Wind power cabin bolt state detection method, detection system and detection equipment

Abstract

The invention relates to the technical field of industrial detection equipment, in particular to a wind power cabin bolt state detection method, system and equipment. An anti-electromagnetic interference signal line is used for constructing a signal loop, a to-be-detected bolt in a wind power cabin is physically connected with the anti-electromagnetic interference signal line, and whether the bolt is broken or not is judged by detecting a signal of the signal loop. Relative displacement of a bolt head and a flange face is monitored through a high-frequency laser displacement sensor, meanwhile, 3D modeling is conducted on a bolt through a laser radar point cloud, the modeling condition of the bolt through the point cloud is analyzed in real time, and looseness detection of the bolt is conducted through a bolt disturbance state detection algorithm based on laser radar point cloud data. According to the invention, micron-level displacement detection and second-level fracture response capability can be realized, through laser displacement and signal detection line dual-channel verification, the fault identification accuracy is improved, early warning of loosening, fracture and other problems is realized, and safe operation of a wind turbine generator is guaranteed.

Description

Technical Field

[0001] This invention relates to the field of industrial testing equipment technology, specifically to a method, system and equipment for detecting the condition of bolts in a wind turbine nacelle. Background Technology

[0002] As wind power equipment becomes increasingly large and complex, its operational stability and maintenance efficiency are particularly critical. Once a wind turbine fails, the cost of downtime repairs is high, the maintenance cycle is long, and power generation revenue is affected. How to improve fault detection accuracy, shorten downtime, reduce labor maintenance costs, and especially reduce unplanned downtime events has become a core challenge that wind power companies and operation and maintenance units urgently need to solve.

[0003] Under the combined effects of strong winds, start-up and shutdown vibrations, and alternating hot and cold temperatures, the connecting bolts inside the nacelle are at risk of failure, including slippage, loosening, and breakage. Currently, most wind turbines rely on manual torque checks or visual inspections, which cannot detect changes in bolt preload or fatigue crack propagation in real time, often only discovering the problem after a structural failure has occurred.

[0004] The breakage or detachment of a single bolt may cause the entire connection structure to loosen, leading to serious accidents such as blade detachment and bearing misalignment, affecting the safety and availability of the wind turbine, and even causing personal safety risks and significant economic losses.

[0005] Existing methods for detecting loose bolts or nuts on wind turbines include manual inspection and online detection technology using electronic systems. Manual inspection is inefficient, untimely, and time-consuming. Online monitoring technology, such as the electrical circuit-based fracture monitoring scheme described in Reference 1, offers a solution.

[0006] Reference 1: Chinese patent document with publication number CN 110778463 A.

[0007] Reference 1 describes a method and device for monitoring and protecting against bolt fracture of wind turbine blades. The wind turbine blades are mounted on the hub of the wind turbine and are fixedly connected to the hub by bolts. The method is characterized by setting up a closed circuit consisting of a power supply and a load connected by wires. When laying the wires, they pass through each wind turbine blade bolt and are fixed to the head of each wind turbine blade bolt with fasteners. A current detection device is set in the closed circuit, and the detected current signal is sent to the wind turbine main control in a logical expression of "1" and "0". When any bolt breaks, the broken bolt head falls out of the bolt hole in the hub under its own weight and the inertia of the wind turbine, breaking off the bolt fixed to it. The current in the closed circuit disappears, and the signal logic sent by the current detection device to the wind turbine main control is "0". The wind turbine main control then controls the wind turbine to stop, avoiding major accidents and ensuring the safe operation of the wind turbine.

[0008] However, while the above-mentioned technical solution can detect bolt fracture, it still has some limitations: First, the solution can only respond to the complete fracture state and has no ability to detect early loosening (micrometer-level displacement); second, the solution relies on rigid optical fibers to construct the circuit, which has poor resistance to bending fatigue and is difficult to adapt to the vibration environment of the nacelle; furthermore, the solution lacks an anti-interference mechanism and is susceptible to false triggering due to blade static electricity and broadband noise interference from the frequency converter. These limitations share the common characteristics of lacking quantitative sensing capability for loosening states and insufficient environmental robustness. Summary of the Invention

[0009] The purpose of this invention is to solve the above-mentioned problems existing in the prior art and to provide a method, system and equipment for detecting the condition of wind turbine nacelle bolts.

[0010] To address the shortcomings of the aforementioned technical problems, the present invention provides a method for detecting the condition of wind turbine nacelle bolts, including detection of bolt breakage and loosening.

[0011] Fracture detection:

[0012] A signal loop is constructed using an anti-electromagnetic interference signal line. The bolts to be tested inside the wind turbine nacelle are physically connected to the anti-electromagnetic interference signal line. The signal from the detection signal loop is used to determine whether a bolt is broken. When a bolt is broken, causing the anti-electromagnetic interference signal line to disconnect, the signal changes accordingly, indicating that a bolt is broken.

[0013] The system acquires environmental noise signals in real time using a high-speed ADC, performs spectrum analysis using window function weighting and FFT module, dynamically identifies the concentrated frequency bands of motor harmonics, and implements automatic spectrum avoidance when a specific motor characteristic frequency is detected to eliminate the interference of electrostatic interference from fan blades and broadband noise from variable frequency motors on the detection results.

[0014] The phase of the output signal is kept synchronized with the phase of the input reference signal by using a phase-locked loop (PLL). A Kalman frequency predictor is built into the PLL to store the most recent 32 frequency points, construct a state prediction vector, and eliminate noise through Kalman filtering to quickly lock the PLL.

[0015] Looseness detection:

[0016] Bolts are modeled in 3D using LiDAR point clouds, and the modeling of bolts by point clouds is analyzed in real time. Bolt loosening is detected using a bolt disturbance state detection algorithm based on LiDAR point cloud data.

[0017] For the same bolt, the distance measurement point cloud set is clustered into 2-3 point cloud sets. The average of the distance measurement data of all point clouds in the set is used to obtain the effective distance measurement data of the current set. When the effective distance measurement data or distance data of the 2-3 point cloud sets are equal or the error value is extremely small, the average of the point cloud cluster sets is calculated separately. The change of the effective distance measurement data over time is the displacement data. The bolt is loosened by using the displacement data.

[0018] The frequency sweep nonlinearity of the DFB laser is optimized by adopting a current-temperature co-modulation method. Combined with the real-time acquisition of grating feedback signal by ADC, the frequency-time curve is dynamically corrected. Combined with real-time phase demodulation algorithm and multi-dimensional error compensation mechanism, the frequency sweep nonlinearity error is suppressed.

[0019] Cross-correlation analysis was performed on the output signals of the reference interferometer and the measuring interferometer. A "vernier effect" model was constructed by extracting the phase difference of the dual-channel interference signals. Combined with the improved CZT spectrum refinement technology, the phase resolution was improved.

[0020] As a further optimization of the wind turbine nacelle bolt condition detection method of the present invention, the specific method of automatic spectrum avoidance is as follows:

[0021] S1. Add a Hanning window to the acquired signal. The specific formula is as follows:

[0022]

[0023] Where 0 ≤ n ≤ N-1;

[0024] Fast algorithms for calculating the Discrete Fourier Transform of a signal:

[0025]

[0026] The sampling point is 1024;

[0027] S2. Define μ as the average energy of background noise and σ as the standard deviation of background noise. Detect energy peaks at integer multiples of the fundamental frequency f0, f0·k, k=1,2,3,... Define an adaptive threshold as Threshold=μ+3σ. If three consecutive harmonic peaks > Threshold, it is determined to be a motor interference frequency band.

[0028] S3. Implement a spectrum avoidance strategy, specifically the frequency jump: f new =f current +Δf (Δf > interference bandwidth), dynamically adjust the center frequency of the bandpass filter to f. new .

