Method and system for monitoring a twisted state of a pitch system cable

Abstract

This application discloses a method and system for monitoring the torsional state of a pitch system cable. It abandons the traditional approach of directly measuring local deformation or stress, instead monitoring the cable's vibration response and identifying its global modal parameters (such as modal frequency and damping). Since any local stiffness change in the cable (such as kinking caused by uneven torsion) will cause a corresponding drift in its overall modal parameters, this solution effectively solves the technical problem that traditional point-based sensing cannot assess overall health by using a point-to-surface global state perception. Furthermore, by deeply mining the physical co-modes in the modal deviation vector, especially specifically diagnosing the characteristic of frequency and damping increasing in the same direction caused by torsional hardening, the degree of torque non-uniformity can be accurately identified and quantified, ultimately achieving sensitive detection and reliable early warning of early local faults in the cable.

Description

Technical Field

[0001] This application relates to the field of pitch system monitoring technology, and more specifically, to a method and system for monitoring the cable twisting state of a pitch system. Background Technology

[0002] Wind turbine generators are key equipment for achieving the clean energy strategy, and their long-term reliable operation is crucial for ensuring energy supply. The pitch system plays a central role in wind turbine generators, optimizing energy capture efficiency and controlling turbine load by adjusting the blade pitch angle in real time. The power and signal cables laid within the pitch system endure complex torsional, bending, and tensile stresses over long periods due to the frequent, large-angle reciprocating rotation of the blades, making them one of the weakest links in the entire system most susceptible to fatigue damage and failure. If the pitch cable fails due to excessive torsion, localized kinking, or fatigue fracture, it will directly lead to pitch failure. This can result in unplanned turbine shutdowns and power generation losses, or even dangerous blade angles leading to overspeeding and other catastrophic accidents, seriously threatening turbine safety.

[0003] To address this challenge, various monitoring methods have been explored in existing technologies. Early solutions primarily relied on indirect cumulative counting methods, such as recording the cumulative rotation angle or number of pitch changes of the pitch motor, triggering a maintenance warning when the cumulative value reaches a preset empirical threshold. The fundamental flaw of these methods lies in their simplification of the complex, elastically distributed cable into an idealized lumped parameter model, assuming uniform cable torsion. However, in actual operating conditions, influenced by gravity, installation processes, temperature variations, and dynamic effects, the torsional stress of the cable often exhibits a highly uneven distribution, easily concentrating at the cable's fixed ends or specific bending sections. Cumulative counting methods are completely unable to detect this fatal localized stress concentration, thus failing to provide effective warnings before actual damage occurs. While subsequent attempts have focused on direct measurement using sensors such as encoders and strain gauges, these solutions are inherently limited by point-based sensing. The information from a single sensor only represents the state at its location, failing to reflect the overall health of the entire cable. Furthermore, large-scale deployment of sensor arrays presents cost and reliability issues. Therefore, existing technologies generally face a core technical challenge: how to effectively monitor the spatial non-uniformity of cable torsion state and identify early damage caused by local stress concentration in a low-cost and highly reliable manner.

[0004] Therefore, an optimized monitoring scheme for the cable torsion status of pitch systems is desired. Summary of the Invention

[0005] To address the aforementioned technical problems, this application is proposed. Embodiments of this application provide a method and system for monitoring the cable twisting state of a pitch system.

[0006] According to one aspect of this application, a method for monitoring the cable twist condition of a pitch system is provided, comprising: Acquire real-time pitch angle and real-time acceleration data; Online modal parameter identification is performed on real-time pitch angle and real-time acceleration data to obtain online modal frequencies and online modal damping; Based on online modal frequencies, online modal damping, and baseline modal maps, the state deviation of the current pitch angle is calculated and evaluated to obtain the deviation vector; Torque non-uniformity is diagnosed on the deviation vector to obtain torque non-uniformity index; An alarm signal is generated based on the comparison between the torque non-uniformity index and the preset static threshold.

[0007] According to another aspect of this application, a monitoring system for the cable twisting state of a pitch system is provided, the system being capable of implementing the method described above, comprising: The real-time data monitoring and acquisition module is used to acquire real-time pitch angle and real-time acceleration data; The online modal parameter identification module is used to perform online modal parameter identification on real-time pitch angle and real-time acceleration data to obtain online modal frequencies and online modal damping; The state deviation calculation module is used to calculate and evaluate the state deviation of the current pitch angle based on online modal frequency, online modal damping and baseline modal map to obtain the deviation vector; The torque non-uniformity diagnosis module is used to diagnose torque non-uniformity in the deviation vector to obtain torque non-uniformity index. The alarm module is used to generate alarm signals based on the comparison between the torque non-uniformity index and the preset static threshold.

[0008] Compared with existing technologies, the method and system for monitoring the torsional state of pitch system cables provided in this application abandon the traditional approach of directly measuring local deformation or stress. Instead, it monitors the vibration response of the cable and identifies its global modal parameters (such as modal frequency and damping). Since any local stiffness change at any location in the cable (such as kinking caused by torsional inhomogeneity) will cause a corresponding drift in its overall modal parameters, this solution achieves global state perception from a point-to-surface perspective, effectively solving the technical problem that traditional point-based sensing cannot assess overall health. Furthermore, this solution deeply mines the physical co-modes in the modal deviation vector, especially specifically diagnosing the characteristic of frequency and damping increasing in the same direction caused by torsional hardening. This allows for accurate identification and quantification of the degree of torque inhomogeneity, ultimately achieving sensitive capture and reliable early warning of early local faults in the cable, overcoming the fundamental deficiency of existing technologies in effectively monitoring torsional inhomogeneity. Attached Figure Description

[0009] The above and other objects, features, and advantages of this application will become more apparent from the more detailed description of the embodiments of this application in conjunction with the accompanying drawings. The drawings are provided to further illustrate the embodiments of this application and form part of the specification. They are used together with the embodiments of this application to explain this application and do not constitute a limitation thereof. In the drawings, the same reference numerals generally represent the same components or steps.

