A vibration monitoring method and system based on a microwave probe

Abstract

This invention provides a vibration monitoring method and system based on microwave probes, belonging to the field of vibration monitoring technology. The method includes: arranging a microwave probe array circumferentially along the root of a wind turbine blade; each microwave probe includes a microwave transmitter and a receiver, arranged in a ring at fixed intervals θ; when the blade is stationary, controlling the microwave probe array to emit linear frequency modulated microwave signals onto the surface of a fiberglass composite material, receiving the reflected echoes, and recording the initial phase of each probe in the corresponding state. This invention achieves high-precision real-time monitoring and intelligent early warning of multi-directional vibration of wind turbine blades in complex environments by fusing phase and time difference dual-mode measurements using a microwave probe array and a sparse reconstruction algorithm, thereby improving fault identification efficiency and equipment safety.

Description

Technical Field

[0001] This invention relates to the field of vibration monitoring technology, and in particular to a vibration monitoring method and system based on a microwave probe. Background Technology

[0002] Wind turbine blades are subject to various factors such as aerodynamic loads and gravity changes during operation, which may cause vibrations of varying degrees. Continuous abnormal vibrations pose a risk of structural fatigue, so effective monitoring of blade vibration is of practical significance. Some existing vibration monitoring methods still have room for improvement in terms of stability and adaptability when dealing with complex natural environments. For example, in a wind farm in the East China coast, moisture condensation is prone to occur on the blade surface during certain seasons, which may interfere with measurement methods that rely on optical reflection principles, affecting signal quality and monitoring continuity.

[0003] In addition, some monitoring methods based on timed sampling may not be ideal in capturing vibration characteristics when faced with low-frequency vibrations due to limitations in sampling methods and signal processing mechanisms, making it difficult to fully reflect the dynamic response characteristics of blades under varying speed and load conditions. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to provide a vibration monitoring method and system based on a microwave probe, which effectively avoids blade breakage accidents caused by vibration instability.

[0005] To solve the above-mentioned technical problems, the technical solution of the present invention is as follows:

[0006] In a first aspect, a vibration monitoring method based on a microwave probe, the method comprising:

[0007] A microwave detection array is arranged circumferentially along the root of the wind turbine blade. Each microwave detection pin contains a microwave transmitter and a receiver, and they are arranged in a ring at fixed intervals θ.

[0008] When the blade is stationary, the microwave detection array is controlled to emit linear frequency modulated microwave signals to the surface of the fiberglass composite material and receive the reflected echoes, and the initial phase of each probe is recorded in the corresponding state.

[0009] Based on the dielectric constant of the blade material, the propagation speed of electromagnetic waves is calculated. Combined with the geometric parameters of the probe installation position, a vibration-free reference model including the reference arrival time and initial phase is established.

[0010] During blade operation, microwave echo signals received by each probe are acquired in real time, and the phase change and arrival time offset relative to the reference signal model are extracted to obtain the undersampled signal.

[0011] The undersampled signal is sparsely reconstructed, and the vibration displacement equation is solved using the reconstruction results to obtain the vibration displacement information of the blade.

[0012] Based on the vibration displacement information of the blades, a two-way vibration model is constructed, and the circumferential vibration components and axial vibration components are analyzed. When the vibration amount in either direction exceeds the preset threshold, a corresponding multi-level alarm mechanism is triggered.

[0013] Secondly, a vibration monitoring system based on a microwave probe includes:

[0014] A microwave detection array module is used to arrange a microwave detection array around the root of a wind turbine blade. Each microwave detection pin contains a microwave transmitter and a receiver, and the microwave detection pins are arranged in a ring at a fixed interval θ.

[0015] The signal processing module is used to control the microwave detection array to emit linear frequency modulated microwave signals to the surface of the fiberglass composite material when the blade is stationary, and to receive the reflected echoes and record the initial phase of each probe in the corresponding state; when the blade is running, it collects the microwave echo signals received by each probe in real time.

[0016] The reference model building module is used to calculate the electromagnetic wave propagation speed based on the dielectric constant of the blade material, and combine it with the geometric parameters of the probe installation position to build a vibration-free reference model that includes the reference arrival time and initial phase.

[0017] The feature extraction module is used to compare the microwave echo signals of each probe with the reference signal model under vibration-free conditions, construct an analytical signal, calculate the phase difference and arrival time offset, and perform undersampling processing to obtain the undersampled signal.

[0018] The signal reconstruction module is used to perform sparse reconstruction of the undersampled signal and solve the vibration displacement equation based on the reconstruction results to obtain the vibration displacement information of the blade.

[0019] The alarm triggering module is used to construct a two-way vibration model based on the vibration displacement information of the blade, and analyze the circumferential vibration component and the axial vibration component. When the vibration amount in either direction exceeds the preset threshold, the corresponding multi-level alarm mechanism is triggered.

[0020] Thirdly, a computing device includes:

[0021] One or more processors;

[0022] A storage device for storing one or more programs that, when executed by one or more processors, cause the one or more processors to implement the method.

[0023] Fourthly, a computer-readable storage medium storing a program that, when executed by a processor, implements the method.

[0024] The above-described solution of the present invention has at least the following beneficial effects:

[0025] Based on the principle of microwave detection, signal propagation and reception are unaffected by environmental factors such as moisture and dust on the blade surface, and do not rely on optical reflection. They can stably penetrate or bypass surface interference media, ensuring signal quality and monitoring continuity in harsh weather conditions, such as high humidity and fog, thus improving adaptability to complex natural environments. A vibration-free reference model containing the reference arrival time and the initial phase after calibration is first established. Then, phase changes and arrival time offsets are extracted by analyzing the signal, allowing for the capture of minute vibration differences. After undersampling the signal, the original signal characteristics are restored through a sparse reconstruction algorithm, overcoming the limitations of timed sampling. This allows for the capture of low-frequency vibrations and a comprehensive reflection of the blade's dynamic response under varying speed and load conditions, avoiding the omission of vibration characteristics. A bidirectional vibration model is constructed. The model decomposes vibration into circumferential and axial components, providing a more comprehensive understanding of blade vibration compared to single-direction monitoring. This avoids structural hazards caused by undetected vibration anomalies in a particular direction. Multiple preset thresholds are set, triggering corresponding alarms when vibration exceeds these thresholds. This allows staff to quickly assess the risk level based on alarm severity, such as minor anomalies, moderate warnings, or emergency shutdowns, enabling timely and targeted interventions. This effectively prevents structural fatigue caused by continuous abnormal vibration, extends blade lifespan, and reduces maintenance costs. The probe integrates a transmitter and receiver, resulting in a compact structure. Furthermore, the microwave signal requires no frequent calibration, reducing on-site maintenance workload and costs compared to optical monitoring equipment. This better meets the practical needs of long-term, stable monitoring in wind power scenarios. Attached Figure Description

[0026] Figure 1 This is a schematic flowchart of a vibration monitoring method based on a microwave probe provided by an embodiment of the present invention.

[0027] Figure 2 This is a schematic diagram of a vibration monitoring system based on a microwave probe provided in an embodiment of the present invention. Detailed Implementation

[0028] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

[0029] like Figure 1As shown, an embodiment of the present invention proposes a vibration monitoring method based on a microwave probe, the method comprising the following steps:

[0030] Step 1: Arrange a microwave detection array around the root of the wind turbine blade. Each microwave detection pin contains a microwave transmitter and a receiver, and they are arranged in a ring at fixed intervals θ.

[0031] Step 2: With the blade stationary, control the microwave detection array to emit linear frequency modulated microwave signals to the surface of the fiberglass composite material, receive the reflected echoes, and record the initial phase of each probe in the corresponding state.

[0032] Step 3: Calculate the electromagnetic wave propagation speed based on the dielectric constant of the blade material, and establish a vibration-free reference model including the reference arrival time and initial phase by combining the geometric parameters of the probe installation position.

[0033] Step 4: Under the condition of blade operation, the microwave echo signals received by each probe are collected in real time, and the phase change and arrival time offset relative to the reference signal model are extracted to obtain the undersampled signal.

[0034] Step 5: Perform sparse reconstruction on the undersampled signal, and solve the vibration displacement equation using the reconstruction results to obtain the vibration displacement information of the blade.

[0035] Step 6: Based on the vibration displacement information of the blade, construct a two-way vibration model and analyze the circumferential vibration component and the axial vibration component. When the vibration amount in either direction exceeds the preset threshold, trigger the corresponding multi-level alarm mechanism.

[0036] In this embodiment of the invention, based on the principle of microwave detection, signal propagation and reception are unaffected by environmental factors such as moisture and dust on the blade surface, do not rely on optical reflection, and can stably penetrate or bypass surface interference media, ensuring signal quality and monitoring continuity in harsh weather conditions, such as high humidity and fog, and improving adaptability to complex natural environments. First, a vibration-free reference model containing the reference arrival time and the initial phase after calibration is established. Then, phase changes and arrival time offsets are extracted by analyzing the signal comparison, which can capture minute vibration differences. After undersampling the signal, the original signal characteristics are restored through a sparse reconstruction algorithm, breaking through the limitations of timed sampling. This not only captures low-frequency vibrations but also comprehensively reflects the dynamic response of the blade under varying speed and load conditions, avoiding the omission of vibration characteristics. The bidirectional vibration model decomposes vibration into circumferential and axial components. Compared with single-direction monitoring, it can provide a more comprehensive understanding of the blade vibration status and avoid structural hazards caused by undetected vibration anomalies in one direction. Multiple preset thresholds are set, triggering corresponding alarms when vibration exceeds the threshold. This allows staff to quickly assess the risk level based on the alarm level, such as minor anomalies, moderate warnings, or emergency shutdowns, enabling timely and targeted measures to effectively prevent structural fatigue caused by continuous abnormal vibration, extend blade life, and reduce maintenance costs. The probe integrates a transmitter and receiver, featuring a compact structure. Furthermore, the microwave signal does not require frequent calibration, reducing on-site maintenance workload and costs compared to optical monitoring equipment, and better meeting the practical needs of long-term, stable monitoring in wind power scenarios.

