. Specifically, the process is as follows: First, the linear noise basis is multiplicatively modulated using the state-dependent coupling factor to obtain an effective noise prediction value that more closely approximates the actual physical process; second, this effective noise prediction value is subtracted from the original main signal to obtain the final purified signal. In this way, it is no longer a simple linear subtraction, but an informed, nonlinearly compensated subtraction, overcoming the signal fidelity degradation problem caused by neglecting the nonlinear coupling relationship between signal and noise in traditional adaptive noise cancellation technology under strong electromagnetic interference environments. Ultimately, a pure signal with extremely high fidelity is obtained, which allows characteristic signals representing early and weak faults in the slip ring (such as voltage fluctuations or transient impacts at the microvolt level) that are submerged in a background of strong noise to be clearly displayed, thus providing a high-quality data foundation for achieving high-sensitivity and high-reliability online condition monitoring and predictive maintenance of slip rings.

[0043] More specifically, in a specific example of the invention, the cancellation process is performed at each sampling time t, and its execution consists of two steps: First, a computing unit receives data from the linear noise basis component prediction step... , and from the state-dependent coupling factor calculation step Perform multiplication modulation to obtain the effective noise prediction value. .

[0044]

[0045] in, The effective noise is the final predicted noise value after nonlinear modulation. The state-dependent coupling factor. The noise basis is linear. For example, in a scenario where the real signal is at a zero-crossing point, and the sanitized signal from the previous moment... It is close to 0V. At this point... The calculated value is close to 1. If the predicted value at this moment is... If it is 50mV, then Calculated as This simulates the situation where the system is most vulnerable when the signal amplitude is close to zero, and the coupling factor is close to 1. In another scenario, the slip ring carries a large actual signal voltage amplitude. It is 5V. At this time, decay to (e.g., 0.1). Even if Still 50mV Then it is calculated as This simulates the physical process by which the system stabilizes as the signal amplitude increases, and the coupling factor smoothly decays to suppress noise coupling. The second step involves the computational unit receiving data from the original signal acquisition step. And subtract the result calculated in the previous step from it. To obtain the final purified signal .

[0046]

[0047] in, This is the purified voltage signal, which is the final pure output signal; This is the original main signal, which is the acquired, noise-contaminated raw signal. Continuing the above scenario example: In the first scenario, if... (Real signal + noise) is 52mV (assuming the real signal is 2mV), then In the second scenario, if If the value is 5005mV (assuming the actual signal is 5000mV), then Under both vastly different operating conditions, the purified signal was restored with high fidelity. The signal was used for subsequent signal quality assessment and fault diagnosis.

[0048] Specifically, this cleaned-up signal (i.e., the final system error) is used to drive the weight update of the adaptive filter. This is to optimize the performance of the adaptive filter by utilizing the more accurate error signal after cleansing. By using this more accurate error to update the filter weights, it is equivalent to providing a better teacher signal for the adaptive learning process, enabling the filter to converge faster and more stably, thereby achieving a more accurate model of the linear propagation path of noise.

[0049] Specifically, in step S400, the purified signal quality is evaluated based on the purified voltage signal and the original main signal to obtain the signal-to-noise ratio improvement. It is understood that while the preceding adaptive nonlinear cancellation step outputs the purified voltage signal, the actual performance of the algorithm in a dynamic electromagnetic environment, such as the degree of noise suppression and the algorithm's convergence, requires an objective quantitative indicator for real-time evaluation and verification. Therefore, in the technical solution of this invention, the purified signal quality is further evaluated based on the purified voltage signal and the original main signal to obtain the signal-to-noise ratio improvement. This allows for the real-time calculation of a quantitative indicator characterizing the noise suppression effect by comparing the signal states before and after purification. This provides the monitoring system with immediate feedback on anti-interference performance, which can be used to monitor the algorithm's convergence status, evaluate the effectiveness of the purification effect, and provide a confidence reference for signal quality for subsequent fault diagnosis.