[0010] Figure 1 This is a flowchart of a method for monitoring the cable twisting state of a pitch system according to an embodiment of this application; Figure 2 This is a data flow diagram of a method for monitoring the cable twisting state of a pitch system according to an embodiment of this application; Figure 3 This is a flowchart illustrating the online modal parameter identification of real-time pitch angle and real-time acceleration data to obtain online modal frequency and online modal damping in the method for monitoring the cable torsion state of a pitch system according to an embodiment of this application. Figure 4 This is a flowchart illustrating a method for monitoring the cable torsion state of a pitch system according to an embodiment of this application, which diagnoses torque non-uniformity by analyzing a deviation vector to obtain a torque non-uniformity index. Figure 5 This is a block diagram of a monitoring system for the cable twisting state of a pitch system according to an embodiment of this application. Detailed Implementation

[0011] Hereinafter, exemplary embodiments according to this application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of this application, and not all embodiments of this application. It should be understood that this application is not limited to the exemplary embodiments described herein.

[0012] As indicated in this application and claims, unless the context clearly indicates otherwise, the words "a," "an," "an," and / or "the" are not specifically singular and may include plural forms. Generally speaking, the terms "comprising" and "including" only indicate the inclusion of explicitly identified steps and elements, which do not constitute an exclusive list, and the method or apparatus may also include other steps or elements.

[0013] While this application makes various references to certain modules of the systems according to embodiments of this application, any number of different modules can be used and run on user terminals and / or servers. The modules described are merely illustrative, and different aspects of the systems and methods may use different modules.

[0014] Flowcharts are used in this application to illustrate the operations performed by the system according to embodiments of this application. It should be understood that the preceding or following operations are not necessarily performed in exact order. Instead, various steps can be processed in reverse order or simultaneously as needed. Furthermore, other operations can be added to these processes, or one or more steps can be removed from them.

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

[0016] Existing monitoring technologies for pitch cables are generally limited by their indirectness and idealized assumptions about torsional uniformity, thus failing to effectively identify and warn of the critical failure mode of torsional non-uniformity caused by localized stress concentration. To address this technical challenge, this application proposes a method for monitoring the torsional state of pitch system cables. This method no longer tracks macroscopic rotation angles but treats the cable as a precise vibrating structure, capturing its vibration response in real time using accelerometers deployed on it. First, online modal parameter identification is performed on the collected acceleration data to accurately extract a dynamic fingerprint characterizing the current global physical state of the cable, namely, multi-order modal frequencies and modal damping. Subsequently, these real-time identified parameters are precisely compared with a pre-constructed baseline modal map covering all pitch angles to quantify and generate a multi-dimensional state deviation vector. Furthermore, instead of simply evaluating the magnitude of the deviation, a deep diagnosis is performed on this vector. Through a specially constructed torque non-uniformity diagnostic model, it specifically identifies and amplifies those physical co-deviation modes (such as simultaneous positive deviations in frequency and damping) that indicate the risk of localized kinking and hardening. Ultimately, the process generates a quantitative index that is highly sensitive to torsional non-uniformity, thereby enabling accurate location and reliable early warning of early, localized lesions in cables.

[0017] Figure 1 This is a flowchart of a method for monitoring the cable twisting state of a pitch system according to an embodiment of this application. Figure 2 This is a schematic diagram of the data flow in a method for monitoring the cable twisting state of a pitch system according to an embodiment of this application. Figure 1 and Figure 2 As shown, the method for monitoring the cable twist state of a pitch system according to an embodiment of this application includes the following steps: S100, acquiring real-time pitch angle and real-time acceleration data. S200, performing online modal parameter identification on the real-time pitch angle and real-time acceleration data to obtain online modal frequency and online modal damping. S300, calculating and evaluating the state deviation of the current pitch angle based on the online modal frequency, online modal damping, and baseline modal map to obtain a deviation vector. S400, performing torque non-uniformity diagnosis on the deviation vector to obtain a torque non-uniformity index. S500, generating an alarm signal based on a comparison between the torque non-uniformity index and a preset static threshold.

[0018] Specifically, in step S100, real-time pitch angle and real-time acceleration data are acquired. It should be understood that, as the pitch system cable is a flexible structure, its dynamic characteristics, especially its modal parameters, change significantly with the degree of torsion. The real-time pitch angle is the most direct physical quantity characterizing this degree of torsion. Therefore, in the technical solution of this application, by synchronously acquiring real-time pitch angle and real-time acceleration data, data input with accurate operating conditions is provided for subsequent online modal parameter identification, ensuring that the identified modal parameters can be accurately correlated with their corresponding physical state, i.e., the torsion angle. This establishes an accurate and comparable benchmark for subsequent state deviation calculations, fundamentally guaranteeing the effectiveness and reliability of the entire monitoring method.