[0037] In a preferred embodiment of the present invention, step 1, arranging a microwave detection array circumferentially along the root of the wind turbine blade, wherein each microwave detection pin includes a microwave transmitter and a receiver, and is arranged in a ring at fixed intervals θ; step 2, controlling the microwave detection array to emit linear frequency modulated microwave signals onto the surface of the fiberglass composite material when the blade is stationary, receiving the reflected echoes, and recording the initial phase of each detection pin in the corresponding state, may include:

[0038] In this embodiment of the invention, when planning the circumferential layout at the root of a wind turbine blade, it is first necessary to measure the circumference of the blade root and calculate the required number of microwave probes based on a preset fixed interval. The number of probes is then calculated as follows: (Round the result up) During the actual installation process, installation clamps are used to ensure that each microwave probe is accurately installed in the predetermined position. The installation angle of the microwave probe must be strictly controlled to ensure that the microwave transmitter and receiver can effectively transmit and receive microwave signals. All probes must maintain the same installation angle to ensure measurement accuracy and consistency. Simultaneously, proper fixing and protection measures must be taken for the probes to prevent loosening or detachment during blade operation. The electrical connection of the probes must also be ensured to be stable and reliable, and to correctly connect them to subsequent signal processing equipment. For example, taking a certain type of wind turbine blade as an example, its root circumference is measured to be 10 meters. With a fixed interval of 0.5 meters, calculations show that 20 microwave probes are needed. During installation, technicians use customized ring clamps to evenly fix the microwave probes on the circumference of the blade root. The distance between each probe is precisely controlled at 0.5 meters, and the installation angle is perpendicular to the blade surface. After connecting the electrical wiring, preliminary signal testing is conducted to ensure that each probe functions normally.

[0039] Step 2: When the blade is stationary, the control device of the microwave detection array is activated. This control device generates control commands for a linear frequency modulated (LFM) microwave signal according to preset parameters and sends them to the transmitter of each microwave probe. The frequency variation range and modulation slope of the LFM microwave signal are preset according to the characteristics of the blade and the measurement requirements. After receiving the command, the microwave transmitter begins to emit the LFM microwave signal onto the surface of the fiberglass composite material. When the microwave signal encounters the blade surface, it is reflected. The reflected echo is received by the receiver of the microwave probe. The receiver amplifies and filters the received weak echo signal to remove noise and interference signals. The preprocessed echo signal is sent to the phase measurement circuit. At the same time, the reference waveform of the transmitted signal is also synchronously input into the circuit. The phase measurement circuit captures the phase difference between the echo signal and the transmitted signal at the same point in time by comparing the waveform changes of the two signals. Specifically, the circuit... The transmitted signal is used as a reference, and its phase value at each moment is recorded. The phase value of the echo signal at the same moment is also recorded. The difference between the two is the initial phase of the probe when the blade is stationary. Finally, the initial phase data obtained from these measurements are sent to the data storage device through the data transmission line for storage. Taking the wind turbine blade mentioned above as an example, the control device of the microwave detection array is set to transmit a linearly frequency-modulated microwave signal from 10GHz to 10.5GHz with a modulation slope of 100MHz / μs. When the blade is stationary, the control device sends a transmission command to 20 microwave probes. After the transmitter of each probe transmits the signal, the receiver successively receives the reflected echo. For example, after the fifth probe receives the echo, after amplification and filtering, its initial phase is measured to be 30° by the phase measurement circuit. The initial phase data of all 20 probes are recorded and stored in the database of the monitoring system.

[0040] In a preferred embodiment of the present invention, step 3 above, which involves calculating the electromagnetic wave propagation speed based on the dielectric constant of the blade material and establishing a vibration-free reference model including the reference arrival time and initial phase, in conjunction with the geometric parameters of the probe installation position, may include:

[0041] Step 301: By analyzing the time difference between the transmitted signal and the reflected echo signal, and combining this with the propagation speed of the microwave signal in the medium, the time elapsed from transmission to reception is calculated, i.e., the reference arrival time of each probe. Specifically, in the signal processing system, the time measurement unit records the timestamps of the transmitted signal and the received reflected echo signal. When the microwave probe transmits a signal, the time measurement unit starts timing; once the reflected echo signal is received, timing stops immediately, and the time difference between the two is obtained. Since the propagation speed of the microwave signal in the medium is closely related to the dielectric constant of the blade material, according to the formula... = (in It is the speed of light in a vacuum. The propagation speed of the microwave signal in the blade material is calculated using the dielectric constant of the blade material. Given the distance the microwave signal travels (which can be determined based on the distance from the probe to the blade surface and the signal propagation path), and combined with the previously obtained time difference, the time elapsed from transmission to reception can be calculated, i.e., the reference arrival time of each probe.

[0042] Step 302: Perform multiple measurements on each probe, record the time difference of each measurement, and calculate the average of all time differences. Use the average of all time differences as the final reference arrival time. Specifically, to improve the accuracy and reliability of the reference arrival time, perform multiple measurements on each probe. Under the same vibration-free conditions, repeat the operation of transmitting and receiving microwave signals. Record the time difference between the transmitted signal and the reflected echo signal each time. Record the time difference obtained from each measurement in a special data table. Then, add up all the time differences obtained from the measurements and divide by the number of measurements to obtain the average of all time differences. This average can effectively reduce the measurement error and is finally used as the final reference arrival time of the probe.

[0043] Step 303: For the initial phase recorded by each probe, the phase difference is calculated by comparing it with a reference signal of one phase, and the initial phase is compensated to obtain the calibrated and compensated initial phase. Specifically, this includes: after completing the reference arrival time calculation, the calibration and compensation of the initial phase begins. First, a stable phase reference signal needs to be set. This reference signal is generated by a high-precision signal generator, and its frequency must be consistent with the center frequency of the linear frequency modulated microwave signal emitted by the microwave probe to ensure the effectiveness of the phase comparison. Taking a 10GHz to 10.5GHz linear frequency modulated signal as an example, the frequency of the phase reference signal is set to 10.25GHz, which is the frequency of the linear frequency modulated signal. The center frequency is set, and the initial phase of the reference signal is fixed at 0°. Throughout the calibration process, the frequency and phase are kept constant through the constant temperature control and anti-interference design of the signal generator to avoid reference signal deviation caused by external environmental fluctuations, ensuring its stability and accuracy. Then, the initial phase recorded by each probe in the stationary state of the blade is compared with this preset phase reference signal. The comparison process is implemented through a phase comparison circuit, which simultaneously receives two signals: one is the microwave echo signal corresponding to the initial phase recorded by the probe, and the other is the 10.25GHz, 0° phase reference signal output from the high-precision signal generator. Internally, a phase detector is used to implement the comparison. The circuit captures the phase state of two signals at the same point in time. For example, at a certain moment, the phase of the reference signal is 0°, while the phase of the probe echo signal is 30°. The circuit calculates the difference between the two as 30°, which is the phase difference between the initial phase of the probe and the phase of the reference signal. If the phase of a probe echo signal is -15°, that is, 15° ahead of the reference signal, the calculated phase difference is -15°. Based on the calculated phase difference, the initial phase is adjusted accordingly. If the phase difference is positive, such as the 30° mentioned above, it means that the initial phase recorded by the probe lags behind the reference signal phase. In this case, the initial phase of the probe needs to be added to the phase difference. That is, 30° (initial phase) + 30° (phase difference) = 60°, so that the compensated phase is synchronized with the reference signal phase in terms of propagation characteristics. If the phase difference is negative, such as -15°, it means that the initial phase of the probe is ahead of the reference signal phase. In this case, the absolute value of the phase difference needs to be subtracted from the initial phase, that is, -15° (initial phase) - 15° (absolute value of phase difference) = -30°. Phase synchronization is also achieved through such adjustment. Through this compensation process, the phase deviation caused by hardware manufacturing errors of different probes, such as the phase response difference between the transmitter and receiver, or slight interference in the installation environment, such as local temperature fluctuations, can be effectively eliminated, and finally the calibrated and compensated initial phase is obtained.

[0044] Step 304: Based on the reference arrival time and the calibrated and compensated initial phase, determine the reference signal model parameters under vibration-free conditions. Specifically, this includes using the reference arrival time and the calibrated and compensated initial phase obtained in the previous steps as core parameters, and combining them with other stable characteristic parameters of the microwave signal under vibration-free conditions, such as signal amplitude and frequency, to form a parameter set. When determining the parameter weights, it is necessary to base them on the physical principles of microwave detection and actual monitoring needs. That is, the reference arrival time directly reflects the time characteristics of signal propagation and is the key to judging the position change caused by blade vibration, so it is given the highest weight and set to 0.4. The calibrated initial phase has a significant impact on the accuracy of vibration displacement calculation, so its weight is second, set to 0.3. The signal amplitude is mainly used to judge whether the signal strength meets the requirements of subsequent processing, so its weight is set to 0.2. The signal frequency is basically constant under vibration-free conditions and has little impact on the overall accuracy of the model, so its weight is set to 0.1. The weight allocation needs to be verified through multiple experiments to ensure that the contribution of each parameter to the model accuracy matches the weight, and finally a complete reference signal model parameter set under vibration-free conditions is formed.

[0045] Step 305: Based on the reference signal model parameters under vibration-free conditions, including the reference arrival time and the calibrated initial phase, construct the reference signal model under vibration-free conditions. Specifically, taking the aforementioned probe as an example, its reference arrival time is... The initial phase after calibration is 60°. The signal amplitude under vibration-free conditions was measured to be 2V using signal acquisition equipment. The microwave signal frequency was set to 10GHz (corresponding to angular frequency). = According to the above weight allocation, these parameters and their corresponding weights are organized into a parameter set to provide complete parameter support for the construction of the reference signal model. Finally, the reference signal model under vibration-free conditions is constructed. Based on the determined parameter characteristics of the reference signal model, a suitable mathematical model is selected. Considering that microwave signals exhibit periodic changes during propagation, a sine wave model is chosen. ,in For signal amplitude, Angular frequency, For time, To accurately describe the phase variation, a sinusoidal wave model was chosen. The parameters from the parameter set were substituted into the model, with the signal amplitude A set to the measured value of 2V and the angular frequency... Calculated based on microwave signal frequency phase The initial phase after calibration is 60°, which needs to be converted to radians, i.e. The time t corresponding to the reference arrival time takes the value of Substituting these parameters into the sinusoidal wave model, we obtain the complete reference signal model under vibration-free conditions as follows: .