[0050] More specifically, in a specific example of the present invention, the evaluation process is executed in an evaluation calculation unit at a preset time window period. First, within the current time window, the evaluation unit caches the original main signal sequence and the purified voltage signal sequence from the corresponding data stream. Then, the unit based on... The relationship is established by subtracting the original main signal sequence from the purified voltage signal sequence point by point to calculate the effective noise signal sequence within the time window. Subsequently, the unit calculates the average power of the purified voltage signal (considered the desired signal) within that window. and the average power of the effective noise signal (considered as canceled noise) within this window. Ultimately, the evaluation unit calculated... and The ratio, and convert it to a decibel value, that is The result is then determined as the improvement in signal-to-noise ratio for the current time window and output to the system log or monitoring interface.

[0051] Specifically, in step S500, fault feature extraction and diagnosis are performed on the purified voltage signal to obtain its health status. It is understood that since the preceding adaptive cancellation step has already provided a high-fidelity purified voltage signal, although this signal recovers the weak fault features submerged by strong noise, it is still time-domain waveform data and has not yet been quantified into an indicator that can be directly used to judge the equipment status. Therefore, in the technical solution of this invention, fault feature extraction and diagnosis are further performed on the purified voltage signal to obtain its health status. This allows for the analysis of the high-fidelity signal using signal processing algorithms to extract sensitive feature quantities that characterize early weak faults in the slip ring, and diagnostic criteria are executed based on these feature quantities. In this way, the purified signal data can be transformed into an intuitive and clear conclusion about the equipment's operating status, ultimately achieving highly sensitive early fault warning and online monitoring.

[0052] More specifically, in a specific example of the present invention, the extraction and diagnosis process is executed periodically within a diagnostic processing unit, with a preset time window as the cycle. First, a feature extraction unit receives the purified voltage signal sequence within the time window from the preceding steps. This unit performs statistical calculations on the sequence data to extract time-domain features that sensitively reflect transient impacts, such as calculating the kurtosis and root mean square (RMS) values ​​of the sequence. Next, a diagnostic logic unit receives the kurtosis and RMS values ​​output by the feature extraction unit. This diagnostic unit has a built-in preset diagnostic rule set. For example, the rule set determines that: if the kurtosis value is greater than a first threshold, such as 6.0, used to detect impact signals, and the RMS value is greater than a second threshold, such as 0.5mV, used to detect energy anomalies, then the health status is determined to be an alarm; if both the kurtosis and RMS values ​​are lower than their respective thresholds, then the health status is determined to be healthy. The diagnostic logic unit finally outputs the health or alarm status code as the final result of this monitoring method.

[0053] In summary, the online monitoring method for the pitch hub slip ring according to an embodiment of the present invention is explained. It first constructs an enhanced reference vector containing noise waveform and precise timing information by synchronously acquiring the original master signal, EMI reference signal, and PWM synchronization pulse. To address the problems of nonlinear coupling between signal and noise and the failure of traditional linear models under strong interference, this concept introduces a nonlinear modulation stage based on signal feedback. It uses the purified signal from the previous moment to dynamically calculate a state-dependent coupling factor and uses this factor to modulate the linear noise floor predicted based on the enhanced reference vector in real time. Finally, this nonlinearly compensated effective noise is subtracted from the original master signal, thereby obtaining a high-fidelity purified signal under strong electromagnetic interference, overcoming the noise residue defect caused by model mismatch in traditional linear cancellation techniques.

[0054] Furthermore, an online monitoring system for the pitch hub slip ring is also provided.

[0055] Figure 6 This is a block diagram of an online monitoring system for a pitch hub slip ring according to an embodiment of the present invention. Figure 6 As shown, the online monitoring system 100 for the pitch hub slip ring according to an embodiment of the present invention includes: a data acquisition module 110 for acquiring the original main signal, electromagnetic interference reference signal, and PWM synchronization pulse; a reference signal preprocessing and feature construction module 120 for performing reference signal preprocessing and feature construction on the electromagnetic interference reference signal and PWM synchronization pulse to obtain an enhanced reference vector; an adaptive noise prediction and cancellation module 130 for performing adaptive noise prediction and cancellation on the original main signal based on the enhanced reference vector to obtain a purified voltage signal; a purified signal quality assessment module 140 for performing purified signal quality assessment based on the purified voltage signal and the original main signal to obtain a signal-to-noise ratio improvement; and a fault feature extraction and diagnosis module 150 for extracting and diagnosing fault features on the purified voltage signal to obtain a health status.