[0019] More specifically, in a particular example of this application, the acquisition process first includes securely mounting one or more multi-axis accelerometers at predetermined locations on the pitch cable bundle to be monitored, such as its midpoint or a location with significant vibration response. The accelerometers are connected via data cables to a data acquisition unit located within the hub. Simultaneously, the data acquisition unit accesses the wind turbine's internal control network, such as a CAN bus or industrial Ethernet, to read high-precision real-time pitch angle data from the servo controller of the pitch drive motor or its built-in absolute position encoder. The data acquisition unit has a built-in synchronous clock that, upon receiving a monitoring command, acquires acceleration data for a specific duration, such as 5 seconds, at a preset sampling frequency, such as 200Hz, and records the instantaneous pitch angle value obtained from the control network at the beginning or center of this acquisition period. Finally, the acquisition unit encapsulates the time-series acceleration data and the single pitch angle value as an indivisible data pair, forming a measurement data packet with a working condition label, and outputs it to the subsequent online modal parameter identification module.

[0020] Specifically, in step S200, online modal parameter identification is performed on the real-time pitch angle and real-time acceleration data to obtain online modal frequencies and online modal damping. It should be understood that the original real-time acceleration data is a complex time-domain signal containing multiple vibration modes, background noise, and measurement interference. It directly reflects the instantaneous motion state of the cable rather than its inherent structural health properties, making it unsuitable for direct quantitative condition assessment. Therefore, in the technical solution of this application, online modal parameter identification is further performed on the real-time pitch angle and real-time acceleration data to obtain online modal frequencies and online modal damping. This deconstructs the mixed vibration signal and transforms it into a set of structured physical parameters that can stably characterize the global dynamic characteristics of the cable. This allows for the acquisition of a series of quantitative indicators highly sensitive to changes in structural stiffness, mass, and damping distribution, providing a reliable and comparable data basis for subsequent accurate condition deviation calculations and fault diagnosis.

[0021] Figure 3 This document presents a flowchart illustrating the online modal parameter identification of real-time pitch angle and real-time acceleration data to obtain online modal frequencies and online modal damping, based on the method for monitoring the cable torsion state of a pitch system according to an embodiment of this application. Figure 3 As shown, step S200 includes: S210, performing signal preprocessing and cross-power spectral density matrix construction on the real-time acceleration data to obtain the cross-power spectral density matrix; S220, performing singular value decomposition and modal peak identification on the cross-power spectral density matrix to obtain singular value spectrum data and a peak frequency list; S230, accurately estimating the modal frequencies and damping ratios based on the singular value spectrum data and the peak frequency list to obtain the online modal frequencies and online modal damping.

[0022] Accordingly, in step S210, the real-time acceleration data undergoes signal preprocessing and cross-power spectral density matrix construction to obtain the cross-power spectral density matrix. It should be understood that since the real-time acceleration data directly acquired from the sensor is a raw time-domain waveform, it not only includes the coupled responses of the cable's multiple vibration modes but also background noise from environmental excitations such as wind turbine rotation and the sensor's own random electrical noise. Therefore, in the technical solution of this application, the real-time acceleration data is further preprocessed and a cross-power spectral density matrix is ​​constructed to obtain the cross-power spectral density matrix. This purifies the signal, suppresses random interference, and transforms the analysis domain from the time domain to the frequency domain, constructing a mathematical expression that comprehensively characterizes the system's energy distribution and inter-channel correlation characteristics in the frequency domain. This provides a stable, reliable, and information-rich input for subsequent modal parameter identification based on frequency domain decomposition, a necessary prerequisite for accurately separating each mode and identifying its characteristics.

[0023] Specifically, in a concrete example of this application, the process first performs signal preprocessing independently on each channel of the input multi-channel real-time acceleration data sequence. This preprocessing includes applying a high-pass filter to remove DC bias and low-frequency drift near zero frequency, and applying a low-pass filter to remove high-frequency noise beyond the analysis range. Subsequently, the preprocessed time-domain data is segmented and windowed using, for example, the Welch averaging method. A fast Fourier transform is performed on each windowed data segment to convert it into a frequency domain signal. Finally, by calculating the ensemble average of the frequency domain signals of all data segments, the cross-power spectral density and auto-power spectral density of each channel are obtained, and these spectral density values ​​are combined at each discrete frequency point into a Hermitian matrix, which is the final output cross-power spectral density matrix. The diagonal elements of this matrix represent the frequency distribution of the energy of the vibration signal of each channel, while the off-diagonal elements contain the amplitude and phase relationships between the signals of different channels, providing complete information for subsequent mode decomposition.

[0024] Accordingly, in step S220, singular value decomposition and modal peak identification are performed on the cross-power spectral density matrix to obtain singular value spectrum data and a list of peak frequencies. It should be understood that because the cross-power spectral density matrix contains the coupled response information of all vibration modes at each frequency point, the energy contributions of each mode are intertwined, making it difficult to directly and clearly separate and identify the center frequency corresponding to a single mode. Therefore, in the technical solution of this application, singular value decomposition and modal peak identification are further performed on the cross-power spectral density matrix to obtain singular value spectrum data and a list of peak frequencies. This decouples the complex response of the multi-degree-of-freedom system into a series of independent, energy-ordered approximate representations of the single-degree-of-freedom system response, thereby highlighting the location of the system modal frequencies. In this way, the modal information hidden in the matrix can be visualized as significant peaks on the spectrum, providing a clear and intuitive basis for subsequent accurate positioning of each modal frequency.