[0046] In this embodiment, an accurate reference arrival time helps to more precisely analyze the propagation characteristics of microwave signals in the blade, thereby improving the accuracy of blade vibration monitoring and enabling timely detection of minute vibration changes, providing strong support for early fault warning. Calibration and compensation of the initial phase eliminates phase deviations caused by equipment or environmental factors, making the initial phase more accurate and reliable. An accurate initial phase is crucial for constructing a reference signal model, helping to more accurately reflect the true characteristics of microwave signals in a vibration-free state. This allows for more sensitive detection of phase changes during blade vibration monitoring, improving the ability to identify vibration signals. Determining the reference signal model parameters in a vibration-free state comprehensively considers multiple key factors related to microwave signals, enabling the model to more comprehensively and accurately describe the characteristics of microwave signals in a vibration-free state. This helps to more accurately determine the vibration state of the blade, improving the accuracy and reliability of the monitoring system. During blade operation, comparing the real-time acquired signals with this reference model allows for rapid determination of whether the blade is vibrating, as well as the degree and type of vibration, improving the efficiency and accuracy of vibration monitoring. This helps to promptly detect abnormal blade vibrations and ensure the safe and stable operation of the equipment.

[0047] In a preferred embodiment of the present invention, step 4 above, which involves real-time acquisition of microwave echo signals received by each probe while the blade is in operation, and extraction of the phase change and arrival time offset relative to the reference signal model to obtain an undersampled signal, may include:

[0048] Step 401: Compare the microwave echo signals of each probe with the reference signal model under vibration-free conditions. For both the reference signal model and the real-time acquired microwave echo signals, construct analytical signals. Specifically, when processing the microwave echo signals received by each probe, preprocessing is required. In actual monitoring scenarios, the microwave echo signals received by the probes are affected by various interference factors, such as electromagnetic radiation generated by power transmission equipment in wind farms and thermal noise from the probe equipment's own circuitry. These interferences can cause high-frequency noise (frequency higher than the upper limit of the microwave signal operating frequency band) and unnecessary clutter in the signal, such as 50Hz power frequency interference from the power grid and low-frequency signals caused by airflow disturbances on the blade surface. These impurity signals can mask the characteristics of the effective echo, leading to deviations in subsequent phase calculations and vibration analysis. Therefore, filtering algorithms must be used to remove these interferences to improve signal quality. When selecting a filtering algorithm, it will be tailored to the specific type of interference. If the signal mainly contains interference higher than the microwave operating frequency band... For example, for high-frequency noise previously set between 10GHz and 10.5GHz, a low-pass filter is used. First, the received signal is analyzed using a spectrum analyzer to determine the lowest frequency of the high-frequency noise. Assuming that the high-frequency noise starts to appear at 15GHz, the cutoff frequency of the low-pass filter is set to 15GHz. This allows effective microwave signals below 15GHz to pass completely while filtering out high-frequency noise above 15GHz. If the signal contains noise of a specific frequency, such as 50Hz power frequency interference, a band-pass filter is used instead. The passband frequency range of the filter is strictly set to the operating frequency band of the microwave signal, 10GHz to 10.5GHz, allowing only effective signals within this band to pass while blocking power frequency interference and useless signals from other frequency bands. The filtering operation is implemented by a hardware filtering unit in the monitoring system. The unit processes the input echo signal in real time. After filtering, the signal-to-noise ratio can usually be improved to more than 25dB, effectively ensuring the accuracy of subsequent processing.

[0049] For the reference signal model constructed under vibration-free conditions, it is stored in the monitoring system database in the form of an explicit mathematical expression. The amplitude of the reference signal needs to be determined through actual measurements when the blade is stationary. During the construction of the vibration-free reference model, microwave probes are controlled to stably emit microwave signals towards the blade surface, while the amplitude data of the reflected echo signals received by each probe receiver are recorded. To ensure amplitude accuracy, the same probe is measured repeatedly (at least 20 times). Then, outliers caused by transient electromagnetic interference or equipment fluctuations are removed from these measurement data, such as values ​​significantly higher than 2.5V or significantly lower than 1.5V. The remaining valid data are then summed, and the sum is divided by the number of valid data points to obtain the final value. The average value is the amplitude of the reference signal. For example, if two outliers are removed from 20 measurements, the values ​​of the remaining 18 data points are between 1.98V and 2.02V. These values ​​are added together to get 36V, and then divided by 18 to get the amplitude of the reference signal as 2V. When processing the reference signal, the sampling frequency of the real-time acquired signal is used, such as a frequency of 21GHz with 2.1 billion data points per second. The system's built-in calculation program calls the mathematical expression of the reference signal, substituting each sampling moment, the determined amplitude, angular frequency, and phase into the expression to calculate the reference signal amplitude corresponding to each sampling moment. This generates a reference signal data sequence that is completely consistent with the sampling rhythm of the real-time signal, ensuring that the reference signal and the real-time signal can be directly compared in the time dimension.

[0050] For the real-time acquired microwave echo signal, after completing the above filtering preprocessing, an analytic signal needs to be constructed using Hilbert transform. First, the filtered real-time signal is discretized and sampled at the same sampling frequency (21 GHz) as the reference signal, that is, every approximately 4.76 × The actual amplitude of the signal is sampled once every second (calculated by dividing 1 by the 21GHz sampling frequency), resulting in a series of discrete real-part signal values. These values ​​are arranged strictly in the order of sampling, ultimately forming a sequence of real-part signals. The system then calls the Hilbert transform algorithm. The core of this algorithm for processing the real-part signal sequence is a pre-defined transform kernel function. This kernel function takes different forms depending on the time dimension of the signal (continuous or discrete). In the continuous time dimension, the Hilbert transform kernel function has a clear mathematical formula, specifically... Where h(t) represents the output value of the continuous kernel function, π is the mathematical constant pi, and t represents the continuous time variable, this formula describes the mathematical relationship between the kernel function and time in continuous time. However, in actual monitoring, discrete sampled signals are processed, and continuous kernel functions cannot be used directly. Therefore, the kernel function needs to be adjusted to adapt to discrete signals. At this time, the input variable of the kernel function becomes the index n of the discrete sampling point (n is an integer, representing the nth sampling point). The corresponding discrete kernel function formula has two cases: when n is odd... That is, the kernel function value is equal to 2 divided by (pi multiplied by the index n); when n is even, h(n) = 0, that is, the kernel function output is 0. This adjustment is to match the sampling characteristics of discrete signals and ensure accurate transformation calculation; then the algorithm will perform convolution operation to generate the imaginary part signal sequence. For each value in the real part signal sequence (assuming the real part value of the kth sampling point is...), Each time, the kernel function value h(k) corresponding to the sampling point index will be found in the discrete kernel function, and then... Multiply by h(k), then add the results of multiplying the real part values ​​of all sampling points in the real part signal sequence by the kernel function value. The sum obtained is the imaginary part signal value corresponding to each sampling point. Calculate the imaginary part value for each sampling point in the real part signal sequence in this way, and then arrange these imaginary part values ​​in the order of sampling to form the imaginary part signal sequence. Finally, take each value in the real part signal sequence as the real part of the complex signal, and take the corresponding value in the imaginary part signal sequence as the imaginary part of the complex signal. Combine them according to the rules of constructing the complex signal as real part + imaginary part (indicated by the imaginary unit j). For example, the analytic signal of the kth sampling point is... +j×the kth value of the imaginary part, thus forming a complete analytical signal. This analytical signal can completely preserve the key features of the real-time microwave echo signal, such as phase and amplitude.

[0051] Step 402: By calculating the arctangent of the ratio of the imaginary part to the real part in the analytic signal, the phases of the real-time acquired signal and the reference signal model are obtained. Based on the phases of the real-time acquired signal and the reference signal model, the phase difference is calculated. Specifically, after constructing the analytic signals of the real-time acquired signal and the reference signal model, the phase calculation and phase difference solution are performed. First, the analytic signal of the real-time acquired signal is processed. This type of analytic signal consists of both real and imaginary values, and each sampling point corresponds to a set of matching real and imaginary data. When calculating the phase of the real-time signal, each sampling point needs to be processed one by one. First, the imaginary and real values ​​in the analytic signal of that sampling point are extracted, and the imaginary value is divided by the real value. The ratio of the two values ​​is then calculated, and an arctangent operation is performed on this ratio. It's important to note that the result of the arctangent operation is within the range of -π to π. If a phase jump occurs after the operation, such as a sudden jump from a value close to π to a value close to -π, phase expansion processing is required. For example, when the phase difference between adjacent sampling points exceeds π, 2π is added to the phase of the next sampling point to adjust the discontinuous phase into a continuous phase curve, ensuring that the phase of each sampling point accurately reflects the actual phase state of the signal. This ultimately yields the phase data of the real-time acquired signal at each sampling point. Next, the phase of the reference signal model is calculated using the same method. The analytic signal of the reference signal model is generated based on a pre-constructed mathematical expression, which is in sine wave form. Specifically... ,in The amplitude of the reference signal. This is the amplitude of the reference signal. This value is obtained by averaging the amplitude of the reflected echo signal multiple times while the blade is stationary, for example, 2V. Angular frequency, where t is the sampling time. This is the calibrated initial phase (obtained through comparison and compensation with the phase reference signal); based on this mathematical expression, the real part sequence of the reference signal is first generated, i.e., the sequence corresponding to each sampling time t. The numerical values ​​are then transformed by Hilbert to obtain the corresponding imaginary part sequence, which ultimately constitutes the analytical signal of the reference signal model. The number of sampling points and the sampling time are completely consistent with the analytical signal of the real-time signal.

[0052] When calculating the phase of the reference signal model, for each sampling point, the imaginary and real parts of the reference analytic signal are extracted, their ratio is calculated, and then an arctangent operation is performed. If a phase jump occurs, the same phase expansion process as the real-time signal is used (adding 2π when the adjacent phase difference exceeds π) to eliminate phase discontinuities, obtaining the phase data of the reference signal model at each corresponding sampling point. The reason for using completely consistent calculation methods for the real-time and reference signals is to avoid additional phase deviations caused by differences in calculation logic, such as different phase expansion rules or different arctangent operation precision settings. This ensures that the two sets of phase data are generated under the same standard and have the conditions for direct comparison. Finally, the phase difference is calculated. Since the phase data of the real-time signal and the reference signal correspond one-to-one at the sampling points... Each sampling moment has a set of real-time phase values ​​and a set of reference phase values. Therefore, the two sets of phase data at each same sampling moment are calculated by subtracting the phase value of the reference signal model from the phase value of the real-time acquired signal at that moment. The result is the phase difference at that sampling moment. If the phase difference is positive, it means that the phase of the real-time acquired signal leads the phase of the reference signal model at that sampling moment. If the phase difference is negative, it means that the phase of the real-time acquired signal lags the phase of the reference signal model. This phase lead or lag essentially reflects the change in the propagation path of the microwave signal caused by vibration during blade operation. Blade vibration will change the relative distance between the probe and the blade surface, thereby changing the propagation time of the microwave signal. The change in propagation time will be directly reflected in the phase difference.