[0029] As a further optimization of the wind turbine nacelle bolt condition detection method of the present invention: the Kalman predictive phase-locked loop specifically involves: establishing a Kalman filter spatial soft-state model, and recursively calculating the Kalman gain K using Kalman filtering. k And calculate the voltage phase-locked loop control (VCO) using the following formula: V ctrl (k)=K p ·θ e (k)+K i ∑θ e (k), where θ e (k) represents the phase error output by the phase detector.

[0030] As a further optimization of the wind turbine nacelle bolt condition detection method of the present invention: the phase detector output is monitored in real time, and when a phase change caused by an electrostatic pulse is detected, a compensation current is automatically injected to cancel the disturbance and maintain the continuity of the phase-locked state, specifically as follows:

[0031] Mutation detection is achieved using difference, when Δθ=|θ e (t)-θ e When (t-1)|>30°, the compensation current is used for calculation; I comp =K c ·sgn(Δθ)·e -α∣Δθ∣ ;

[0032] Where α is the attenuation coefficient, and α is taken as 0.05; K c To compensate for the gain;

[0033] The corrected VCO input is: V ctrl′ (t)=V ctrl (t)+β·I comp , where β is the current-to-voltage conversion coefficient.

[0034] As a further optimization of the wind turbine nacelle bolt condition detection method of the present invention, the bolt disturbance condition detection algorithm based on lidar point cloud data is as follows:

[0035] S1. Locating the core point: Collect the point cloud data. In the k-th frame, the point cloud set for the bolt region is P. k ={p k,1 ,p k,2 ,…,p k,m}, where p k,i =(x k,i ,y k,i ,z k,i ) represents the three-dimensional coordinates of the point, and m represents the total number of point clouds in this frame;

[0036] For point p∈P k Its ε-neighborhood is defined as: N ε (p)={q∈Pk |dist(p,q)≤ε};

[0037] Where dist(p,q) is the three-dimensional Euclidean distance. ε represents the neighborhood radius, which is a preset parameter;

[0038] If the number of points contained in the ε-neighborhood of point P satisfies |Nε(p)|≥MinPts, then p is a core point;

[0039] Wherein, MinPts represents the minimum number of points threshold, which is a preset parameter;

[0040] S2. Noise Determination and Point Cloud Filtering: If point P does not belong to any core point and is not within the ε-neighborhood of any core point, then p is a noise point, denoted as p∈Noise; after DBSCAN denoising in the k-th frame, the effective point cloud set of the bolt surface is: Among them, Noise k Let be the set of noise points in the k-th frame;

[0041] S3. Bolt surface point density and growth rate, and anomaly detection: In the k-th frame, the effective point cloud quantity of the i-th bolt is: Let i be a subset of the point cloud of the i-th bolt;

[0042] The point cloud increment is defined as follows: the increase in point cloud size of the i-th bolt in frame k relative to frame (k-1) is Δn. i,k =n i,k -n i,k-1 Calculate the cumulative average growth of the i-th bolt from the initial frame to the (k-1)-th frame. Define the ratio of the growth of the i-th bolt in the K-th frame to the cumulative average growth.

[0043] If r i,j ≤α or r i,j If ≥β, then the i-th bolt is determined to be in t. k Displacement anomalies exist at all times;

[0044] Where α is the lower threshold and β is the upper threshold, and α and β are set according to historical data of the bolt under stable conditions.

[0045] As a further optimization of the wind turbine nacelle bolt condition detection method of the present invention, the method for suppressing frequency sweep nonlinear error specifically includes the following steps:

[0046] S1, System Initialization

[0047] Set the initial current I0 and temperature T0, and load the calibration parameters;

[0048] S2, Signal Acquisition and Processing

[0049] The ADC acquires the grating feedback signal in real time with a sampling rate ≥100MSa / s;

[0050] S3, Real-time Phase Demodulation

[0051] Window function optimization and real-time FFT are used to accelerate computation:

[0052]

[0053] The instantaneous phase is calculated as follows:

[0054]

[0055] S4. Frequency Error Calculation

[0056] e f (t)=f ideal (t)-f actual (t)

[0057] S5, Multi-dimensional Error Compensation

[0058] The error model is defined as follows:

[0059] Calculate the comprehensive compensation terms for temperature drift, current fluctuation, etc., using the following formula:

[0060] I comp (t)=I0(t)+K I ·E(t)

[0061] T comp (t)=T0(t)+K T ·E(t)

[0062] S6, Current-Temperature Co-modulation

[0063] Define the cooperative control equations, calculate the modulation of current and temperature, and calculate ΔI and ΔT based on the control matrix:

[0064]

[0065] Among them, e f (t) represents the frequency error, e φ (t) represents the phase error;

[0066] S7, Dynamically Update Parameters

[0067] The algorithm parameters are adaptively adjusted according to environmental changes;

[0068] S8, Frequency-Time Curve Correction

[0069] Calculate the corrected actual frequency: f actual (t)=f ideal (t)+Δf comp (t)

[0070] The compensation item is calculated as follows:

[0071] Legendre orthogonal polynomials are used instead of traditional polynomial fitting to avoid matrix ill-conditioning and improve fitting stability.

[0072] Update the output frequency curve and return to step S2.

[0073] As a further optimization of the wind turbine nacelle bolt condition detection method of the present invention: in the method of improving phase resolution, phase-locked loop or Kalman filtering technology is used to compensate for the phase noise of the local oscillator and the signal source, that is, compensation is performed for each frame of signal, and the compensation accuracy is found to be optimal in successive iterations; at the same time, a window function is added, and after each complete cycle of signal is acquired, Fourier transform is performed on the periodic signal to eliminate interference frequency bands.

[0074] The present invention also provides a wind turbine nacelle bolt condition detection system, including a data application layer, a data transmission layer, a data acquisition layer and a data perception layer;

[0075] The data application layer includes cloud servers and field terminals. The field terminals are responsible for providing users with an entry point for real-time monitoring data access. Users can log in by entering their account and password, and then remotely monitor the real-time status of bolts inside the wind turbine nacelle through a human-computer interaction display interface.

[0076] The data transmission layer is responsible for transmitting various parameter data collected by the data acquisition layer to the data application layer via wireless signals;

[0077] The data acquisition layer includes several data acquisition devices, including a bolt breakage detection module and a bolt loosening detection module. The bolt breakage detection module determines whether a bolt is broken by detecting the signal in the signal circuit, and the bolt loosening detection module determines whether a bolt is loose by using a laser displacement sensor and a lidar.

[0078] The data sensing layer includes several groups of detection units, the same number as the data acquisition equipment. Each group of detection units includes several detection units, and the detection units in the same group can be connected in series to form a signal loop.

[0079] The present invention also provides a wind turbine nacelle bolt condition detection device, including a cloud server, a station terminal, an AP gateway device, several detection hosts and several detection switches;

[0080] The cloud server includes a data center module, an intelligent analysis and decision-making module, and a device management and collaboration module;

[0081] The station terminal includes a human-computer interaction module and a communication module;

[0082] The AP gateway device includes a wireless WIFI module, a 4G module, a LoRa module, and an RS-485 interface module;

[0083] The detection host includes a laser emission and modulation module, an optical path and scanning module, a receiving and detection module, a scanning control and positioning module, a data processing and control module, and a power supply module;

[0084] The detection switch includes a housing, a detection signal line, and springs. The housing is a cylindrical structure, and several springs are arranged inside the housing. One end of the spring is fixedly connected to the inner wall of the housing, and the other end of the spring is a free end. A bolt clamping cavity is formed between the several springs. The detection signal line has a loop part and two terminals. The loop part is embedded in the housing, and the two terminals are used to connect to adjacent detection switches or detection host.

[0085] As a further optimization of the wind turbine nacelle bolt condition detection device of the present invention, the site terminal also includes an identity authentication and login module and an access control module.

[0086] As a further optimization of the wind turbine nacelle bolt condition detection device of the present invention: the end face of the spring facing the bolt clamping cavity is provided with a locking wing, the locking wing is perpendicular to the axis of the shell, and the outer edge of the locking wing has a sawtooth structure.