[0025] In a specific example of this application, the process performs singular value decomposition on the cross-power spectral density matrix at each discrete frequency point within a preset analysis frequency band. Since the first singular value physically represents the most significant vibrational energy contribution at that frequency, the system constructs a curve representing the sequence of first singular values ​​at all frequency points, which varies with frequency; this is the singular value spectrum data. On this spectrum, the cable's natural modal frequencies exhibit sharp peaks with concentrated energy due to resonance. Subsequently, an automatic peak finding algorithm is applied to this singular value spectrum data. This algorithm pre-sets criteria such as minimum peak height and minimum peak spacing to ensure that only truly independent modal peaks with significant energy are identified, while filtering out spurious peaks caused by noise fluctuations. The output of this algorithm is a list of peak frequencies containing the approximate frequency values ​​corresponding to all identified modal peaks; this list serves as input for the next step of precise parameter estimation.

[0026] Accordingly, in step S230, the modal frequencies and damping ratios are precisely estimated based on the singular value spectrum data and the peak frequency list to obtain the online modal frequencies and online modal damping. It should be understood that since the peak frequency list only provides a preliminary estimate of the modal frequencies, its accuracy is limited by the frequency domain resolution and fails to reveal the key parameter characterizing the system's energy dissipation characteristics, namely the modal damping ratio, and frequency information alone is insufficient to comprehensively assess the structural health status. Therefore, in the technical solution of this application, the modal frequencies and damping ratios are further precisely estimated based on the singular value spectrum data and the peak frequency list to obtain the online modal frequencies and online modal damping. This allows for refined analysis of each identified modal peak, simultaneously extracting high-precision natural frequencies and damping ratios from its frequency domain morphology. In this way, the preliminary, qualitative modal localization results can be transformed into a structured set of final parameters that can accurately quantify the current dynamic characteristics of the cable, providing decisive data support for subsequent state deviation assessments.

[0027] Specifically, in this embodiment, the precise estimation of modal frequencies and damping ratios based on singular value spectrum data and a peak frequency list to obtain online modal frequencies and online modal damping includes: performing modal peak isolation and free decay response generation on the singular value spectrum data and peak frequency list to obtain an autocorrelation function set; calculating the logarithmic decay rate and identifying the modal damping ratio on the autocorrelation function set to obtain a damped parameter set; and performing precise modal frequency calculation and parameter integration on the damped parameter set to obtain online modal frequencies and online modal damping. It is worth noting that here, the damped parameter set includes modal damping ratios and damped vibration frequencies. That is, more specifically, this precise estimation process iteratively processes each peak frequency in the peak frequency list. First, a frequency window is extracted from the singular value spectrum data, centered on the current peak frequency to be processed. The data within this window is treated as the frequency domain response of a single-degree-of-freedom system and isolated. Then, an inverse fast Fourier transform is performed on this window to generate an autocorrelation function curve in the time domain, which represents an ideal free-dampening vibration waveform. Next, the logarithmic decay method is applied to this autocorrelation function curve. By picking the amplitudes of multiple consecutive peaks, the logarithmic decay rate is calculated, thereby identifying the damping ratio of this mode. Simultaneously, the damped vibration frequency is obtained by measuring the time interval between adjacent peaks. Both of these constitute a damped parameter set. Finally, based on the damping ratio and damped vibration frequency in this damped parameter set, the more fundamental natural mode frequencies, unaffected by damping, are calculated using structural dynamics formulas. These precise mode frequencies are then paired with the identified modal damping ratios. After iterative processing of all peaks, the process ultimately outputs a complete set of online modal frequencies and online modal damping for each mode.

[0028] Specifically, in step S300, based on the online modal frequency, online modal damping, and baseline modal map, the current pitch angle is calculated and evaluated to obtain a deviation vector. It should be understood that since the absolute values ​​of the online modal parameters are not only a function of the cable's health status but also strongly dependent on the current pitch angle operating condition, directly applying threshold judgments to the modal parameters without eliminating this conditional dependency will result in numerous false alarms or missed alarms. Therefore, in the technical solution of this application, the current pitch angle is further calculated and evaluated based on the online modal frequency, online modal damping, and baseline modal map to obtain a deviation vector. This transforms the original, condition-coupled modal parameters into a standardized feature representation that only reflects the degree of deviation from the health status. In this way, a stable health indicator unaffected by the pitch angle can be generated and directly used for subsequent fault diagnosis and trend analysis.

[0029] More specifically, in this embodiment, based on online modal frequencies, online modal damping, and a baseline modal map, a state deviation calculation and evaluation is performed on the current pitch angle to obtain a deviation vector. This includes: accurately interpolating the current pitch angle based on the baseline modal map to obtain interpolation reference parameters, wherein the interpolation reference parameters include interpolated healthy modal frequencies and interpolated healthy modal damping. State deviation calculations are then performed on the interpolated healthy modal frequencies and interpolated healthy modal damping, as well as the online modal frequencies and online modal damping, to obtain a deviation vector, wherein the deviation vector includes relative frequency deviation and relative damping deviation. That is, more specifically, the process first receives an online modal parameter set and the current pitch angle synchronized with it. Based on the current pitch angle value, the system retrieves and locates two adjacent discrete reference angle points in the pre-stored baseline modal map. Subsequently, the healthy modal frequencies and healthy modal damping corresponding to these two reference angle points are extracted, and for each mode, a linear interpolation algorithm is applied to calculate the theoretically required healthy reference parameters, i.e., the interpolation reference parameters, at the current pitch angle. After obtaining a health baseline that precisely matches the current operating conditions, the online modal parameters and interpolated baseline parameters are compared sequentially. The relative frequency deviation and relative damping deviation of each mode are obtained by calculating the difference between the two and dividing by the baseline value. Finally, the relative frequency deviation and relative damping deviation of all modes are arranged in a predefined fixed order and combined into a single, one-dimensional real vector. This is the final output deviation vector that comprehensively quantifies the current system's health deviation state.