[0053] Step 403: Based on the phase difference, analyze the difference in arrival time between the real-time acquired signal and the reference signal model to determine the peak position of the signal. Based on the peak position, determine the arrival time offset of each probe. Specifically, after calculating the phase difference and arrival time offset, proceed to the undersampling process. The core objective of this step is to reduce the amount of data in subsequent signal processing while ensuring that key characteristics of blade vibration, such as vibration frequency and amplitude trends, are not lost. This avoids redundant data consuming excessive computational resources, thereby improving the overall processing efficiency of the monitoring system. Before performing the undersampling operation, a preset sampling interval needs to be determined. For example, the highest common vibration frequency of monitored wind turbine blades is 10Hz. According to the basic principles of signal sampling, the sampling frequency after undersampling must be at least twice the highest vibration frequency, i.e., 20Hz, to fully capture the vibration characteristics. Combined with the sampling frequency of the original data, such as the previously mentioned 21GHz, this corresponds to a sampling interval of 4.76 × 1000 Hz. One point is collected every second, and it is deduced that a valid data point needs to be selected every 5 raw data points. The preset sampling interval is thus determined to be one point every 5 data points. The data to be processed consists of two types: a phase difference data sequence arranged in chronological order of sampling time, and an arrival time offset data sequence corresponding one-to-one with the phase difference data. Both types of data are of the same length, and each data point is labeled with its corresponding sampling time. For example, if the raw data has 1000 sampling points, there are 1000 phase difference values ​​and 1000 arrival time offset values. All data are arranged sequentially from earliest to latest sampling time to ensure consistency in the time dimension. Next, the specific processing... The undersampling operation starts from the first data point of the two data sequences, skips the next four data points, and selects the 1st, 6th, 11th, 16th, and so on data points. The selected data points will be used as the valid data after undersampling. During the extraction process, the sampling time of each selected data point needs to be recorded synchronously to ensure that the phase difference data and the arrival time offset data after undersampling still maintain the relationship of corresponding data at the same sampling time, avoiding data misalignment. For example, the first sampling time corresponds to a phase difference of 20° and an arrival time offset of 0.2ms, the sixth sampling time corresponds to a phase difference of 24° and an arrival time offset of 0.23ms, and so on.

[0054] Taking a single probe monitoring a wind turbine blade as an example, the relevant calculations have already been completed. Specifically, through signal analysis, the phase of the real-time acquired signal is 30°, and the phase of the reference signal model is 10°. Subtracting the two yields a phase difference of 20°. When using a threshold-based peak detection algorithm, the peak detection threshold is first set to 80% of the signal's maximum amplitude. Then, the signal amplitude curve is scanned, and points exceeding the threshold where the amplitude decreases at adjacent points are identified as peak values. This method determines the peak time corresponding to the reference signal model. The peak value of the real-time acquired signal is 10ms. The time difference is 10.2ms, and the difference between the two is 0.2ms.

[0055] In this embodiment, by constructing an analytical signal, the real signal is converted into a complex signal, which facilitates the analysis of the signal phase. In actual vibration monitoring, the phase change is one of the important pieces of information reflecting the vibration state of the blade. The analytical signal can more comprehensively represent the characteristics of the signal and improve the accuracy of vibration signal analysis. Accurately calculating the phase difference can intuitively reflect the phase change of the real-time acquired signal relative to the reference signal model. The phase difference is a direct manifestation of the change in the microwave signal propagation path caused by blade vibration. By monitoring and analyzing the phase difference, minute vibrations of the blade can be detected in a timely manner, providing an important basis for early fault warning and helping to improve the sensitivity of the monitoring system. Determining the arrival time offset can further quantify the impact of blade vibration on microwave signal propagation time. The arrival time offset is related to the vibration displacement of the blade. By analyzing the arrival time offset, information such as the vibration amplitude and frequency of the blade can be understood more accurately, improving the accuracy of blade vibration state assessment and helping to more accurately determine whether the blade is in normal operating condition. Undersampling reduces the amount of data, reduces the complexity and computation of data processing, and improves the processing efficiency of the system. While ensuring the acquisition of key information, undersampling can reduce the pressure of data storage and transmission, enabling the monitoring system to process large amounts of vibration monitoring data more in real time and efficiently, while also reducing the system cost.

[0056] In another preferred embodiment of the present invention, the process of constructing the analytic signal includes transforming the original signal from a single real number form to a form containing both real and imaginary parts. During the transformation process, a complementary signal in another dimension, i.e., a complementary partner, is sought, and the complementary signal is fused with the original signal to construct the analytic signal. This process may include:

[0057] In this embodiment of the invention, in signal processing, real-valued signals, such as microwave echoes, after filtering and preprocessing, mostly contain only single-dimensional information, namely the pattern of signal amplitude change over time. This single-dimensional characteristic makes it impossible for real-valued signals to directly and completely represent the phase characteristics of the signal, such as the phase shift caused by path changes during signal propagation. Phase information is crucial for subsequent analysis of dynamic changes such as blade vibration. Therefore, it is necessary to generate a complementary signal that can supplement the characteristics of the real-valued signal in another dimension. The Hilbert transform is the core means to achieve this goal. Its essence is to perform a specific set of filtering operations on the original real-valued signal, generating a complementary signal through these operations. In the implementation of the complementary signal, the first step is to design a specialized Hilbert filter based on the frequency range of the original real-valued signal. For example, if the microwave echo signal processed previously has a frequency range of 10GHz to 10.5GHz, this filter will not change the amplitude of any frequency component in the original signal; it will only adjust the phase state of different frequency components to ensure that the complementary signal maintains consistency with the original signal in amplitude characteristics, complementing it only in the phase dimension. The phase adjustment rule of the filter is fixed and uniform. For all non-DC frequency components in the original real-valued signal, that is, for each frequency point within the range of 10GHz to 10.5GHz, the filter will uniformly delay the phase of each frequency component by 90°. And for... The DC component (a component with a frequency of 0, which can be filtered out in advance in high-frequency signals such as microwaves to avoid interference) that may exist in the signal will be directly shielded or its phase will be kept unchanged to prevent the DC component from affecting the phase matching relationship between the complementary signal and the original signal. After the filter is designed, the original real-valued signal is input into this Hilbert filter. Inside the filter, the signal will be processed one frequency component at a time according to the preset phase adjustment rules. For example, if the initial phase of the 10 GHz frequency component in the original signal is E, the phase will be delayed to E-90° after filtering; if the initial phase of the 10.25 GHz intermediate frequency component is R, the phase will be delayed to R-90° after filtering, and so on. Components within the target frequency range will undergo a consistent 90° phase delay. After such filtering, the new signal obtained from the filter output is the complementary signal of the original real-valued signal. This complementary signal forms a clear dimensional complementary relationship with the original real-valued signal. That is, the original real-valued signal serves as the foundation, carrying the core information of the signal amplitude changing over time; the complementary signal serves as a supplement, carrying the phase information after 90° phase adjustment. The two always maintain a fixed 90° phase difference. Ultimately, this complementary signal generated by the Hilbert transform, together with the original real-valued signal, constitutes an analytic signal containing both real and imaginary parts. The original real-valued signal serves as the real part of the analytic signal, and the complementary signal serves as the imaginary part of the analytic signal.

[0058] The original signal and the complementary signal obtained through Hilbert transform are fused together to form a complex signal, i.e., an analytic signal. In this way, the original single real-valued microwave echo signal is transformed into an analytic signal containing both real and imaginary parts. The analytic signal constructed in this way retains the amplitude information of the original signal and introduces additional phase information through the imaginary part. In the vibration monitoring scenario, the phase change of the microwave echo signal caused by blade vibration is analyzed, and then the vibration state of the blade is inferred.

[0059] In a preferred embodiment of the present invention, step 5 above, which involves sparsely reconstructing the undersampled signal and solving the vibration displacement equation using the reconstruction result to obtain the vibration displacement information of the blade, may include:

[0060] Step 501 involves evaluating the characteristics of the undersampled signal, including analyzing its amplitude range, frequency distribution, and noise level to obtain signal features. Specifically, this includes: firstly, evaluating the amplitude range, frequency distribution, and noise level of the undersampled signal; secondly, processing the amplitude value of each sampling point in the undersampled signal individually when calculating the amplitude range, specifically by traversing the data sequence of the undersampled signal, such as a combined data sequence containing phase difference and arrival time offset; thirdly, recording the maximum and minimum amplitude values ​​encountered during the traversal; and finally, determining the signal amplitude range by using the recorded maximum value as the upper limit and the minimum value as the lower limit. For example, using the undersampled signal of a wind turbine blade as an example, by traversing its data sequence, the amplitude range is ultimately calculated to be between -2V and 2V. When analyzing the frequency distribution, a Fast Fourier Transform (FFT) is used to convert the signal from the time domain to the frequency domain. The time-domain data sequence of the undersampled signal is first input into the FFT algorithm unit, which automatically divides the frequency intervals according to the data length. By calculating the amplitude corresponding to each frequency point, a spectrum is generated, from which the signal can be visually observed. The distribution of energy across different frequency components helps determine the dominant frequency components. For example, the undersampled signal of the wind turbine blade, after FFT processing, shows that its main frequency components are concentrated between 10Hz and 50Hz, and the amplitudes of these frequency components are significantly higher than other frequencies, indicating that the blade vibration is mainly in this frequency band. When estimating the noise level, since the signal noise in actual monitoring is often close to Gaussian white noise, the noise intensity can be quantified by calculating the standard deviation of the signal. First, calculate the average amplitude value of all sampling points of the undersampled signal, then calculate the difference between the amplitude value of each sampling point and the average value, square the difference, sum the results, and divide by the total number of sampling points to obtain the variance. The square root of the variance is the standard deviation, which can approximately represent the noise intensity. At the same time, the spectrum can be used to help judge the noise impact. In the high-frequency part of the spectrum, such as the energy distribution in the frequency range above 50Hz, if the energy in the high-frequency part is low and uniformly distributed, it indicates that the noise interference to the signal is small. The undersampled signal of the wind turbine blade, after calculation, has a standard deviation of 0.1V, and the energy in the high-frequency part of the spectrum is weak, indicating that the noise level is low.