[0087] As a further optimization of the wind turbine nacelle bolt condition detection device of the present invention: the shell is made of PA66 plastic or PC plastic, and the spring is made of spring steel with a thickness of 1mm.

[0088] The present invention has the following beneficial effects:

[0089] I. This invention enables micron-level displacement detection and second-level fracture response, supporting 24 / 7 online monitoring of critical bolts. Through dual-channel verification using laser displacement and signal detection lines, it improves fault identification accuracy, provides early warning of loosening and breakage issues, and ensures the safe operation of wind turbine units.

[0090] II. Laser Displacement Detection: By installing high-precision laser displacement sensors near critical bolts, real-time monitoring of minute displacement changes is achieved. When a bolt becomes loose, develops microcracks, or breaks, changes in its structural stress will trigger minute displacement changes. The laser sensor can quickly capture these displacement changes and trigger an alarm. The system uses an embedded processing unit to filter and threshold the sensor data, combined with a sudden change detection algorithm, effectively reducing the false alarm rate caused by environmental interference. The laser detection system can operate independently or be deployed in conjunction with enameled wire breakage monitoring technology to build a redundant monitoring system, further improving reliability and early warning capabilities.

[0091] Third, based on traditional laser interferometric ranging technology, this system achieves high-precision, wide-range, and robust ranging capabilities in dynamic environments by integrating a frequency-sweeping light source, a real-time phase demodulation algorithm, and a multi-dimensional error compensation mechanism.

[0092] IV. This invention uses LiDAR point clouds to create 3D models of bolts and analyzes the bolt modeling in real time. The point cloud data is sampled at 10 frames per second, and then outliers are removed from the sparse point cloud using an adaptive filtering algorithm to eliminate noise interference. Attached Figure Description

[0093] Figure 1 This is a logic block diagram of the automatic spectrum avoidance strategy in the detection method;

[0094] Figure 2 The logic block diagram of the Kalman predictive phase-locked loop in the detection method is shown below.

[0095] Figure 3 This is a schematic diagram of the detection system.

[0096] Figure 4 This is a schematic diagram of the detection switch in the detection equipment;

[0097] Figure 5 A schematic diagram of the structure for detecting the installation status of the switch;

[0098] Marked in the image:

[0099] 1. Shell;

[0100] 2. Detection signal line;

[0101] 201. Ring-shaped part;

[0102] 202. Connector;

[0103] 3. Shrapnel;

[0104] 4. Lock the winglets. Detailed Implementation

[0105] To better understand the present invention, the following embodiments further illustrate the content of the present invention, but the content of the present invention is not limited to the following embodiments.

[0106] <Methods for Inspecting the Condition of Bolts in Wind Turbine Nacelles>

[0107] Fracture detection:

[0108] A signal loop is constructed using an anti-electromagnetic interference signal line. The bolts to be tested inside the wind turbine nacelle are physically connected to the anti-electromagnetic interference signal line. The signal from the detection signal loop is used to determine if a bolt is broken. When a bolt is broken, causing the anti-electromagnetic interference signal line to disconnect, the signal changes accordingly, indicating that a bolt is broken.

[0109] Detection switches are installed on each bolt, and these switches are connected in series via an electromagnetic interference (EMI) shielded signal line, ultimately leading to the detection host. The EMI shielded signal line connects the detection switches into a continuous signal loop. The detection host determines the bolt's condition by sending a specific detection signal to this loop and monitoring the signal return. Under normal circumstances, the signal loop is intact, the detection signal transmits smoothly, and the detection host receives stable signal feedback. However, when a bolt breaks, the detection switch installed on that bolt is pulled apart due to the breakage, causing the EMI shielded signal line connected to it to disconnect. This disconnection of the EMI shielded signal line directly interrupts the signal loop. The detection host, after sending a detection signal, cannot receive the expected signal feedback, thus determining that an abnormal change has occurred in the signal loop.

[0110] The detection host is equipped with a high-precision signal detection circuit that can monitor the status of the signal loop in real time. Once a signal loop interruption is detected, the detection host will immediately trigger an alarm mechanism. The alarm information is uploaded to the cloud server via the AP gateway device. The cloud server's intelligent analysis and decision-making module further analyzes the alarm information to confirm the type and severity of the alarm. Subsequently, the cloud server displays the bolt breakage alarm information and the specific bolt location to maintenance personnel through the human-machine interface module of the site terminal. Maintenance personnel can then quickly locate the broken bolt based on this information and take appropriate maintenance measures, such as replacing the bolt or checking the connections of related components.

[0111] During the operation of wind turbine nacelle bolt condition detection equipment, the detection host needs to monitor the signal circuit in real time to determine whether the bolts have broken. However, the operating environment of wind turbine generators is complex, with various interference sources. Among them, electrostatic interference from the turbine blades and broadband noise from the variable frequency motor are two common interference factors. These interferences may affect the signal detection results of the detection host, leading to misjudgments in continuity detection.

[0112] During high-speed rotation, wind turbine blades generate static electricity through friction with the air. This static electricity accumulates on the blade surface, forming electrostatic charges. When the static electricity accumulates to a certain level, it may discharge through the electromagnetic interference (EMI) suppression signal lines or other metal components within the wind turbine nacelle. Since the signal circuit of the detection unit is connected via EMI suppression signal lines, these lines can become pathways for electrostatic discharge. Electrostatic discharge generates instantaneous high-voltage pulses that propagate along the EMI suppression signal lines and enter the signal detection circuit of the detection unit, thus interfering with signal detection. During signal detection, the detection unit typically sends a low-voltage detection signal and monitors the continuity of the signal circuit. If electrostatic interference generated by the wind turbine blades enters the signal circuit, it may cause the detection signal to be momentarily pulled high or low. For example, when a high-voltage pulse generated by electrostatic discharge is superimposed on the detection signal, the detection unit may misjudge the state of the signal circuit. If the detection signal is pulled high, the detection unit may mistakenly believe that the signal circuit is conductive, even if a bolt has actually broken, causing the EMI suppression signal line to disconnect. Conversely, if the detection signal is pulled low, the detection host may mistakenly interpret it as a broken signal loop, even if the bolt is not actually broken. This misjudgment can lead to maintenance personnel receiving incorrect alarm messages, thus affecting the normal maintenance and operation of the equipment.

[0113] The variable frequency motor in a wind turbine generator set is an indispensable component during operation, controlling its speed and power through a frequency converter. The frequency converter works by converting direct current (DC) to alternating current (AC) and regulating voltage and frequency using high-frequency switching elements (such as IGBTs). This high-frequency switching operation generates a significant amount of electromagnetic noise over a wide frequency range, commonly referred to as broadband noise. Broadband noise can propagate to other equipment within the wind turbine nacelle and to electromagnetic interference suppression signal lines through electromagnetic induction, conduction, or radiation.

[0114] The electromagnetic interference-resistant signal lines in the signal circuit of the detection host may be affected by the broadband noise of the variable frequency motor. When broadband noise couples into the signal circuit, it superimposes a series of high-frequency noise components onto the detection signal. These noise components interfere with the detection host's judgment of the signal's on / off state. For example, high-frequency noise may cause changes in the amplitude and frequency of the detection signal, making it impossible for the detection host to accurately identify the true state of the signal circuit. If the noise amplitude is large, the detection host may misjudge the signal circuit as open, even if the signal circuit is actually conductive. This misjudgment will cause the detection host to incorrectly issue a bolt breakage alarm, when the bolt has not actually broken. Conversely, if the noise causes a decrease in the amplitude of the detection signal, the detection host may misjudge the signal circuit as conductive, thus missing the bolt breakage fault.

[0115] To address the interference issues caused by electrostatic discharge from wind turbine blades and broadband noise from variable frequency motors, a program with the following data processing methods is built into the detection host:

[0116] The system acquires environmental noise signals in real time using a high-speed ADC, performs spectrum analysis using window function weighting and FFT modules, dynamically identifies the concentrated frequency bands of motor harmonics, and implements automatic spectrum avoidance when a specific motor characteristic frequency is detected to eliminate the interference of electrostatic interference from fan blades and broadband noise from variable frequency motors on the detection results.