[0030] Specifically, in step S400, torque non-uniformity diagnosis is performed on the deviation vector to obtain a torque non-uniformity index. It should be understood that since the deviation vector is a multi-dimensional data volume containing information on all modal deviations, the complex fault modes it contains, especially the specific physical synergistic effects indicating torsional non-uniformity, cannot be directly revealed by simple mathematical norms or linear thresholds. That is, in the scenario of monitoring torsional non-uniformity in pitch cables, the traditional approach is to generate a health index by calculating the L2 norm of the deviation vector. This method treats the contribution of all modal deviations equally. However, in structural dynamics, higher-order modes, due to their shorter wavelengths, naturally exhibit higher sensitivity to subtle changes in local stiffness or mass (such as early kinking). This equalization approach will obscure the valuable early fault signs carried by higher-order modes. Furthermore, this method directly adds the squares of the frequency deviation and the damping deviation, failing to reflect the physical correlation between the two under specific fault modes. For example, localized kinking and hardening in cables is often accompanied by a synergistic phenomenon of increased stiffness (positive frequency deviation) and increased internal friction (positive damping deviation). Simple sum-of-squares operations cannot capture or enhance this synergistic effect, which has clear diagnostic implications, thus failing to effectively distinguish between different fault modes. Furthermore, crudely compressing a high-dimensional deviation vector rich in diagnostic information into a scalar results in severe information loss, causing similar indices to correspond to drastically different physical states, thereby reducing the specificity of the diagnosis.

[0031] To overcome the aforementioned technical challenges, a preferred embodiment of this solution proposes a weighted collaborative anomaly index construction process that reflects modal sensitivity differences and enhances physical synergy. This process achieves accurate quantitative assessment of cable health status through a series of tightly integrated data processing steps. In the technical solution of this application, torque non-uniformity diagnosis is further performed on the deviation vector to obtain a torque non-uniformity index. Based on preset physical prior knowledge and diagnostic rules, deep feature extraction and pattern recognition are performed on the deviation vector, thereby specifically amplifying the deviation component highly correlated with torsional non-uniformity faults. In this way, a generalized deviation metric can be transformed into a torque non-uniformity index with clear physical meaning and high diagnostic specificity, greatly improving the sensitivity and accuracy of fault monitoring.

[0032] Figure 4 This is a flowchart illustrating the method for monitoring the cable torsion state of a pitch system according to an embodiment of this application, which diagnoses torque non-uniformity by analyzing the deviation vector to obtain a torque non-uniformity index. Figure 4As shown, step S400 includes: S410, extracting the total number of modes from the deviation vector; S420, constructing a modal sensitivity weight vector based on the total number of modes; S430, calculating a physical coordination anomaly score on the deviation vector to obtain a physical coordination anomaly score vector; and S440, performing a weighted aggregation of the modal sensitivity weight vector and the physical coordination anomaly score vector to obtain a weighted coordination anomaly index as the torque non-uniformity index.

[0033] Accordingly, in steps S410 and S420, the total number of modes is extracted from the deviation vector, and a modal sensitivity weight vector is constructed based on the total number of modes. It should be understood that since the deviation vector contains information on multiple modes from low to high order, traditional evaluation methods often assign equal weights to all modes when processing this vector. However, according to prior physical knowledge of structural dynamics, higher-order modes, due to their shorter wavelengths, naturally exhibit much higher sensitivity to local stiffness changes in cables (e.g., early kinks caused by torsional inhomogeneity) than lower-order modes. This equal weighting approach would obscure the highly valuable early fault symptoms carried by higher-order modes. Therefore, in the technical solution of this application, the total number of modes is further extracted from the deviation vector, and a modal sensitivity weight vector is constructed based on the total number of modes. This establishes an active weighting mechanism by assigning differentiated importance to the deviation information of different modes based on this prior physical knowledge. This ensures that subsequent diagnostic calculations act like a mathematical magnifying glass with adjustable focus, concentrating on the high-frequency details of cable vibration, prioritizing and amplifying the local, subtle abnormal changes in the cable reflected by higher-order modes, thereby significantly improving the overall assessment system's sensitivity to detecting local early faults such as torsional inhomogeneity.