[0061] Step 502: Based on the signal characteristics, analyze the sparsity of the undersampled signal in different transform domains and identify the transform domain for the final sparse representation. Specifically, the transform domains include Fourier transform domain, wavelet transform domain, and discrete cosine transform domain. First, input the undersampled signal into the corresponding transform unit to complete the conversion from the time domain to each transform domain. That is, the Fourier transform domain is suitable for analyzing periodic signals and will decompose the signal into sine / cosine components of different frequencies; the wavelet transform domain is good at processing non-stationary signals and can simultaneously reflect the local characteristics of the signal in both the time and frequency domains; the discrete cosine transform domain is good at smoothing signals. The signal exhibits excellent performance in reduction and analysis. After the transformation, the number of non-zero coefficients of the signal is counted in each transform domain. Non-zero coefficients refer to coefficients whose amplitude after transformation is greater than a preset small threshold, such as 0.001V. The fewer the number of non-zero coefficients, the sparser the signal is in that transform domain, meaning that most of the signal's energy is concentrated on a few coefficients. By comparing the number of non-zero coefficients in the three transform domains, it was found that the undersampled signal of the wind turbine blade had the fewest non-zero coefficients in the wavelet transform domain, with its energy mainly concentrated on about 20 coefficients. Therefore, the wavelet transform domain was chosen as the transform domain for the final sparse representation.

[0062] Step 503: Based on the transform domain and signal characteristics of the final sparse representation, basis pursuit (BP) is used as the sparse reconstruction algorithm, and the parameters of the sparse reconstruction algorithm are determined, including the number of iterations and the sparsity estimate. Specifically, this includes: using basis pursuit (BP) as the sparse reconstruction algorithm, the goal of which is to find the sparsest solution to approximate the original signal under a given measurement matrix; when determining the number of iterations, different iteration numbers are selected, such as 50, 80, 100, and 120, and the undersampled signal is input into the basis pursuit algorithm for reconstruction. After each reconstruction, the error between the reconstructed signal and the original undersampled signal is compared, such as amplitude error and frequency error. The amplitude error is approximately 12% when the number of iterations is 50; it decreases to 8% when the number of iterations is 80; and stabilizes below 5% when the number of iterations is 100. Increasing the number of iterations further to 120 does not significantly reduce the error, so 100 iterations are chosen as the preset number of iterations. When estimating sparsity, considering the signal characteristics in the wavelet transform domain in step 502—that is, most of the signal energy is concentrated on 20 non-zero coefficients, and the amplitudes of these coefficients account for more than 95% of the total energy—the sparsity estimate is set to 20. This means that only 20 main non-zero coefficients need to be retained during the reconstruction process to fully restore the signal characteristics.

[0063] Step 504: Based on the sparse reconstruction algorithm and its parameters, and combined with the arrangement of the microwave probes and the signal acquisition process, a measurement matrix is ​​constructed. Specifically, this includes: First, the microwave probes are uniformly distributed circumferentially along the blade root, with fixed intervals between adjacent probes, resulting in a periodic distribution of signal acquisition positions. This periodicity is directly reflected in the row vector design of the measurement matrix. The row vectors corresponding to the data acquired by adjacent probes in the matrix have completely consistent numerical distribution patterns, forming a symmetrical structure. This matches the periodic characteristics of the acquisition positions, ensuring that the matrix accurately reflects the influence of the spatial arrangement of the probes on signal acquisition. Second, the attenuation and scattering effects during signal acquisition need to be considered and modeled. After the microwave signal is emitted from the probes, it will experience energy attenuation due to the dielectric constant of the blade material. Based on the dielectric constant of the blade material, the attenuation coefficient is determined to be 0.98. When the signal encounters the blade… Scattering occurs on the surface of the probe, causing slight changes in amplitude. Based on the scattering characteristic test results, a scattering correction factor of 0.99 is set. When constructing the matrix, for each probe's corresponding signal propagation path, the element value at the corresponding position in the matrix is ​​multiplied by the attenuation coefficient and the scattering correction factor. This method quantifies the actual impact of attenuation and scattering on the signal, ensuring that the matrix closely matches the real signal propagation process. Finally, the number of rows and columns of the measurement matrix is ​​determined. The number of rows must match the length of the undersampled signal, as each row corresponds to one sampling point of the undersampled signal. In the wind turbine example, the undersampled signal contains 100 sampling points, so the number of rows in the matrix is ​​set to 100. The number of columns must match the length of the original signal, as each column corresponds to one sampling point of the original signal. In the example, the original signal contains 200 sampling points, so the number of columns in the matrix is ​​set to 200. Finally, a 100×200 measurement matrix is ​​constructed.

[0064] Step 505: Using the measurement matrix, the undersampled signal is input into the sparse reconstruction algorithm for iterative calculation, gradually recovering the sparse representation of the original signal. The signal estimate is continuously updated during the reconstruction process until a preset number of iterations is reached, yielding the reconstructed signal. Specifically, this includes: After inputting the undersampled signal and the constructed measurement matrix into the basis pursuit algorithm, the sparse reconstruction calculation officially begins. The algorithm first initializes the signal estimate as a zero vector. This is because at the start of the iteration, there is no valid information, and the zero vector serves as the most basic starting point. The algorithm then enters the iterative process, with each iteration following a fixed logical progression. Specifically, in each iteration, the algorithm first multiplies the current signal estimate by the measurement matrix to obtain a projected signal. This projected signal can be understood as a simulation of the undersampled signal based on the current estimate. Then, the actual undersampled signal is subtracted from this simulated projected signal, and the difference is the error value for the current iteration. The magnitude of the error value directly reflects the degree of deviation between the current estimate and the actual signal. When adjusting the signal estimate, the key indicator of signal sparsity needs to be considered. Signal sparsity is specifically determined by non- The sparsity index is represented by the number of zero coefficients. It is determined in the wavelet transform domain by first setting a threshold, such as 0.001V. Coefficients with amplitudes greater than this threshold in the signal estimate are considered non-zero coefficients. The total number of these non-zero coefficients is then counted; this count is the current sparsity index. The algorithm aims to minimize the sum of this sparsity index (number of non-zero coefficients) and the previously calculated error value. Based on this goal, the signal estimate is adjusted, focusing on retaining coefficients with higher energy in the wavelet transform domain, i.e., amplitudes much greater than the threshold, because these coefficients are key features of the signal. The algorithm weakens small coefficients whose amplitudes are close to or below a threshold. These coefficients are mostly noise or redundant information. Through this adjustment, the signal estimate is continuously updated and gradually approaches the true signal. As the number of iterations increases, the error value gradually decreases. For example, after 50 iterations, the error value drops to 0.05V; after 80 iterations, the error value further drops to 0.03V; when the iteration reaches the preset 100 iterations, the error value stabilizes within 0.02V. At this point, continuing to iterate has little effect on improving accuracy, so the algorithm stops iterating and finally obtains the reconstructed complete signal.

[0065] Step 506 involves comparing the features of the reconstructed signal with those of the original undersampled signal, including amplitude, frequency, and phase, to verify the effectiveness and accuracy of the reconstructed signal and obtain the verification results. Specifically, when comparing the amplitude features of the reconstructed signal and the original undersampled signal, it is essential to first establish a one-to-one correspondence between the sampling points of the two signals. That is, the first sampling point of the reconstructed signal corresponds to the first sampling point of the original undersampled signal, the second sampling point corresponds to the second sampling point, and so on, ensuring that the signal amplitude is compared at the same moment. Then, each corresponding sampling point is processed sequentially, starting with reading the original... The absolute difference between the amplitude value of the undersampled signal at that sampling point (e.g., the amplitude of the 5th sampling point of the original signal is 1.2V) and the amplitude value of the corresponding sampling point of the reconstructed signal (e.g., the amplitude of the 5th sampling point of the reconstructed signal is 1.17V) is calculated. This absolute difference is then divided by the amplitude value of the original undersampled signal at that sampling point. The result is multiplied by 100% to obtain the amplitude error percentage. After traversing all sampling points in this way, for example, if the amplitude error percentage of all sampling points is controlled within 5%, it means that the deviation between the reconstructed signal and the original signal in the amplitude dimension is extremely small, and the amplitude characteristics of the original signal can be well restored.

[0066] When comparing frequency characteristics, a Fast Fourier Transform (FFT) is first performed on the reconstructed signal and the original undersampled signal, respectively. The time-domain data (i.e., the amplitude value sequence that changes over time) of the two signals are input into the FFT processing unit. The unit converts the time-domain signal into a frequency-domain spectrum graph according to the sampling frequency and data length of the signal. The horizontal axis of the spectrum graph is frequency, and the vertical axis is the signal amplitude (i.e., energy) at the corresponding frequency point. Then, the main peak position is found in the two spectrum graphs. The main peak position refers to the frequency point with the highest amplitude and the most concentrated energy in the spectrum graph. This frequency point represents the main vibration frequency of the signal. Assuming that the main peak frequency of the original undersampled signal is 30Hz and the main peak frequency of the reconstructed signal is 30.5Hz, the absolute difference between the two is calculated to be 0.5Hz. This difference is the frequency error. Since the error is within 1Hz, it indicates that the reconstructed signal accurately retains the main frequency characteristics of the original signal, without obvious frequency shift, and has a high degree of matching with the main frequency of the actual vibration of the blade.

[0067] When conducting phase feature comparison, the phase difference between the reconstructed signal and the corresponding sampling points of the original undersampled signal is calculated one by one based on the one-to-one correspondence of sampling points. First, the phase value of a certain sampling point in the original signal is read (e.g., the phase of the 10th sampling point in the original signal is 25°) and the phase value of the corresponding sampling point in the reconstructed signal (e.g., the phase of the 10th sampling point in the reconstructed signal is 27.2°). The absolute difference between the two is calculated. After traversing all sampling points, the largest absolute phase difference is statistically assumed to be 4.8°, and this value is controlled within 5°. Phase is a key factor reflecting signal propagation characteristics and time offset. The indicators show that a small phase error indicates that the reconstructed signal deviates little from the original signal in terms of time dimension and propagation characteristics, and can accurately reflect the signal phase change caused by blade vibration. Considering the three error indicators of amplitude, frequency and phase, an amplitude error of less than 5% indicates accurate signal amplitude restoration, a frequency error of less than 1Hz indicates complete preservation of the main vibration frequency, and a maximum phase error of no more than 4.8° indicates good matching of time and propagation characteristics. All three indicators prove that the reconstructed signal effectively preserves the core characteristics of the original undersampled signal, and its effectiveness and accuracy meet the needs of blade vibration monitoring.