[0117] like Figure 1 As shown, the specific method for automatic spectrum avoidance is as follows:

[0118] S1. Add a Hanning window to the acquired signal. The specific formula is as follows:

[0119]

[0120] Where 0 ≤ n ≤ N-1.

[0121] Fast algorithms for calculating the Discrete Fourier Transform of a signal:

[0122]

[0123] The number of sampling points is 1024.

[0124] S2. Define μ as the average energy of background noise and σ as the standard deviation of background noise. Detect energy peaks at integer multiples of the fundamental frequency f0, f0·k, k=1,2,3,... Define an adaptive threshold as Threshold=μ+3σ. If three consecutive harmonic peaks > Threshold, it is determined to be a motor interference frequency band.

[0125] S3. Implement a spectrum avoidance strategy, specifically the frequency jump to f. new =f current +Δf (Δf > interference bandwidth), dynamically adjust the center frequency of the bandpass filter to f. new .

[0126] After calculating the interference frequency band, the circuit board's ADC module automatically switches to the effective frequency band that eliminates the interference frequency band based on the frequency adjustment of the embedded software.

[0127] A phase-locked loop (PLL) is used to keep the phase of the output signal synchronized with the phase of the input reference signal. A Kalman frequency predictor is built into the PLL to store the most recent 32 frequency points and construct a state prediction vector. The PLL is then quickly locked by a Kalman filter to reduce the phase-locking time.

[0128] like Figure 2As shown, the specific method for noise elimination using Kalman filtering is as follows: Establish a spatial soft-state model of the Kalman filter, and recursively calculate the Kalman gain K using Kalman filtering. k And calculate the voltage phase-locked loop control (VCO) using the following formula: V ctrl (k)=K p ·θ e (k)+K i ∑θ e (k), where θ e (k) represents the phase error output by the phase detector.

[0129] The construction process of the Kalman filter spatial soft-state model is as follows:

[0130] S1. Divide the continuous spatial domain into discrete grid points, with each grid point representing a spatial location.

[0131] S2. Mesh generation is performed using Finite Difference (FDM).

[0132] S3. Establish the state equations:

[0133] Equation 1: x k =F k x k-1 +B k u k +w k .

[0134] x k Let be the state vector at time k.

[0135] F k Let be the state transition matrix, which describes the evolution of the state from time k-1 to time k.

[0136] B k The control input matrix describes the effect of the control input uk on the state.

[0137] w k This is process noise.

[0138] Equation 2: x k,i =∑ j∈N(i) a ij x k-1,j +w k,i .

[0139] x k,i This represents the state of the i-th grid point at time k.

[0140] N(i) is the set of neighboring grid points of the i-th grid point.

[0141] a ijis the spatial autoregressive coefficient, describing the influence of the j-th grid point on the i-th grid point.

[0142] w k,i Let be the phase error of the i-th grid point at time k.

[0143] S4. Establish the observation model:

[0144] y k =H k x k +v k .

[0145] y k Let be the observation vector at time k.

[0146] H k The observation matrix describes how the observation data is obtained from the state vector.

[0147] v k For observation noise, describe the random errors in the observation process.

[0148] S5. Establish a spatial soft-state model:

[0149]

[0150] x k，i This represents the state of the i-th grid point at time k.

[0151] N(i) is the set of neighboring grid points of the i-th grid point.

[0152] a ij is the spatial autoregressive coefficient, describing the influence of the j-th grid point on the i-th grid point.

[0153] The Kalman filter algorithm estimates the system state recursively, including two steps: prediction and update. After obtaining the phase error using the Kalman filter algorithm, regression optimization methods can be used to optimize the spatial soft-state model, thereby improving the model's accuracy.

[0154] The phase detector output is monitored in real time. When a phase change caused by an electrostatic pulse is detected, a compensation current is automatically injected to cancel the disturbance and maintain the continuity of the phase-locked state. Specifically:

[0155] Mutation detection is achieved using difference, when Δθ=|θ e (t)-θ e When (t-1)|>30°, the compensation current is used for calculation; I comp =K c ·sgn(Δθ)·e -α∣Δθ∣ .

[0156] Where α is the attenuation coefficient, and α is taken as 0.05; K c To compensate for the gain.

[0157] The corrected VCO input is: V ctrl′ (t)=V ctrl (t)+β·I comp , where β is the current-to-voltage conversion coefficient.

[0158] Looseness detection:

[0159] Bolts are modeled in 3D using LiDAR point cloud data. The modeling of bolts using point cloud data is analyzed in real time. Bolt loosening is detected using a bolt disturbance state detection algorithm based on LiDAR point cloud data.

[0160] The algorithm for detecting bolt disturbance status based on lidar point cloud data is as follows:

[0161] S1. Locating the core point: Collect the point cloud data. In the k-th frame, the point cloud set for the bolt region is P. k ={p k,1 ,p k,2 ,…,p k,m}, where p k,i =(x k,i ,y k,i ,z k,i ) represents the three-dimensional coordinates of the point, and m represents the total number of point clouds in this frame;

[0162] For point p∈P k Its ε-neighborhood is defined as: N ε (p)={q∈P k |dist(p,q)≤ε};

[0163] Where dist(p,q) is the three-dimensional Euclidean distance. ε represents the neighborhood radius, which is a preset parameter;

[0164] If the number of points contained in the ε-neighborhood of point P satisfies |Nε(p)|≥MinPts, then p is a core point;

[0165] MinPts represents the minimum number of points threshold, which is a preset parameter.

[0166] S2. Noise Determination and Point Cloud Filtering: If point P does not belong to any core point and is not within the ε-neighborhood of any core point, then p is a noise point, denoted as p∈Noise; after DBSCAN denoising in the k-th frame, the effective point cloud set of the bolt surface is: Among them, Noise k Let be the set of noise points in the k-th frame.

[0167] S3. Bolt surface point density and growth rate, and anomaly detection: In the k-th frame, the effective point cloud quantity of the i-th bolt is: Let i be a subset of the point cloud for the i-th bolt.

[0168] The point cloud increment is defined as follows: the increase in point cloud size of the i-th bolt in frame k relative to frame (k-1) is Δn. i,k =n i,k -n i,k-1 Calculate the cumulative average growth of the i-th bolt from the initial frame to the (k-1)-th frame. Define the ratio of the growth of the i-th bolt in the K-th frame to the cumulative average growth.

[0169] If r i,j ≤α or r i,j If ≥β, then the i-th bolt is determined to be in t. k Displacement anomalies are always present.

[0170] Where α is the lower threshold and β is the upper threshold, and α and β are set according to historical data of the bolt under stable conditions.

[0171] For the same bolt, the distance measurement point cloud set is clustered into 2-3 point cloud sets. The average of the distance measurement data of all point clouds in the set is used to obtain the effective distance measurement data of the current set. When the effective distance measurement data or distance data of the 2-3 point cloud sets are equal or the error value is extremely small, the average of the point cloud cluster sets is calculated separately. The change of the effective distance measurement data over time is the displacement data. The bolt loosening is determined by this displacement data.

[0172] To correct point cloud data drift, the data in non-dense areas of the point cloud generally drifts and has a large error, forming ghost images. A filtering algorithm should be used to filter out the point cloud data in non-dense areas.

[0173] A current-temperature co-modulation method is adopted to optimize the sweep frequency nonlinearity of the DFB laser. Combined with the real-time acquisition of the grating feedback signal by ADC, the frequency-time curve is dynamically corrected. Combined with the real-time phase demodulation algorithm and multi-dimensional error compensation mechanism, the sweep frequency nonlinearity error is suppressed.

[0174] Continuously collect hardware sweep frequency signals at room temperature in a constant temperature chamber to obtain initial values ​​or correct sweep frequency linear signals.

[0175] Establish the correlation curve between temperature change and frequency sweep signal per unit time as the initial value for compensation.