[0056] Furthermore, the reference signal preprocessing and feature construction module 120 includes: The timing phase feature extraction unit is used to extract timing phase features from the PWM synchronization pulse based on the PWM period and sampling frequency to obtain normalized PWM phase features; The EMI delay vector extraction unit is used to extract the EMI delay vector from the electromagnetic interference reference signal based on the delay line order. The vector concatenation unit is used to concatenate the normalized PWM phase characteristics and EMI delay vector to obtain an enhanced reference vector.

[0057] Furthermore, the adaptive noise prediction and cancellation module 130 includes: The linear noise basis component prediction unit is used to predict the linear noise basis components of the enhancement reference vector based on the filter weight vector to obtain the linear noise basis. The state-dependent coupling factor calculation unit is used to calculate the state-dependent coupling factor based on the purified signal from the previous time step. The signal modulation noise cancellation unit is used to perform modulation noise cancellation on the original master signal based on the state-dependent coupling factor and the linear noise basis to obtain a purified voltage signal.

[0058] As described above, the online monitoring system 100 for the pitch hub slip ring 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 a pitch system, or an industrial computer for wind farm-level monitoring. In one possible implementation, the online monitoring system 100 for the pitch hub slip ring 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 online monitoring system 100 for the pitch hub slip ring can be a software module in the control firmware of the computing device or control unit, configured to perform synchronous acquisition of multi-channel signals, extraction of timing phase characteristics of PWM synchronization pulses, construction of EMI delay vectors, and execution of an adaptive nonlinear noise cancellation algorithm. This algorithm includes linear noise basis prediction, calculation of state-dependent coupling factors based on clean signal feedback, and modulation noise cancellation and weight update. Alternatively, it can be a high-fidelity signal cleansing algorithm program developed for the computing device or control unit. Of course, the online monitoring system 100 for the pitch hub slip ring can also be one of the many hardware modules of the computing device or control unit, for example, implemented as a dedicated digital signal processor to efficiently perform the vector dot product and nonlinear function operations required for the adaptive filtering, or embedded in a field-programmable gate array circuit to process the sample counter state update, tap delay line construction and weight update based on nonlinear error in a pipelined manner, or as an application-specific integrated circuit.

[0059] 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. An online monitoring method for a pitch hub slip ring, characterized in that, include: Acquire the original master signal, electromagnetic interference reference signal, and PWM synchronization pulse; Reference signal preprocessing and feature construction are performed on electromagnetic interference reference signal and PWM synchronization pulse to obtain enhanced reference vector; Based on the enhanced reference vector, adaptive noise prediction and cancellation are performed on the original master signal to obtain the purified voltage signal; The purification signal quality is evaluated based on the purified voltage signal and the original main signal to obtain the signal-to-noise ratio improvement. Fault features are extracted and diagnosed from the purified voltage signal to determine its health status.

2. The online monitoring method for the pitch hub slip ring according to claim 1, characterized in that, The electromagnetic interference reference signal and PWM synchronization pulse are preprocessed and feature-constructed to obtain an enhanced reference vector, including: Based on the PWM period and sampling frequency, the timing phase features of the PWM synchronization pulse are extracted to obtain the normalized PWM phase features. Based on the delay line order, the EMI delay vector is extracted from the electromagnetic interference reference signal; The normalized PWM phase characteristics and EMI delay vector are concatenated to obtain an enhanced reference vector.