[0068] Step 507: Based on the verification results, establish the mathematical relationship between the vibration displacement and the reconstructed signal, i.e., the vibration displacement equation. Substitute the reconstructed signal into the vibration displacement equation for calculation to obtain the vibration displacement information of the blade. Step 5071: Determine the real and imaginary parts of the analytic signal in the real-time acquired signal to obtain the phase change of the real-time acquired signal; determine the real and imaginary parts of the reference signal model to obtain the phase change of the reference signal model. Specifically, when processing the real-time acquired signal, first assume that its analytic signal form is... ,in This represents the real part of the signal acquired and analyzed in real time. Representing the imaginary part of the analytic signal, to obtain the phase change of the real-time acquired signal, we need to use the formula... Calculate its phase correlation, where, The adjustment coefficient is set to 0.2. This coefficient adjusts the influence of the sinusoidal function component in the imaginary part of the real-time acquired signal on the phase change calculation. Its value is not arbitrarily set, but determined through multiple measurements of microwave signal changes caused by blade vibration under different operating conditions, such as small-amplitude and large-amplitude blade vibration, or wind speeds of 5 m / s and 10 m / s. These measurements were statistically analyzed and error verified to ensure that the calculated phase correlation accurately reflects the phase characteristics of the actual signal. The adjustment coefficient γ is set to 0.15. This coefficient mainly affects the cosine function component in the real part of the real-time acquired signal, and it interacts with... Mutual cooperation, when When enhancing the influence of the sine function on phase calculation, γ balances the overall calculation by adjusting the weight of the cosine function, jointly optimizing the calculation results of phase changes and making the results more closely resemble the actual signal state during blade vibration; angular frequency The value is 2π×50rad / s, which corresponds to a frequency of 50Hz. 50Hz was chosen because during the operation of this wind turbine, the blade vibration is accompanied by multiple frequency components. Among them, 50Hz is a key frequency related to the main vibration mode of the blade. Using the angular frequency corresponding to this frequency can more accurately capture the signal changes caused by the blade vibration, thereby improving the accuracy of the phase change value calculation.

[0069] For the reference signal model, the form of its analytic signal is set as follows: ,in , The real and imaginary parts of the reference signal model are analyzed separately. When calculating the phase correlation of the reference signal model, the core logic consistent with that of the real-time acquired signal is used, first calculating... and Multiply Sum of the products, then calculate and Multiply The sum of the products of the first and second parts is then divided by the result of the first part. Finally, the arctangent of the quotient obtained from this division is performed. The final value is the phase correlation quantity of the reference signal model. The adjustment coefficient δ is set to 0.3. The core function of this coefficient is to adjust the imaginary part of the reference signal model. The contribution of this sinusoidal function to the phase change calculation results, when As the value increases, the influence of the sine function on phase calculation strengthens; when When the coefficient is reduced, its effect weakens. The purpose is to ensure that the contribution of the sine function matches the actual characteristics of the reference signal model, and to avoid inaccurate calculation of phase correlation due to coefficient deviation; adjustment coefficient The value is set to 0.2, and this coefficient specifically acts on the real part of the reference signal model. This cosine function part, and Coordinated adjustment, when When increasing the value to enhance the contribution of the sine function to phase calculation, By setting the value to 0.2, the cosine function maintains a reasonable weight, preventing it from being excessively weakened; when When decreasing, It also maintains stability and prevents the cosine function from having too much influence. Together, they avoid phase calculation imbalance, optimize the results, and make the phase correlation quantity more accurately reflect the true phase characteristics of the reference signal model. This is the angular frequency corresponding to the reference signal model's base frequency. The determination of the base frequency needs to be based on the monitoring scenario. If it needs to be adapted to the key frequencies of the real-time acquired signals, the base frequency can be set to 50Hz (consistent with the key frequencies related to the main vibration of the blades in the real-time acquired signals). The value is 2π × 50 rad / s; if the microwave signal is set based on the state of the blade without vibration, the reference frequency can be set to 10 GHz (consistent with the operating frequency of microwave detection). The value is 2π× rad / s, Its purpose is to ensure and The frequency characteristics are consistent with the overall frequency characteristics of the reference signal model. For example, when the reference frequency is 50Hz, these two trigonometric functions can simulate the periodic phase fluctuations in the reference signal related to the stable operation of the blade. When the reference frequency is 10GHz, they can simulate the high-frequency periodic characteristics of the microwave signal itself, thereby allowing the trigonometric functions to accurately match the periodic components related to the phase in the reference signal model, further improving the calculation accuracy of the phase correlation quantity. After the phase correlation quantity of the reference signal model is calculated, combined with the previously calculated phase correlation quantity of the real-time acquired signal, the phase change difference between the two is obtained by subtracting the phase correlation quantity of the reference signal model from the phase correlation quantity of the real-time acquired signal. This difference directly reflects the phase shift caused by blade vibration, that is, blade vibration will change the microwave propagation path length, causing the phase of the real-time signal to shift relative to the reference signal model (the signal under no vibration or stable state).

[0070] Step 5072: Determine the time corresponding to the peak position of the real-time acquired signal and the time corresponding to the peak position of the reference signal model, and obtain the time difference between the modulated real-time acquired signal and the reference signal model. Specifically, this includes: determining the time corresponding to the peak position of the real-time acquired signal. In this process, peak detection needs to be performed on the real-time acquired signal. Specifically, starting from the signal's initial moment, the signal strength value at each sampling moment is recorded sequentially. The current moment's strength is compared with the strengths of the previous and next moments. When the current moment's strength is simultaneously greater than the strengths of the previous and next moments, that moment is determined to be the signal peak position and recorded as [peak value]. In actual monitoring, real-time acquired signals are often mixed with thermal noise from equipment circuits, such as minute voltage fluctuations caused by the thermal motion of resistors, and environmental electromagnetic interference, such as electromagnetic radiation from surrounding equipment. These interferences can cause small, irregular fluctuations in the signal within ±5%. Direct detection can easily misinterpret these fluctuations as the true peak values ​​generated by blade vibration. Therefore, preprocessing with filtering techniques is necessary. Low-pass filters can block noise with frequencies higher than the main frequency of blade vibration, such as high-frequency interference above 100Hz. Band-pass filters retain only frequency components from 10Hz to 50Hz (the main frequency range of blade vibration). By filtering out irrelevant frequency bands, false peak values ​​caused by noise are reduced, ensuring... It reflects the actual peak vibration time.

[0071] To further eliminate the impact of system latency, it is necessary to... Added correction items ,in The correction factor is set to 0.06. This value was determined by repeatedly measuring the propagation delay of microwaves emitted from the probe, reflected from the blade, and reaching the receiver in this wind turbine blade vibration monitoring scenario (each measurement was taken 10 minutes apart, for a total of 100 measurements). The average delay was calculated to be 0.06 times the time correction corresponding to the sinusoidal function period. Therefore, this value was set as... =0.06 to compensate for the delay; The angular frequency of the sine function in the correction term is determined by real-time acquisition of the signal and FFT analysis. Its main frequencies are concentrated between 10Hz and 50Hz, with the core frequency of 50Hz used for calculation. ,Right now =2 × pi × 50 rad / s, ensuring the periodicity of the correction term matches the signal oscillation period, and then combine this correction term with... Add them together to obtain the real-time acquisition signal correlation time value including the correction.

[0072] For the reference signal model, which is the baseline signal when the blade is not vibrating, the intensity variation pattern is stable. Similarly, the time corresponding to the peak position is determined by peak detection. ; followed by Added correction items ,in The correction factor is set to 0.05. This value is obtained by statistically analyzing the transmission delay of the reference signal from the database to the processing unit (measured over 1000 calls), resulting in a correction amount corresponding to an average delay of 0.05 cosine function periods. This value is also determined by considering the time error of the signal processing chip (measured error range of 0.01 to 0.09). =0.05 to compensate for these errors, the angular frequency of the cosine function in the correction term. With the real-time acquisition signal correction item To maintain consistency, i.e., 2 × pi × 50 rad / s, and ensure unified correction logic, this correction term is aligned with... Adding them together yields the reference signal model correlation time values, including the corrections.

[0073] Step 5073: Based on the phase change of the real-time acquired signal and the reference signal model, and the time difference between the modulated real-time acquired signal and the reference signal model, the final vibration displacement is obtained. Specifically, after obtaining the phase change difference from step 5071 and the time difference from step 5072, these two parameters are substituted into the vibration displacement equation for calculation. The equation is in the form x(t) = α × [ ]×[Real-time acquired signal phase correlation - Reference signal model phase correlation]+(1-α)×[ [×[Real-time value including correction - Reference time value including correction]], where α is a weighting coefficient, ranging from 0 to 1. Its specific value is based on the influence of the phase change difference and time difference on the vibration displacement calculation result. If the phase change difference is more significantly affected by the blade vibration and the calculation error is smaller, α will be set closer to 1; if the time difference measurement is more accurate and has a more direct impact on the displacement, α will be set closer to 0. λ is the wavelength of the microwave signal, and c is the propagation speed of the microwave signal in the medium. Both of these parameters are known physical quantities. λ can be calculated from the microwave operating frequency and the speed of light in a vacuum. For example, when the operating frequency is 10 GHz, λ = 3 × m / s÷ Hz and c can be calculated based on the dielectric constant of the blade material. For example, when the dielectric constant is 4, c = 3 × m / s÷ After substituting each parameter into the equation, the calculation is performed in the order of first calculating the values ​​inside the parentheses, then multiplication and division, and finally addition and subtraction. The final value x(t) is the vibration displacement of the blade at time t. This value directly reflects the distance of the blade from its initial position without vibration at time t. By recording and analyzing x(t) at different times, the vibration state of the blade can be fully understood.

[0074] This embodiment analyzes the characteristics of undersampled signals, selects appropriate sparse reconstruction algorithms and parameters such as sparse representation transform domain and basis pursuit, constructs an accurate measurement matrix, iteratively recovers the original signal, verifies the effectiveness of the reconstructed signal, and establishes a vibration displacement equation, which can accurately solve the blade vibration displacement, promptly detect anomalies and maintain the system; it determines the phase change between real-time and reference signals, considers the real and imaginary parts of the signal and trigonometric function correction, detects the peak time and obtains the time difference, including trigonometric function correction, and combines the two to calculate the vibration displacement. Multi-parameter fusion and flexible weight adjustment improve the displacement calculation accuracy and monitoring reliability, ensuring the safe operation of the equipment.