[0176] The Bragg grating inside a DFB laser reflects light of a specific wavelength, forming a grating feedback signal. An analog-to-digital converter (ADC) is used to acquire this grating feedback signal in real time, converting it from an analog signal to a digital signal.

[0177] The acquired grating feedback signal data is used to monitor the actual frequency changes of the laser in real time. Data analysis is then used to determine whether the frequency changes deviate from the target frequency-time curve.

[0178] The grating feedback signal acquired by the ADC is used to calculate the deviation between the actual frequency and the target frequency in real time. The frequency-time curve is dynamically corrected by adjusting the current and temperature control signals.

[0179] Dynamic calibration requires an efficient algorithm capable of quickly and accurately calculating the adjustment amount of the control signal. Commonly used algorithms include PID control algorithms and fuzzy control algorithms.

[0180] Frequency prediction is performed using an improved Kalman filter algorithm:

[0181] Let the state equation be: X k =A·X k-1 +B·u k +w k ;

[0182] The observation equation is: Z k =H·X k +v k ;

[0183] in: Z is the state vector (frequency and rate of change of frequency). k Observations acquired by the ADC;

[0184] w k ~N(0,Q), v k ~N(0,R) represents process noise and observation noise, and A,B,H represent state transition matrix, control matrix, and observation matrix, respectively.

[0185] The method for suppressing sweep frequency nonlinearity error specifically includes the following steps:

[0186] S1, System Initialization

[0187] Set the initial current I0 and temperature T0, and load the calibration parameters;

[0188] S2, Signal Acquisition and Processing

[0189] The ADC acquires the grating feedback signal in real time with a sampling rate ≥100MSa / s;

[0190] S3, Real-time Phase Demodulation

[0191] Window function optimization and real-time FFT are used to accelerate computation:

[0192]

[0193] The instantaneous phase is calculated as follows:

[0194]

[0195] S4. Frequency Error Calculation

[0196] e f (t)=f ideal (t)-f actual (t)

[0197] S5, Multi-dimensional Error Compensation

[0198] The error model is:

[0199] Calculate the comprehensive compensation terms for temperature drift, current fluctuation, etc., using the following formula:

[0200] I comp (t)=I0(t)+K I ·E(t)

[0201] T comp (t)=T0(t)+K T ·E(t)

[0202] S6, Current-Temperature Co-modulation

[0203] Define the cooperative control equations, calculate the modulation of current and temperature, and calculate ΔI and ΔT based on the control matrix:

[0204]

[0205] Among them, e f (t) represents the frequency error, e φ (t) represents the phase error;

[0206] S7, Dynamically Update Parameters

[0207] The algorithm parameters are adaptively adjusted according to environmental changes;

[0208] S8, Frequency-Time Curve Correction

[0209] Calculate the corrected actual frequency: f actual (t)=f ideal (t)+Δf comp (t)

[0210] The compensation item is calculated as follows:

[0211] Legendre orthogonal polynomials are used instead of traditional polynomial fitting to avoid matrix ill-conditioning and improve fitting stability.

[0212] Update the output frequency curve and return to step S2.

[0213] Cross-correlation analysis was performed on the output signals of the reference interferometer and the measuring interferometer. A "vernier effect" model was constructed by extracting the phase difference of the dual-channel interference signals. Combined with CZT spectrum refinement technology, the phase resolution was improved.

[0214] By performing cross-correlation analysis on the output signals of the reference interferometer and the measurement interferometer, the phase difference information between the two is extracted, and a vernier effect model is constructed using the phase difference information obtained from the cross-correlation analysis.

[0215] The signal under test is mixed with a local oscillator optical signal with a slightly different frequency to generate a lower intermediate frequency (IF) signal. The phase of this IF signal contains the original phase information φ(t). The phase is then extracted from the IF signal using techniques such as phase-locked loop (PLL) or quadrature demodulation. The phase difference represents a "vernier".

[0216] Phase-locked loop (PLL) or Kalman filtering techniques are used to compensate for the phase noise of the local oscillator and the signal source. This involves compensating for each frame of signal, with the compensation accuracy iteratively found to achieve the optimal value. Simultaneously, a window function is added, and after acquiring a complete signal cycle, a Fourier transform is performed on the cycle signal to eliminate interference frequency bands.

[0217] <Wind Turbine Nacelle Bolt Condition Inspection System>

[0218] like Figure 3 As shown, this detection system consists of four core parts: data application layer, data transmission layer, data acquisition layer, and data perception layer. These layers cooperate with each other to achieve comprehensive monitoring and management of the bolt status inside the wind turbine nacelle.

[0219] The data application layer includes cloud servers and site terminals. The site terminals provide users with an entry point for real-time monitoring data access. Users can log in by entering their account and password to remotely monitor the real-time status of bolts inside the wind turbine nacelle through a human-computer interaction display interface.

[0220] The data application layer is located at the top layer of the entire monitoring system. It serves as the user interface and the core area for data processing and storage. It mainly consists of two parts: a cloud server and a site terminal. The site terminal is the front-end device for user operation, providing a convenient access point for real-time monitoring data. Users simply need to enter a pre-set account and password, and after system verification, they can successfully log in. After logging in, users can remotely monitor the real-time status of bolts inside the wind turbine nacelle through a human-machine interface.

[0221] The human-computer interaction interface can present various monitoring data of bolts, including key information such as the degree of looseness and whether the bolts are broken. Users can quickly understand the health status of each bolt through various charts, data tables, and real-time images on the interface.

[0222] The site terminal also has data storage capabilities, enabling local saving of users' historical operation records and monitoring data for easy access and comparative analysis. The cloud server, acting as the "brain" of the entire data application layer, possesses powerful data processing and storage capabilities. After receiving various monitoring data uploaded from the data transmission layer, the cloud server analyzes and processes it. This processing includes data cleaning, filtering, classification, and fault diagnosis based on preset algorithms. Through these processes, the cloud server can accurately determine whether the bolts are in normal condition and generate corresponding early warning information in a timely manner. If a potential risk of bolt breakage or loosening is detected, the cloud server will immediately issue an alarm to the user through the site terminal, reminding the user to take appropriate maintenance measures. Simultaneously, the cloud server is responsible for long-term storage of all monitoring data and analysis results, providing rich data resources for subsequent data mining and equipment health management.

[0223] The data transmission layer is responsible for transmitting various parameter data collected by the data acquisition layer to the data application layer via wireless signals. The data transmission layer acts as a bridge between the data acquisition layer and the data application layer, and its main responsibility is to ensure that the collected parameter data can be transmitted stably and efficiently to the data application layer.

[0224] In the operating environment of wind power equipment, the reliability and stability of data transmission are crucial, as the nacelle is typically located at a high altitude and may be subject to electromagnetic interference. This layer adopts an Industrial Internet of Things (IIoT) architecture, integrating multiple transmission protocols through redundantly designed wireless communication gateways. For high-priority alarm signals (such as fracture events), low-latency 5G URLLC or dedicated LoRa spread spectrum links are used to ensure end-to-end transmission within 200ms; for massive displacement monitoring data, NB-IoT or 4G LTE Cat-M1 wide-area low-power networks are enabled for batch compression transmission.

[0225] During data transmission, data encryption and verification mechanisms can be incorporated to ensure data integrity and security. Data encryption effectively prevents data theft or tampering, protecting the privacy of monitoring data; while data verification mechanisms can detect errors or loss during transmission in real time and correct or retransmit them promptly, ensuring data accuracy and reliability. Furthermore, the data transmission layer also has a data caching function. When wireless signal interference causes transmission interruption, the data collected by the data acquisition layer can be temporarily stored in the cache and transmitted again after the signal is restored, thus preventing data loss.

[0226] The data acquisition layer includes several data acquisition devices, including a bolt fracture detection module and a bolt loosening detection module. The bolt fracture detection module determines whether a bolt is fractured by detecting the signal in the signal circuit, while the bolt loosening detection module determines whether a bolt is loose by using a laser displacement sensor and a lidar.