[0034] Specifically, in a concrete example of this application, the process first parses the total number of modes to be analyzed, denoted as N, from the structural or dimensional information of the input bias vector. For example, if the bias vector is defined as containing the frequency bias and damping bias of the 3rd order modes, then the total number of modes N is determined to be 3. Subsequently, a pre-configured sensitivity amplification factor greater than 0 is loaded. Based on the total number of modes N and the amplification factor A power-law function is used to generate the weight coefficients for each mode, constructing a mode sensitivity weight vector of length N. .along with As the value increases, the weight of higher-order modes grows rapidly in a non-linear manner. This allows subsequent calculations to prioritize and amplify subtle, localized anomalies in the cable reflected by higher-order modes, such as local stiffness hardening points caused by torsional inhomogeneity. In this way, through proactive weighting, the algorithm model ensures that it does not ignore any weak but critical early fault signals originating from higher-order modes, thereby significantly improving the overall assessment system's sensitivity to detecting localized early faults such as torsional inhomogeneity. This provides a solid data foundation for predictive maintenance.

[0035] Accordingly, in step S430, a physical co-anomaly score is calculated on the deviation vector to obtain a physical co-anomaly score vector. It should be understood that since the deviation vector is a multi-dimensional information body containing deviations in various physical quantities such as frequency and damping, traditional evaluation methods, such as simply calculating its sum of squares or L2 norm, fail to reflect the physical correlation of each deviation component under specific fault modes. Specifically, the critical fault mode of localized cable kinking and hardening manifests simultaneously in a physical scenario as increased local stiffness (leading to positive frequency deviation) and increased internal friction (leading to positive damping deviation). Simple mathematical distance metrics cannot capture and specifically enhance this synergistic effect with clear diagnostic indications, resulting in an inability to effectively distinguish different fault modes. Therefore, in the technical solution of this application, a physical co-anomaly score is further calculated on the deviation vector to obtain a physical co-anomaly score vector, thereby transforming the scoring function from a purely mathematical distance metric into a fault mode recognizer with physical insight and diagnostic specificity. In this way, the original deviation vector can be transformed into a completely new scoring vector. In this scoring vector, the modes that closely match the characteristics of torsional non-uniformity faults will be assigned extremely high scores, thereby achieving precise targeted enhancement of specific fault modes at the data level and laying a solid foundation for the final reliable diagnosis.

[0036] Specifically, in a particular example of this application, the process first loads the independent contribution coefficients used to adjust the basic importance from the system configuration. and and a positive synergistic enhancement coefficient. Then, iterate through each mode in the input bias vector, from j=1 to N. For the j-th mode, extract its corresponding frequency relative deviation from the bias vector. and relative deviation of damping Next, a quadratic function containing a cooperative enhancement term is applied to compute the physical cooperative anomaly score for this mode. Due to the gating effect of the Heaviside function, this term is only activated and contributes when both frequency deviation and damping deviation are positive. This design perfectly maps the physical scenario of localized kink hardening in cables, where an increase in local stiffness (leading to positive frequency deviation) and an increase in internal friction (leading to positive damping deviation) occur simultaneously. Under this specific condition, this term produces a significant positive superposition effect, thereby greatly increasing the anomaly score of this mode. In particular, the calculated cooperative anomaly score... It is no longer a vague indicator of the magnitude of the deviation, but has become a highly specific indicator that is highly sensitive to kink hardening, a key failure mode, greatly enhancing the diagnostic certainty of subsequent health indices. After calculating the j-th order... Then, it is stored in the j-th position of a new vector. After traversing all N modes, the process finally outputs an N-dimensional physical cooperative anomaly score vector.

[0037] Accordingly, in step S440, the modal sensitivity weight vector and the physical coordination anomaly score vector are weighted and aggregated to obtain a weighted coordination anomaly index as the torque non-uniformity index. It should be understood that, since the preceding steps have refined the multidimensional deviation information into two independent vectors—namely, the modal sensitivity weight vector representing the importance of different modes and the physical coordination anomaly score vector quantifying the fault characteristics of each mode—this discrete information, with its weights and physical meaning, needs to be integrated into a single, robust, easily interpretable, and trend-analyzable final health indicator for alarm decision-making. Therefore, in the technical solution of this application, the modal sensitivity weight vector and the physical coordination anomaly score vector are further weighted and aggregated to obtain a weighted coordination anomaly index as the torque non-uniformity index, thereby achieving a final integration and enhancement of the entire diagnostic methodology.

[0038] Specifically, in a specific example of this application, the aggregation process receives the modal sensitivity weight vector and physical coordination anomaly score vector output from the preceding steps as input. First, element-wise multiplication is performed on these two vectors; that is, for each modal order from j=1 to N, the product of its weight and score is calculated. Then, the products calculated for all orders are summed to obtain a weighted sum of coordination anomaly scores. Finally, the square root of this sum is performed to obtain the final weighted coordination anomaly index, which is used as a torque non-uniformity index characterizing the degree of cable torsional non-uniformity. That is, the modal sensitivity weight vector and physical coordination anomaly score vector are weighted and aggregated using the following formula:

[0039] in, It is the total number of modes. It is the modal sensitivity weight. It is a score for abnormal physical coordination. It is an index of torque non-uniformity. More specifically, it first calculates the abnormal distribution of each mode that reflects the physical synergistic effect. Then use This amplifier enhances the anomalous components of higher-order modes that typically indicate localized defects, and finally accumulates all weighted anomalous contributions. This generates a comprehensive index that accurately captures early, localized cable health degradation with specific physical characteristics. The resulting weighted co-anomaly index provides a near-exponential amplification response to co-deviations in higher-order modes indicating localized defects and exhibiting torsional hardening characteristics. Therefore, compared to the traditional L2 norm index, the weighted co-anomaly index produces significant and clear numerical jumps in the very early stages of cable fault occurrence, achieving detection performance and diagnostic specificity far exceeding traditional methods. This provides a strong basis for developing precise alarm strategies and predictive maintenance plans.