[0075] In a preferred embodiment of the present invention, step 6 above, which involves constructing a bidirectional vibration model based on the blade's vibration displacement information and analyzing the circumferential and axial vibration components, triggers a corresponding multi-level alarm mechanism when the vibration amount in either direction exceeds a preset threshold. This mechanism may include:

[0076] Step 601: Based on the arrangement of the microwave detection array and the geometric characteristics of the blades, establish a coordinate system, defining the axial direction as the direction perpendicular to the annular plane and the circumferential direction as the circumferential direction within the plane containing the annulus. Specifically, this includes: first, clarifying the key parameters of the microwave detection array and the blades, including the positional distribution of the microwave detection array, such as whether it is arranged circumferentially, the installation angle of each probe, and the spacing between adjacent probes; second, clarifying the geometric characteristics of the blades, such as whether the cross-sectional shape is circular, whether the overall structure is slender, and the installation position of the blades in the equipment; and establishing a Cartesian coordinate system within the annular plane with the center of the annular plane containing the microwave detection array as the origin. Define the direction perpendicular to the annular plane as the axial direction, represented by the z-axis; and define the direction along the circumference within the annular plane as the circumferential direction, corresponding to the angular direction of the polar coordinates in the Cartesian coordinate system, thereby constructing a unified coordinate system.

[0077] Step 602 involves decomposing the blade vibration into two independent components: circumferential vibration and axial vibration. Specifically, this includes observing and decomposing the blade vibration as a whole. The actual vibration of the blade is a three-dimensional spatial motion. Based on the coordinate system established in step 601, it can be decomposed into two independent directions: circumferential motion (along the circumferential direction in the annular plane) and axial motion (along the z-axis direction). Specifically, the principle of vector decomposition is adopted. The vibration displacement vector of the blade at a certain moment is taken. According to the definition of the circumferential and axial directions in the coordinate system, the displacement vector is decomposed into a circumferential component vector and an axial component vector. Through decomposition, mutually independent circumferential vibration components and axial vibration components are obtained.

[0078] Step 603: Based on the vibration displacement information, analyze the amplitude, frequency, and phase characteristic parameters of the blade vibration. For circumferential vibration, determine the amplitude and frequency by analyzing the positional changes of the blade within the annular plane; for axial vibration, determine the amplitude and frequency by analyzing the positional changes of the blade in the direction perpendicular to the annular plane. Using the characteristic parameters, establish sinusoidal function models for circumferential and axial vibrations. Specifically, this includes extracting characteristic parameters from the vibration displacement information and establishing sinusoidal function models, including a circumferential vibration model and an axial vibration model, wherein the circumferential vibration model is represented as... in It is the amplitude of circumferential vibration. It is the circumferential vibration frequency. It is the phase of circumferential vibration; axial vibration can be expressed as ,in It is the axial vibration amplitude. It is the axial vibration frequency. This refers to the axial vibration phase. For the circumferential vibration component, the position of the blade within the annular plane is continuously recorded over time. The vibration period is determined by measuring the time interval between two adjacent peaks on the curve, and the circumferential vibration frequency is calculated based on frequency = 1 / period. The circumferential vibration amplitude is determined by measuring the maximum position offset (distance from the equilibrium position to the peak position) in the position change curve. Simultaneously, the circumferential vibration phase is extracted, specifically by reading the circumferential displacement value at t=0 and substituting it into the circumferential vibration model to obtain... Then, combining the trend of the displacement curve near t=0 (if the curve shows an upward trend, the phase is between 0 and π; if it shows a downward trend, the phase is between π and 2π), determine... The specific values ​​are as follows: For the axial vibration component, the position of the blade along the z-axis is recorded as a function of time. The axial vibration amplitude and frequency are extracted using the same method. When extracting the axial vibration phase, the axial displacement value at t=0 is read and substituted into the axial vibration model to obtain the following values. The rising and falling trend of the axial displacement curve near t=0 is used to determine .

[0079] Step 604: In the polar coordinate system, the vibration displacement vector is decomposed into circumferential and axial components. For the vibration displacement vector at each moment, decomposition is performed according to the geometric relationship of the vector and the principle of trigonometric functions. Specifically, this includes: transforming the vibration displacement information from the rectangular coordinate system to the polar coordinate system. In the polar coordinate system, the vibration displacement vector of the blade is described by the polar radius (the magnitude of the displacement vector, i.e., the total amplitude of the displacement) and the polar angle (the angle between the displacement vector and the circumferential direction, denoted as M). For the vibration displacement vector at each moment... Based on geometric relationships and trigonometric function principles, component decomposition is performed. Since the polar angle M is the angle between the displacement vector and the circumferential direction, the circumferential component is the projection of the polar radius onto the circumferential direction, i.e. The axial component is the projection of the extreme radius onto the axial direction, i.e. This decomposition allows for the determination of the magnitudes of vibration components in two directions within the polar coordinate system.

[0080] Step 605: Determine the preset thresholds for circumferential and axial vibrations. These preset thresholds are divided into multiple levels, each corresponding to different alarm levels and handling measures. Specifically, this includes: determining the safe range and level of circumferential and axial vibrations; referring to relevant industry standards for wind turbine blade vibration and equipment design requirements, determining 1.0 mm for the circumferential vibration and 0.8 mm for the axial vibration as the maximum permissible vibration amplitude; and combining vibration data from the equipment under fault-free operating conditions over the past year, assuming that the statistical analysis shows the circumferential vibration amplitude is concentrated between 0 and 0.5 mm and the axial vibration between 0 and 0.4 mm, dividing the vibration into three safety levels. Level 1 thresholds are as follows: circumferential vibration amplitude ≤ 0.5 mm and axial vibration amplitude ≤ 0.4 mm. Within this range, the blade structural stress is within a safe value, and the equipment operates stably. Level 2 thresholds are circumferential vibration amplitude 0.5 to 1.0 mm and axial vibration amplitude 0.4 to 0.8 mm. At this point, the vibration slightly exceeds the normal level, and the blade stress is close to the fatigue warning value, requiring continuous monitoring. Level 3 thresholds are circumferential vibration amplitude > 1.0 mm and axial vibration amplitude > 0.8 mm. At this point, the vibration exceeds the material's safe bearing capacity, and the blade may suffer structural damage such as cracks, requiring emergency treatment.

[0081] Step 606: Design a multi-level alarm mechanism. When the circumferential or axial vibration component exceeds a preset threshold, a corresponding alarm signal is triggered. The alarm level is divided according to the degree to which the vibration exceeds the threshold. Specifically, alarm rules and signals are set according to the safety range level. When the circumferential vibration amplitude is >0.5mm and ≤1.0mm, or the axial vibration amplitude is >0.4mm and ≤0.8mm, a level one alarm is triggered, and the monitoring system pop-up window displays "[Equipment Number] Circumferential Vibration 0.6mm (Exceeding Level One Threshold 0.5mm." Please observe this specific prompt information. When the circumferential vibration amplitude is >1.0mm or the axial vibration amplitude is >0.8mm, a level three alarm is triggered. The system immediately generates a stop command, and a pop-up window displays "[Equipment Number] Axial Vibration 0.9mm (Exceeding Level Three)." (Threshold 0.8mm) requires emergency shutdown for maintenance; if the secondary threshold is further refined through data optimization, such as 0.7 to 1.0mm circumferentially and 0.6 to 0.8mm axially, a secondary alarm will be triggered in this range. In addition to the prompt message, an inspection instruction will be generated, requiring operators to check on-site for scratches on the blade surface, whether the microwave detection array probe is offset, and whether the equipment connection bolts are loose. Multi-channel alarm signals are configured for each alarm level, and on-site audible and visual alarms are set (level 1 alarm: flashing green light and low-frequency buzzer; level 3 alarm: constant red light and high-frequency buzzer). SMS alarms are sent to maintenance personnel, including the equipment name (XX wind turbine), vibration direction (axial), current value 0.9mm, and alarm level 3. Email alarms are also pushed to ensure that maintenance personnel receive specific over-limit information in a timely manner.

[0082] Step 607: Continuously collect vibration displacement information, update the data of circumferential and axial vibration components in real time, and determine whether the vibration amount exceeds a preset threshold. When the circumferential or axial vibration component is detected to exceed the preset threshold, a corresponding multi-level alarm mechanism is immediately triggered. Specifically, this includes: continuously collecting blade vibration displacement data at a sampling interval of 10 times per second using a microwave detection array. Each sampling will obtain the current vibration displacement vector of the blade, including the total amplitude and the angle with the circumference. Based on this real-time displacement vector, the circumferential and axial vibration components are calculated using the polar coordinate decomposition method in step 604, thereby obtaining the current vibration displacement data. The system records the circumferential and axial vibration components and updates them in real time. The updated components are then compared with the thresholds set in step 605. If the real-time circumferential vibration component is 0.6mm (exceeding the first-level threshold of 0.5mm but below the second-level threshold of 0.7mm), the first-level alarm set in step 606 is immediately triggered, a monitoring pop-up window is displayed, and a green audible and visual alarm is activated. If the real-time axial vibration component is 0.9mm (exceeding the third-level threshold of 0.8mm), the system automatically sends a shutdown command to the wind turbine control system, simultaneously activating a red audible and visual alarm and sending SMS and email alarms containing specific over-limit values ​​to ensure timely shutdown and prevent damage to the blade structure.

[0083] This embodiment decomposes vibration into two independent components, circumferential and axial, which helps to understand the essential characteristics of blade vibration more deeply. Vibrations in different directions may be caused by different reasons, and analyzing them separately can more accurately locate the problem and improve the accuracy of fault diagnosis. Decomposing the vibration displacement vector in polar coordinates can more clearly show the relationship between the circumferential and axial components. Using the principle of trigonometric functions for decomposition makes the calculation process simpler and more accurate, improving analysis efficiency. Determining preset thresholds and classifying them into levels provides a clear standard for the safety assessment of blade vibration. Different threshold levels can correspond to different risk levels, facilitating the implementation of corresponding handling measures to ensure the safe operation of the equipment. The design of a multi-level alarm mechanism can promptly detect abnormal blade vibration and take different alarm measures according to the degree of abnormality, ensuring that relevant personnel can understand the equipment status in a timely manner, take effective handling measures, avoid accidents, and promptly determine whether the threshold is exceeded and trigger the alarm mechanism, enabling rapid response to abnormal situations, reducing equipment damage and downtime, and improving equipment reliability and operating efficiency.