[0227] The data acquisition layer is the core component of the entire detection system, and it is directly responsible for collecting various status parameters of the bolts inside the wind turbine nacelle.

[0228] The bolt fracture detection module is a key device used to monitor whether a bolt has broken. It determines the bolt's condition by detecting changes in the signal circuit. Specifically, when the bolt is in a normal connection state, the signal circuit is intact, and the signal can be transmitted normally; however, once the bolt breaks, the signal circuit is severed, and signal transmission is interrupted. The bolt fracture detection module can monitor the status of the signal circuit in real time and send the detected signal change data to the data transmission layer.

[0229] The bolt loosening detection module is used to monitor whether bolts are loose. It employs laser displacement sensors and lidar technology. The laser displacement sensor can accurately measure the relative displacement change between the bolt and the connected component. When a bolt loosens, the distance between it and the connected component changes slightly; the laser displacement sensor can sensitively detect this change and convert it into an electrical signal output.

[0230] LiDAR, on the other hand, emits a laser beam and receives the reflected signal to accurately measure the three-dimensional position of the bolt. By continuously monitoring the bolt's position, LiDAR can determine whether the bolt has shifted or loosened. The bolt loosening detection module integrates and analyzes the data from the laser displacement sensor and LiDAR to accurately determine the degree of bolt loosening. This detection method based on multi-sensor fusion effectively improves the accuracy and reliability of bolt loosening detection, avoiding false positives and false negatives.

[0231] The data sensing layer includes several groups of detection units, the same number as the data acquisition equipment. Each group of detection units includes several detection units, and the detection units in the same group can be connected in series to form a signal loop.

[0232] The data sensing layer is the underlying foundation of the detection system. Closely connected to the data acquisition layer, it provides the necessary physical support and signal transmission channels. The data sensing layer comprises several groups of detection units, the same number as the data acquisition devices. Each group consists of multiple detection units with different functions. Several detection units within the same group are connected in series to form a signal loop. The detection units utilize robust and durable housing materials, capable of withstanding extreme conditions such as high wind speeds, low temperatures, and high humidity, ensuring stable and reliable performance during long-term operation.

[0233] This wind turbine nacelle bolt condition monitoring system achieves comprehensive real-time monitoring of the bolt condition within the wind turbine nacelle through the collaborative work of its data application layer, data transmission layer, data acquisition layer, and data perception layer. It not only promptly detects bolt breakage and loosening but also provides users with accurate early warning information and detailed data analysis reports, helping them to take proactive maintenance measures and avoid equipment damage and safety accidents caused by bolt failures. Furthermore, the system's intelligent design and advanced technology applications make it highly efficient, reliable, and easy to use, meeting the high requirements of modern wind power equipment operation and maintenance management, and providing strong technical support for the safe and stable development of the wind power industry.

[0234] <Wind turbine nacelle bolt condition testing equipment>

[0235] The testing equipment includes a cloud server, a site terminal, an AP gateway device, several testing hosts, and several testing switches;

[0236] The cloud server includes a data center module, an intelligent analysis and decision-making module, and a device management and collaboration module.

[0237] The data center module stores massive amounts of monitoring data uploaded from the monitoring host, including critical information such as bolt displacement, loosening status, and fracture conditions. Through long-term accumulation and management of this data, the data center module provides abundant resources for subsequent data mining and equipment health management. The intelligent analysis and decision-making module utilizes algorithms and models to perform in-depth analysis of the monitoring data, determining in real-time whether the bolts are in normal condition. Upon detecting potential fault risks, this module immediately generates early warning information and provides corresponding handling suggestions to maintenance personnel based on preset decision rules. The equipment management and collaboration module is responsible for the unified management and coordination of all equipment in the entire monitoring system, ensuring efficient and stable collaborative operation among components.

[0238] The site terminal serves as the user interface for the monitoring system, comprising a human-machine interface (HMI) module and a communication module. The HMI module features an intuitive and user-friendly interface, allowing users to view real-time monitoring data of bolts within the wind turbine nacelle, including real-time bolt status, historical data trends, and early warning information. Furthermore, users can use this module to set parameters and control the monitoring system. The communication module establishes a stable communication connection between the site terminal, the cloud server, and the monitoring host, ensuring real-time and accurate data transmission.

[0239] The site terminal also includes an identity authentication and login module and an access control module. The identity authentication and login module verifies the user's identity information to ensure that only authorized personnel can access the system. When logging in, users need to enter a pre-registered account and password, which the system rigorously compares. If the entered account and password match the authorization information stored in the system, the user is allowed access; if the information does not match, the system rejects the login request and records the unauthorized login attempt for subsequent security audits. To further enhance system security, the identity authentication and login module can also support multi-factor authentication mechanisms. In addition to traditional account and password authentication, users can choose to use auxiliary authentication methods such as mobile SMS verification codes, fingerprint recognition, facial recognition, or security tokens. These multi-factor authentication methods effectively prevent unauthorized access due to leaked account passwords, ensuring system security.

[0240] The access control module is a core component of the system's security architecture. It's responsible for assigning different operational permissions based on user identity and role. In the wind turbine nacelle bolt condition monitoring equipment, different users may have different responsibilities and operational needs. For example, ordinary maintenance personnel may only need to view bolt monitoring data and receive alerts, while system administrators need advanced permissions to configure, maintain, and manage the equipment. The access control module ensures that users can only operate within their authorized scope by assigning specific roles to each user and setting corresponding permissions based on those roles.

[0241] The collaborative operation of the identity authentication and login module and the access control module provides strong protection for the safe operation of the wind turbine nacelle bolt condition monitoring equipment. When a user attempts to log in to the system, the identity authentication and login module first verifies the user's identity. If the user passes the authentication, the access control module will load the corresponding operation interface and functional modules for the user according to their permission settings. All user operations within the system are strictly monitored by the access control module; any operation exceeding the authorized scope will be rejected by the system and recorded in the security log. This strict security management mechanism not only protects the system's data security but also ensures the normal operation of the equipment and the orderly conduct of maintenance work.

[0242] By introducing identity authentication and login modules and access control modules, the field terminal of the wind turbine nacelle bolt condition monitoring equipment not only enhances system security but also provides users with a more flexible and personalized operating experience. Users can securely access various functions of the system according to their responsibilities and needs, while system administrators can effectively manage and monitor system usage through these modules, ensuring stable system operation and data security.

[0243] The AP gateway device serves as a communication bridge connecting the detection host and the cloud server. It includes a wireless Wi-Fi module, a 4G module, a LoRa module, and an RS-485 interface module. These communication modules provide the detection system with multiple communication methods to adapt to different application scenarios and environmental conditions. The wireless Wi-Fi module is suitable for high-speed data transmission over short distances, the 4G module enables stable communication over longer distances, and the LoRa module, with its low power consumption and long-distance transmission capabilities, is suitable for the communication needs of distributed detection hosts. The RS-485 interface module supports traditional wired communication, ensuring reliable data transmission even in situations with poor wireless signal strength.

[0244] The detection host is the core equipment for bolt condition monitoring. It includes a LoRa communication module, an ADC conversion module, a LiDAR module, a CPU module, a cache module, a point cloud computing module, a clock synchronization module, and a power supply module. To achieve comprehensive coverage of all bolts inside the wind turbine nacelle, at least three detection hosts are installed on the inner wall of the nacelle. Through reasonable layout and collaborative operation, these detection hosts ensure that every bolt is within the monitoring range. The detection host not only has traditional fracture detection functions but also LiDAR capabilities. The LiDAR emits a laser beam to scan the three-dimensional contour of the bolt and generates high-precision point cloud data. Through real-time analysis of this point cloud data, the detection host can construct a three-dimensional model of the bolt and use a bolt disturbance state detection algorithm based on LiDAR point cloud data to accurately detect the loosening state of the bolt. This detection algorithm based on LiDAR point cloud data can effectively identify minor loosening of bolts caused by vibration, fatigue, and other factors during operation, providing a more reliable guarantee for the safe operation of the wind turbine nacelle.