3. The online monitoring method for the pitch hub slip ring according to claim 2, characterized in that, Based on the PWM period and sampling frequency, timing phase features of the PWM synchronization pulse are extracted to obtain normalized PWM phase features, including: Based on the PWM period and sampling frequency, extract the sample of the previous PWM pulse; The current sample count value is obtained by updating the sample counter state based on the PWM synchronization pulse and the previous PWM pulse sample. The current sample count value is normalized to obtain the normalized PWM phase characteristics.

4. The online monitoring method for the pitch hub slip ring according to claim 1, characterized in that, Based on the enhanced reference vector, adaptive noise prediction and cancellation are performed on the original master signal to obtain the purified voltage signal, including: Based on the filter weight vector, linear noise basis component prediction is performed on the enhancement reference vector to obtain the linear noise basis; Calculate the state-dependent coupling factor based on the purified signal from the previous moment; Based on the state-dependent coupling factor and the linear noise basis, the original master signal is modulated and noise canceled to obtain the purified voltage signal.

5. The online monitoring method for the pitch hub slip ring according to claim 4, characterized in that, Based on the filter weight vector, linear noise basis component prediction is performed on the enhancement reference vector to obtain a linear noise basis, including: predicting the linear noise basis component of the enhancement reference vector using the following formula, wherein the formula is: in, Here is the filter weight vector. To enhance the reference vector, The noise base is linear.

6. The online monitoring method for the pitch hub slip ring according to claim 4, characterized in that, Based on the purified signal from the previous moment, the state-dependent coupling factor is calculated, including: calculating the state-dependent coupling factor using the following formula, wherein the formula is: in, The minimum coupling coefficient, This represents the absolute value of the purified signal from the previous moment. The state-dependent coupling factor. It is the hyperbolic tangent function. This is the sensitivity parameter.

7. The online monitoring method for the pitch hub slip ring according to claim 4, characterized in that, Based on the state-dependent coupling factor and the linear noise basis, modulation noise cancellation is performed on the original main signal to obtain a purified voltage signal, including: modulation noise cancellation of the original main signal using the following formula, wherein the formula is: in, This is the original main signal. For effective noise, As a linear noise basis, The state-dependent coupling factor. This is the voltage signal after purification.

8. An online monitoring system for a pitch hub slip ring, characterized in that, include: The data acquisition module is used to acquire the original main signal, electromagnetic interference reference signal, and PWM synchronization pulse; The reference signal preprocessing and feature construction module is used to perform reference signal preprocessing and feature construction on the electromagnetic interference reference signal and the PWM synchronization pulse to obtain an enhanced reference vector. An adaptive noise prediction and cancellation module is used to perform adaptive noise prediction and cancellation on the original master signal based on the enhancement reference vector to obtain a purified voltage signal. The purification signal quality assessment module is used to assess the purification signal quality based on the purified voltage signal and the original main signal in order to obtain the signal-to-noise ratio improvement. The fault feature extraction and diagnosis module is used to extract and diagnose fault features from the purified voltage signal to obtain its health status.

9. The online monitoring system for the pitch hub slip ring according to claim 8, characterized in that, The reference signal preprocessing and feature construction module includes: The timing phase feature extraction unit is used to extract timing phase features from the PWM synchronization pulse based on the PWM period and sampling frequency to obtain normalized PWM phase features; The EMI delay vector extraction unit is used to extract the EMI delay vector from the electromagnetic interference reference signal based on the delay line order. The vector concatenation unit is used to concatenate the normalized PWM phase characteristics and EMI delay vector to obtain an enhanced reference vector.

10. The online monitoring system for the pitch hub slip ring according to claim 8, characterized in that, The adaptive noise prediction and cancellation module includes: The linear noise basis component prediction unit is used to predict the linear noise basis components of the enhancement reference vector based on the filter weight vector to obtain the linear noise basis. The state-dependent coupling factor calculation unit is used to calculate the state-dependent coupling factor based on the purified signal from the previous time step. The signal modulation noise cancellation unit is used to perform modulation noise cancellation on the original master signal based on the state-dependent coupling factor and the linear noise basis to obtain a purified voltage signal.

Images