[0040] Through the technical means of the above-mentioned preferred embodiments, the introduced modal sensitivity weighting mechanism can effectively amplify the sensitivity of higher-order modes to local defects, enabling the system to capture early and weak fault signals that are easily overlooked by traditional methods, thus greatly improving the sensitivity of fault detection. Simultaneously, the design of the physical co-anomaly scoring function allows the method to no longer merely assess the magnitude of deviations, but to identify the patterns of deviations, particularly providing specific enhancement to the frequency-damped double positive deviation pattern that indicates the risk of kink hardening. This significantly improves the accuracy and specificity of diagnosis, effectively distinguishing different types of health degradation. Finally, the aggregated weighted co-anomaly index, as a comprehensive health indicator, exhibits a clear degradation trend in the early stages of fault evolution due to its sensitive amplification effect on key fault modes, providing a possibility for shifting from passive response to proactive prediction in maintenance. In summary, the technical objective of this method is to construct a cable health assessment system with clear physical meaning, high sensitivity to key fault modes, and early warning capabilities. The achieved technical effect is a significant improvement in the early warning capability, diagnostic accuracy, and overall reliability of the monitoring system, ultimately serving the overall goal of improving the operating efficiency of wind turbine generators and reducing maintenance costs.

[0041] Specifically, in step S500, an alarm signal is generated based on a comparison between the torque non-uniformity index and a preset static threshold. It should be understood that since the torque non-uniformity index calculated in the preceding steps is a continuous value quantifying the cable's health status, it does not directly constitute a decision instruction. The monitoring system needs a clear criterion to determine the risk level represented by this value. Therefore, in the technical solution of this application, an alarm signal is further generated based on a comparison between the torque non-uniformity index and a preset static threshold, thereby transforming the continuous health assessment results into discrete alarm levels with clear operational guidance. This closes the complete loop from data monitoring to maintenance decision-making, enabling automatic early warning of cable health degradation and providing the final execution basis for implementing condition-based maintenance and preventative maintenance strategies.

[0042] More specifically, in a particular example of this application, the process first relies on a pre-set multi-level static threshold system. This system includes at least a lower-level warning threshold and a higher-level alarm threshold. These thresholds are determined based on a comprehensive analysis of extensive historical operating data, simulation analysis, and an expert knowledge base of cable failure modes. Once the current torque non-uniformity index is calculated, the decision logic proceeds as follows: The index value is first compared with the highest-level alarm threshold. If the index value is greater than or equal to the alarm threshold, a serious health risk or impending failure of the cable is identified, and a high-level alarm signal is immediately generated. This signal is sent to the wind turbine's main controller to trigger necessary risk mitigation measures, such as limiting turbine power or performing a safe shutdown. If the index value does not reach the alarm threshold, it is then compared with a lower-level warning threshold. If the index value is greater than or equal to the warning threshold, the cable is determined to have entered an early degradation stage with potential risk, and a medium-level warning signal is generated. This warning signal is sent to the wind farm's operation and maintenance management system to prompt maintenance personnel to conduct a detailed inspection of the cable during future maintenance windows. If the indicator value is less than all preset thresholds, the cable is considered to be in normal condition, no alarm signal is generated, and the indicator value is only recorded in the historical database for long-term trend analysis.

[0043] In summary, the monitoring method for the torsional state of a pitch system cable according to the embodiments of this application is explained. It abandons the traditional approach of directly measuring local deformation or stress, instead monitoring the cable's vibration response and identifying its global modal parameters (such as modal frequency and damping). Since any local stiffness change at any point in the cable (such as kinking caused by torsional inhomogeneity) will cause a corresponding drift in its overall modal parameters, this solution achieves global state perception from a point-to-surface perspective, effectively solving the technical problem that traditional point-based sensing cannot assess overall health. Furthermore, this solution deeply mines the physical co-modes in the modal deviation vector, particularly performing specific diagnosis on the characteristic of frequency and damping increasing in the same direction caused by torsional hardening. This allows for accurate identification and quantification of the degree of torque inhomogeneity, ultimately achieving sensitive capture and reliable early warning of early local faults in the cable, overcoming the fundamental deficiency of existing technologies in effectively monitoring torsional inhomogeneity.

[0044] Furthermore, a monitoring system for the cable twisting status of a pitch system is also provided.

[0045] Figure 5 This is a block diagram of a monitoring system for the cable twisting condition of a pitch system according to an embodiment of this application. Figure 5 As shown, the pitch system cable torsion monitoring system 100 according to an embodiment of this application includes: a real-time data monitoring and acquisition module 110 for acquiring real-time pitch angle and real-time acceleration data; an online modal parameter identification module 120 for performing online modal parameter identification on the real-time pitch angle and real-time acceleration data to obtain online modal frequency and online modal damping; a state deviation calculation module 130 for calculating and evaluating the state deviation of the current pitch angle based on the online modal frequency, online modal damping, and baseline modal map to obtain a deviation vector; a torque non-uniformity diagnosis module 140 for diagnosing torque non-uniformity on the deviation vector to obtain a torque non-uniformity index; and an alarm module 150 for generating an alarm signal based on a comparison between the torque non-uniformity index and a preset static threshold.