[0245] The detection switches are critical components installed on each bolt. These switches are connected in series to the detection host, forming a complete signal loop. When a bolt inside the wind turbine nacelle breaks, it snaps the detection signal wire, causing the signal loop to be interrupted. By monitoring the status of the signal loop in real time, the detection host can quickly detect the signal anomaly and determine that a bolt has broken. This signal loop-based breakage detection method is simple, reliable, and has a fast response time. It can issue an alarm instantly upon bolt breakage, providing valuable processing time for maintenance personnel.

[0246] like Figure 4 and 5 As shown, the detection switch includes a housing 1, a detection signal line 2, and springs 3. The housing 1 is a cylindrical structure, and several springs 3 are disposed inside the housing 1. One end of the spring 3 is fixedly connected to the inner wall of the housing 1, and the other end of the spring 3 is a free end. A bolt clamping cavity is formed between the several springs 3. The detection signal line 2 has a ring part 201 and two terminals 202. The ring part 201 is embedded in the housing 1, and the two terminals 202 are used to connect adjacent detection switches or detection host.

[0247] The detection signal line 2 is a uniquely designed signal line with a wire diameter of <0.24mm. It has flame-retardant and electromagnetic interference shielding functions. Its fiber core is made of silver. The connector 202 adopts the R11 plug interface standard and can disconnect from the plug under a torque of >4N·m.

[0248] Compared to conventional signal cables, it has the following advantages:

[0249] 1. When a bolt breaks, the connector comes off rather than the wire breaks;

[0250] 2. It can reduce electromagnetic interference;

[0251] 3. The strength is within a certain upper and lower threshold, and it breaks when the plug breaks;

[0252] 4. After the wire breaks, it will not cause leakage current due to the broken wire swinging and coming into contact with other electronic components or electrical equipment;

[0253] 5. Under normal cutting, swinging, and friction, there is sufficient insulation layer strength to protect against false alarms;

[0254] 6. Vibration-resistant design, ensuring secure fastening of the slots even under high-intensity vibration conditions.

[0255] A locking wing 4 is provided perpendicularly to the end face of the spring piece 3 facing the bolt clamping cavity. The locking wing 4 is perpendicular to the axis of the housing 1, and the outer edge of the locking wing 4 has a serrated structure. The housing 1 is made of PA66 plastic or PC plastic, and the spring piece 3 is made of spring steel with a thickness of 1mm.

[0256] When the bolt is pressed in, it needs to pass a radial force of about 5 Newtons and an axial force of about 50 Newtons. The spring 3 and locking wing 4 can be used to achieve a tight connection between the detection switch and the bolt to be detected.

[0257] The operation process of the wind turbine nacelle bolt condition detection equipment is as follows: When the wind turbine generator is operating normally, the detection host continuously monitors the bolt condition through laser displacement sensors and lidar. The laser displacement sensor collects the relative displacement data between the bolt head and the flange surface in real time and transmits it to the data processing and control module. The data processing and control module analyzes the displacement data to determine whether there is any abnormal displacement of the bolt. At the same time, the lidar scans the bolt surface to generate point cloud data and transmits it to the data processing and control module. The data processing and control module uses a bolt disturbance state detection algorithm to analyze the point cloud data to determine whether there is any loosening of the bolt. Once an abnormal displacement or loosening of the bolt is detected, the data processing and control module will immediately generate an alarm signal and upload the alarm information and related monitoring data to the cloud server through the AP gateway device. After receiving the alarm information, the cloud server's intelligent analysis and decision-making module will further analyze the alarm, determine the severity of the alarm, and generate corresponding handling suggestions according to preset decision rules. The equipment management and collaboration module will then display the alarm information and handling suggestions to the maintenance personnel through the human-machine interaction module of the site terminal, reminding them to take appropriate maintenance measures in a timely manner.

[0258] Furthermore, the detection switch plays a crucial role in bolt breakage detection. When a bolt breaks, the signal detection wire of the switch is severed, causing an interruption in the signal circuit. The detection host, by monitoring the status of the signal circuit in real time, can quickly detect the signal anomaly and determine that a bolt has broken. The detection host will immediately generate a breakage alarm signal and upload the alarm information to the cloud server. Upon receiving the breakage alarm, the cloud server will immediately notify the maintenance personnel, reminding them to take emergency measures to handle the broken bolt to prevent equipment damage and safety accidents caused by bolt breakage.

[0259] In terms of maintenance and management, wind turbine nacelle bolt condition monitoring equipment also offers significant advantages. Through the equipment management and collaboration module on a cloud server, maintenance personnel can remotely manage and maintain various devices within the monitoring system. They can view the equipment's operating status in real time through the human-machine interface module on the site terminal, including information such as the operating temperature of the monitoring host, power supply voltage, and communication status. When equipment malfunctions, maintenance personnel can quickly locate the cause of the fault and take appropriate repair measures through remote diagnostic functions. This remote management and maintenance method greatly reduces maintenance costs and workload, while improving maintenance efficiency.

[0260] The specific embodiments of the present invention have been described above. It should be understood that the present invention is not limited to the specific embodiments described above, and those skilled in the art can make various modifications or variations within the scope of the claims, which do not affect the essence of the present invention.

Claims

1. A method for detecting the condition of bolts in a wind turbine nacelle, characterized in that: This includes detection of broken and loose bolts in wind turbine nacelles; Fracture detection: A signal loop is constructed using an anti-electromagnetic interference signal line. The bolts to be tested inside the wind turbine nacelle are physically connected to the anti-electromagnetic interference signal line. The signal from the detection signal loop is used to determine whether a bolt is broken. When a bolt is broken, causing the anti-electromagnetic interference signal line to disconnect, the signal changes accordingly, indicating that a bolt is broken. The system acquires environmental noise signals in real time using a high-speed ADC, performs spectrum analysis using window function weighting and FFT module, dynamically identifies the concentrated frequency bands of motor harmonics, and implements automatic spectrum avoidance when a specific motor characteristic frequency is detected to eliminate the interference of electrostatic interference from fan blades and broadband noise from variable frequency motors on the detection results. The phase of the output signal is kept synchronized with the phase of the input reference signal by using a phase-locked loop (PLL). A Kalman frequency predictor is built into the PLL to store the most recent 32 frequency points, construct a state prediction vector, and eliminate noise through Kalman filtering to quickly lock the PLL. Looseness detection: Bolts are modeled in 3D using LiDAR point clouds, and the modeling of bolts by point clouds is analyzed in real time. Bolt loosening is detected using a bolt disturbance state detection algorithm based on LiDAR point cloud data. For the same bolt, the distance measurement point cloud set is clustered into 2-3 point cloud sets. The average of the distance measurement data of all point clouds in the set is used to obtain the effective distance measurement data of the current set. When the effective distance measurement data or distance data of the 2-3 point cloud sets are equal or the error value is extremely small, the average of the point cloud cluster sets is calculated separately. The change of the effective distance measurement data over time is the displacement data. The bolt loosening is determined by this displacement data. The frequency sweep nonlinearity of the DFB laser is optimized by adopting a current-temperature co-modulation method. Combined with the real-time acquisition of grating feedback signal by ADC, the frequency-time curve is dynamically corrected. Combined with real-time phase demodulation algorithm and multi-dimensional error compensation mechanism, the frequency sweep nonlinearity error is suppressed. Cross-correlation analysis was performed on the output signals of the reference interferometer and the measuring interferometer. A "vernier effect" model was constructed by extracting the phase difference of the dual-channel interference signals. Combined with CZT spectrum refinement technology, the phase resolution was improved.