[0046] As described above, the pitch system cable torsion monitoring system 100 according to the embodiments of this application can be implemented in various wireless terminals, such as servers with pitch system cable torsion monitoring algorithms. In one possible implementation, the pitch system cable torsion monitoring system 100 according to the embodiments of this application can be integrated into the wireless terminal as a software module and / or hardware module. For example, the pitch system cable torsion monitoring system 100 can be a software module in the operating system of the wireless terminal, or it can be an application developed for the wireless terminal; of course, the pitch system cable torsion monitoring system 100 can also be one of many hardware modules of the wireless terminal.

[0047] The various embodiments of this disclosure have been described above. These descriptions are exemplary and not exhaustive, nor are they limited to the disclosed embodiments. Many modifications and variations will be apparent to those skilled in the art without departing from the scope and spirit of the described embodiments. The terminology used herein is chosen to best explain the principles, practical application, or improvement of the technology in the market, or to enable others skilled in the art to understand the embodiments disclosed herein.

Claims

1. A method for monitoring the cable twist condition of a pitch system, characterized in that, include: Acquire real-time pitch angle and real-time acceleration data; Online modal parameter identification is performed on real-time pitch angle and real-time acceleration data to obtain online modal frequencies and online modal damping; Based on online modal frequencies, online modal damping, and baseline modal maps, the state deviation of the current pitch angle is calculated and evaluated to obtain the deviation vector; Torque non-uniformity is diagnosed on the deviation vector to obtain torque non-uniformity index; An alarm signal is generated based on the comparison between the torque non-uniformity index and the preset static threshold.

2. The method for monitoring the cable twisting state of a pitch system according to claim 1, characterized in that, Online modal parameter identification is performed on real-time pitch angle and real-time acceleration data to obtain online modal frequencies and online modal damping, including: Signal preprocessing and cross-power spectral density matrix construction are performed on real-time acceleration data to obtain the cross-power spectral density matrix; Singular value decomposition and modal peak identification are performed on the cross power spectral density matrix to obtain singular value spectral data and a list of peak frequencies; Accurate estimation of modal frequencies and damping ratios is performed based on singular value spectral data and a list of peak frequencies to obtain online modal frequencies and online modal damping.

3. The method for monitoring the cable twisting state of a pitch system according to claim 2, characterized in that, Accurate estimation of modal frequencies and damping ratios is performed based on singular value spectral data and a peak frequency list to obtain online modal frequencies and online modal damping, including: Modal peak isolation and free decay response generation are performed on singular value spectral data and peak frequency list to obtain a set of autocorrelation functions; Logarithmic decay rate calculation and modal damping ratio identification are performed on the autocorrelation function set to obtain the damped parameter set; Accurate modal frequency calculation and parameter integration are performed on the damped parameter set to obtain online modal frequencies and online modal damping.

4. The method for monitoring the cable twisting state of a pitch system according to claim 3, characterized in that, The damped parameter set includes modal damping ratio and damped vibration frequency.

5. The method for monitoring the cable twisting state of a pitch system according to claim 1, characterized in that, Based on online modal frequencies, online modal damping, and baseline modal maps, the state deviation of the current pitch angle is calculated and evaluated to obtain the deviation vector, including: Based on the baseline modal map, the current pitch angle is precisely interpolated to obtain interpolation reference parameters, which include interpolation healthy mode frequency and interpolation healthy mode damping. A deviation vector is obtained by calculating the state deviation of the interpolated healthy mode frequency and interpolated healthy mode damping, as well as the online mode frequency and online mode damping. The deviation vector includes the relative frequency deviation and the relative damping deviation.

6. The method for monitoring the cable twisting state of a pitch system according to claim 1, characterized in that, Torque non-uniformity is diagnosed by analyzing the deviation vector to obtain torque non-uniformity indices, including: Extract the total number of modes from the deviation vector; Based on the total number of modes, construct a modality sensitivity weight vector; Physical coordination anomaly scoring is performed on the deviation vector to obtain the physical coordination anomaly score vector; The modal sensitivity weight vector and the physical coordination anomaly score vector are weighted and aggregated to obtain a weighted coordination anomaly index, which serves as the torque non-uniformity index.

7. The method for monitoring the cable twisting state of a pitch system according to claim 6, characterized in that, The modal sensitivity weight vector and the physical coordination anomaly score vector are weighted and aggregated to obtain a weighted coordination anomaly index as the torque non-uniformity index. This includes weighting and aggregating the modal sensitivity weight vector and the physical coordination anomaly score vector using the following formula: in, It is the total number of modes. It is the modal sensitivity weight. It is a score for abnormal physical coordination. It is an indicator of torque non-uniformity.

8. A monitoring system for the cable twisting state of a pitch system, the system being capable of implementing the method described in any one of claims 1-7, characterized in that, include: The real-time data monitoring and acquisition module is used to acquire real-time pitch angle and real-time acceleration data; The online modal parameter identification module is used to perform online modal parameter identification on real-time pitch angle and real-time acceleration data to obtain online modal frequencies and online modal damping; The state deviation calculation module is used to calculate and evaluate the state deviation of the current pitch angle based on online modal frequency, online modal damping and baseline modal map to obtain the deviation vector; The torque non-uniformity diagnosis module is used to diagnose torque non-uniformity in the deviation vector to obtain torque non-uniformity index. The alarm module is used to generate alarm signals based on the comparison between the torque non-uniformity index and the preset static threshold.

Images