2. The method for detecting the condition of wind turbine nacelle bolts as described in claim 1, characterized in that: The specific methods for automatic spectrum avoidance are as follows: S1. Add a Hanning window to the acquired signal. The specific formula is as follows: Where 0 ≤ n ≤ N-1; Fast algorithms for calculating the Discrete Fourier Transform of a signal: The sampling point is 1024; S2. Define μ as the average energy of background noise and σ as the standard deviation of background noise. Detect energy peaks at integer multiples of the fundamental frequency f0, f0·k, k=1,2,3,... Define an adaptive threshold as Threshold=μ+3σ. If three consecutive harmonic peaks > Threshold, it is determined to be a motor interference frequency band. S3. Implement a spectrum avoidance strategy, specifically the frequency jump to f. new =f current +Δf(,Δf(>interference bandwidth,dynamically adjust the center frequency of the bandpass filter to f new .

3. The method for detecting the condition of wind turbine nacelle bolts as described in claim 1, characterized in that: The specific method for noise elimination using Kalman filtering is as follows: Establish a spatial soft-state model of the Kalman filter, and recursively calculate the Kalman gain K using Kalman filtering. k And calculate the voltage phase-locked loop control (VCO) using the following formula: V ctrl (k)=K p ·θ e (k)+K i ∑θ e (k), θ e (k) represents the phase error output by the phase detector; The phase detector output is monitored in real time. When a phase change caused by an electrostatic pulse is detected, a compensation current is automatically injected to cancel the disturbance and maintain the continuity of the phase-locked state. Specifically: Mutation detection is achieved using difference, when Δθ=|θ e (t)-θ e When (t-1)|>30°, the compensation current is used for calculation; I comp =K c ·sgn(Δθ)·e -α∣Δθ∣ ; Where α is the attenuation coefficient, and α is taken as 0.05; K c To compensate for the gain; The corrected VCO input is: V ctrl′ (t)=V ctrl (t)+β·I comp , where β is the current-to-voltage conversion coefficient.

4. The method for detecting the condition of wind turbine nacelle bolts as described in claim 1, characterized in that: The algorithm for detecting bolt disturbance status based on lidar point cloud data is as follows: S1. Locating the core point: Collect the point cloud data. In the k-th frame, the point cloud set for the bolt region is P. k ={p k,1 ,p k,2 ,…,p k,m }, where p k,i =(x k,i ,y k,i ,z k,i ) represents the three-dimensional coordinates of the point, and m represents the total number of point clouds in this frame; for Point p∈P k Its ε-neighborhood is defined as: N ε (p)={q∈P k |dist(p,q)≤ε}; Where dist(p,q) is the three-dimensional Euclidean distance. ε represents the neighborhood radius, which is a preset parameter; If the number of points contained in the ε-neighborhood of point P satisfies |Nε(p)|≥MinPts, then p is a core point; Wherein, MinPts represents the minimum number of points threshold, which is a preset parameter; S2. Noise Determination and Point Cloud Filtering: If point P does not belong to any core point and is not within the ε-neighborhood of any core point, then p is a noise point, denoted as p∈Noise; after DBSCAN denoising in the k-th frame, the effective point cloud set of the bolt surface is: Among them, Noise k Let be the set of noise points in the k-th frame; S3. Bolt surface point density and growth rate, and anomaly detection: In the k-th frame, the effective point cloud quantity of the i-th bolt is: Let i be a subset of the point cloud of the i-th bolt; The point cloud increment is defined as follows: the increase in point cloud size of the i-th bolt in frame k relative to frame (k-1) is Δn. i,k =n i,k -n i,k-1 Calculate the cumulative average growth of the i-th bolt from the initial frame to the (k-1)-th frame. Define the ratio of the growth of the i-th bolt in the K-th frame to the cumulative average growth. If r i,j ≤α or r i,j If ≥β, then the i-th bolt is determined to be in t. k Displacement anomalies exist at all times; Where α is the lower threshold and β is the upper threshold, and α and β are set according to historical data of the bolt under stable conditions.

5. The method for detecting the condition of wind turbine nacelle bolts as described in claim 1, characterized in that: The method for suppressing sweep frequency nonlinearity error specifically includes the following steps: S1, System Initialization Set the initial current I0 and temperature T0, and load the calibration parameters; S2, Signal Acquisition and Processing The ADC acquires the grating feedback signal in real time with a sampling rate ≥100MSa / s; S3, Real-time Phase Demodulation Window function optimization and real-time FFT are used to accelerate computation: The instantaneous phase is calculated as follows: S4. Frequency Error Calculation e f (t)=f ideal (t)-f actual (t) S5, Multi-dimensional Error Compensation The error model is defined as follows: Calculate the combined compensation term for temperature drift and current fluctuation using the following formula: I comp (t)=I0(t)+K I ·E(t) T comp (t)=T0(t)+K T ·E(t) S6, Current-Temperature Co-modulation Define the cooperative control equations, calculate the modulation of current and temperature, and calculate ΔI and ΔT based on the control matrix: Among them, e f (t) represents the frequency error, e φ (t) represents the phase error; S7, Dynamically Update Parameters The algorithm parameters are adaptively adjusted according to environmental changes; S8, Frequency-Time Curve Correction Calculate the corrected actual frequency: f actual (t)=f ideal (t)+Δf comp (t) The compensation item is calculated as follows: Legendre orthogonal polynomials are used instead of traditional polynomial fitting to avoid matrix ill-conditioning and improve fitting stability. Update the output frequency curve and return to step S2.

6. The method for detecting the condition of wind turbine nacelle bolts as described in claim 1, characterized in that: In methods to improve phase resolution, phase-locked loops or Kalman filtering techniques are used to compensate for the phase noise of the local oscillator and the signal source. That is, compensation is performed for each frame of signal, and the compensation accuracy is found to be optimal through iterative iterations. At the same time, a window function is added, and after acquiring a complete cycle of signal, Fourier transform is performed on the cycle signal to eliminate interference frequency bands.

7. A wind turbine nacelle bolt condition detection system, characterized in that: It includes the data application layer, data transmission layer, data acquisition layer, and data perception layer; The data application layer includes cloud servers and field terminals. The field terminals are responsible for providing users with an entry point for real-time monitoring data access. Users can log in by entering their account and password, and then remotely monitor the real-time status of bolts inside the wind turbine nacelle through a human-computer interaction display interface. The data transmission layer is responsible for transmitting various parameter data collected by the data acquisition layer to the data application layer via wireless signals; The data acquisition layer includes several data acquisition devices, including a bolt breakage detection module and a bolt loosening detection module. The bolt breakage detection module determines whether a bolt is broken by detecting the signal in the signal circuit, and the bolt loosening detection module determines whether a bolt is loose by using a laser displacement sensor and a lidar. The data sensing layer includes several groups of detection units, the same number as the data acquisition equipment. Each group of detection units includes several detection units, and the detection units in the same group can be connected in series to form a signal loop.

8. A wind turbine nacelle bolt condition inspection device, characterized in that: Includes cloud servers, site terminals, AP gateway devices, several detection hosts, and several detection switches; The cloud server includes a data center module, an intelligent analysis and decision-making module, and a device management and collaboration module; The station terminal includes a human-computer interaction module and a communication module; The AP gateway device includes a wireless WIFI module, a 4G module, a LoRa module, and an RS-485 interface module; The detection host includes a LoRa communication module, an ADC conversion module, a LiDAR module, a CPU module, a cache module, a point cloud computing module, a clock synchronization module, and a power supply module. The detection switch includes a housing, a detection signal line, and springs. The housing is a cylindrical structure, and several springs are arranged inside the housing. One end of the spring is fixedly connected to the inner wall of the housing, and the other end of the spring is a free end. A bolt clamping cavity is formed between the several springs. The detection signal line has a loop part and two terminals. The loop part is embedded in the housing, and the two terminals are used to connect to adjacent detection switches or detection host.

9. The wind turbine nacelle bolt condition detection device as described in claim 8, characterized in that: The end face of the spring sheet facing the bolt clamping cavity is provided with a locking wing, which is perpendicular to the axis of the housing and has a serrated outer edge.

10. The wind turbine nacelle bolt condition detection device as described in claim 9, characterized in that: The shell is made of PA66 plastic or PC plastic, and the spring is made of spring steel with a thickness of 1mm.

Images