[0084] like Figure 2 As shown, embodiments of the present invention also provide a vibration monitoring system based on a microwave probe, comprising:

[0085] A microwave detection array module is used to arrange a microwave detection array around the root of a wind turbine blade. Each microwave detection pin contains a microwave transmitter and a receiver, and the microwave detection pins are arranged in a ring at a fixed interval θ.

[0086] The signal processing module is used to control the microwave detection array to emit linear frequency modulated microwave signals to the surface of the fiberglass composite material when the blade is stationary, and to receive the reflected echoes and record the initial phase of each probe in the corresponding state; when the blade is running, it collects the microwave echo signals received by each probe in real time.

[0087] The reference model building module is used to calculate the electromagnetic wave propagation speed based on the dielectric constant of the blade material, and combine it with the geometric parameters of the probe installation position to build a vibration-free reference model that includes the reference arrival time and initial phase.

[0088] The feature extraction module is used to compare the microwave echo signals of each probe with the reference signal model under vibration-free conditions, construct an analytical signal, calculate the phase difference and arrival time offset, and perform undersampling processing to obtain the undersampled signal.

[0089] The signal reconstruction module is used to perform sparse reconstruction of the undersampled signal and solve the vibration displacement equation based on the reconstruction results to obtain the vibration displacement information of the blade.

[0090] The alarm triggering module is used to construct a two-way vibration model based on the vibration displacement information of the blade, and analyze the circumferential vibration component and the axial vibration component. When the vibration amount in either direction exceeds the preset threshold, the corresponding multi-level alarm mechanism is triggered.

[0091] It should be noted that this system is a system corresponding to the above method. All implementation methods in the above method embodiments are applicable to this embodiment and can achieve the same technical effect.

[0092] Embodiments of the present invention also provide a computing device, including: a processor and a memory storing a computer program, wherein the computer program, when executed by the processor, performs the method described above. All implementations in the above method embodiments are applicable to this embodiment and can achieve the same technical effects.

[0093] Embodiments of the present invention also provide a computer-readable storage medium storing instructions that, when executed on a computer, cause the computer to perform the method described above. All implementations in the above method embodiments are applicable to this embodiment and can achieve the same technical effects.

[0094] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method of vibration monitoring based on a microwave probe needle, characterized by, The method comprises: Arranging a microwave detection array circumferentially along the root of a wind power blade, each microwave detection needle containing a microwave transmitter and receiver and arranged in a ring distribution with a fixed interval θ; In a static state of the blade, controlling the microwave detection array to emit a linear frequency modulation microwave signal to the surface of the glass fiber reinforced plastic composite material and receive the reflected echo, and recording the initial phase of each detection needle in the corresponding state; According to the dielectric constant of the blade material, calculating the electromagnetic wave propagation speed, combining the geometric parameters of the installation position of the detection needle, and establishing a non-vibration reference model containing the reference arrival time and the initial phase; In a running state of the blade, real-time collecting the microwave echo signals received by each detection needle, extracting the phase change and arrival time offset of the real-time collected signals relative to the reference signal model, and obtaining an under-sampling signal; Performing sparse reconstruction on the under-sampling signal, and solving the vibration displacement equation through the reconstruction result to obtain the vibration displacement information of the blade; According to the vibration displacement information of the blade, constructing a bidirectional vibration model, and analyzing the circumferential vibration component and the axial vibration component, when the vibration amount in any direction exceeds a preset threshold, triggering a corresponding multi-level alarm mechanism.

2. The microwave probe needle-based vibration monitoring method according to claim 1, characterized by, According to the dielectric constant of the blade material, calculating the electromagnetic wave propagation speed, combining the geometric parameters of the installation position of the detection needle, and establishing a non-vibration reference model containing the reference arrival time and the initial phase, comprising: By analyzing the time difference between the transmitted signal and the reflected echo signal, combining the propagation speed of the microwave signal in the medium, calculating the time elapsed from transmission to reception, i.e. the reference arrival time of each detection needle; For each detection needle, record the time difference of each measurement and calculate the average of all time differences, and take the average of all time differences as the final reference arrival time; For the initial phase recorded for each detection needle, calculate the phase difference by comparing it with a reference signal of a phase, and compensate the initial phase to obtain the calibrated and compensated initial phase; According to the reference arrival time and the calibrated and compensated initial phase, determine the reference signal model parameters in the non-vibration state; According to the reference signal model parameters in the non-vibration state, including the reference arrival time and the calibrated initial phase, construct the reference signal model in the non-vibration state.

3. The microwave probe needle-based vibration monitoring method of claim 2, wherein, In a running state of the blade, real-time collecting the microwave echo signals received by each detection needle, extracting the phase change and arrival time offset of the real-time collected signals relative to the reference signal model, and obtaining an under-sampling signal, comprising: Comparing the microwave echo signals of each detection needle with the reference signal model in the non-vibration state, and constructing an analytical signal for the reference signal model and the real-time collected microwave echo signal respectively; By calculating the inverse tangent value of the ratio of the imaginary part to the real part in the analytical signal, obtaining the phase of the real-time collected signal and the reference signal model, and according to the phase of the real-time collected signal and the phase of the reference signal model, calculating the phase difference; According to the phase difference, analyzing the difference in arrival time between the real-time collected signal and the reference signal model, determining the peak position of the signal, and according to the peak position of the signal, determining the arrival time offset of each detection needle; Performing under-sampling processing on the phase difference and the arrival time offset to obtain an under-sampling signal.

4. The microwave probe needle-based vibration monitoring method according to claim 3, characterized by, The construction process of the analytic signal includes transforming the original signal, so that the signal changes from a single real number form to a form containing real and imaginary parts, and in the transformation process, the complementary signal, i.e. the complementary partner, of the signal in another dimension is sought, and the complementary signal is fused with the original signal to construct the analytic signal.

5. The microwave probe needle-based vibration monitoring method of claim 4, wherein, The undersampled signal is subjected to sparse reconstruction, and the vibration displacement equation is solved by using the reconstruction result to obtain the vibration displacement information of the blade, including: The characteristics of the undersampled signal are evaluated, including analyzing the amplitude range, frequency distribution and noise level of the signal to obtain the signal characteristics; According to the signal characteristics, the sparsity of the undersampled signal in different transform domains is analyzed to identify the transform domain of the final sparse representation; According to the transform domain of the final sparse representation and the signal characteristics, the basis pursuit (BP) is used as the sparse reconstruction algorithm, and the parameters of the sparse reconstruction algorithm, including the iteration number and the sparsity estimation value, are determined; According to the sparse reconstruction algorithm and the parameters of the sparse reconstruction algorithm, and in combination with the arrangement mode of the microwave detection needle and the signal acquisition process, a measurement matrix is constructed; The undersampled signal is input into the sparse reconstruction algorithm using the measurement matrix, and iterative calculation is performed to gradually recover the sparse representation of the original signal, and the signal estimate is continuously updated during the reconstruction process until the preset iteration number is reached to obtain the reconstructed signal; The characteristics of the reconstructed signal and the original undersampled signal are compared, including amplitude, frequency and phase, to verify the effectiveness and accuracy of the reconstructed signal, and a verification result is obtained; According to the verification result, a mathematical relationship between the vibration displacement and the reconstructed signal, i.e. a vibration displacement equation, is established, and the reconstructed signal is substituted into the vibration displacement equation to calculate the vibration displacement information of the blade.

6. The microwave probe needle-based vibration monitoring method of claim 5, wherein, According to the verification result, a mathematical relationship between the vibration displacement and the reconstructed signal, i.e. a vibration displacement equation, is established, and the reconstructed signal is substituted into the vibration displacement equation to calculate the vibration displacement information of the blade, including: The real part and the imaginary part of the analytic signal in the real-time collected signal are determined to obtain the phase change of the real-time collected signal; the real part and the imaginary part of the reference signal model are determined to obtain the phase change of the reference signal model; The time corresponding to the peak value position of the real-time collected signal and the time corresponding to the peak value position of the reference signal model are determined to obtain the difference between the modulated real-time collected signal time and the reference signal model time; According to the phase changes of the real-time collected signal and the reference signal model, and the difference between the modulated real-time collected signal time and the reference signal model time, the final vibration displacement is obtained.

7. The microwave probe needle-based vibration monitoring method of claim 6, wherein, According to the vibration displacement information of the blade, a bidirectional vibration model is constructed, and the circumferential vibration component and the axial vibration component are analyzed, and when the vibration amount in any direction exceeds a preset threshold, a corresponding multi-level alarm mechanism is triggered, including: According to the arrangement mode of the microwave detection array and the geometric characteristics of the blade, a coordinate system is established, and the axial direction is defined as the direction perpendicular to the annular plane, and the circumferential direction is defined as the circumferential direction in the annular plane; The vibration of the blade is decomposed into two independent components, i.e. the circumferential vibration and the axial vibration; According to the vibration displacement information, the amplitude, frequency and phase characteristic parameters of the blade vibration are analyzed, for the circumferential vibration, the amplitude and frequency are determined by analyzing the position change of the blade in the annular plane; for the axial vibration, the amplitude and frequency are determined by analyzing the position change of the blade in the direction perpendicular to the annular plane; the characteristic parameters are used to establish the sine function model of the circumferential vibration and the axial vibration; In the polar coordinate system, the vibration displacement vector is decomposed into circumferential component and axial component, for the vibration displacement vector at each time, the decomposition is carried out according to the geometric relationship of the vector and the principle of trigonometric function; The preset threshold values of the circumferential vibration and the axial vibration are determined, the preset threshold values are divided into multiple levels, each level corresponds to different alarm levels and processing measures; A multi-level alarm mechanism is designed, when the circumferential vibration component or the axial vibration component exceeds the preset threshold value, the corresponding alarm signal is triggered, and the alarm level is divided according to the degree of vibration exceeding the threshold value; The vibration displacement information is continuously collected, the data of the circumferential vibration component and the axial vibration component are updated in real time, and it is judged whether the vibration exceeds the preset threshold value, when the circumferential vibration component or the axial vibration component exceeds the preset threshold value is detected, the corresponding multi-level alarm mechanism is triggered immediately.

8. A microwave probe based vibration monitoring system implementing the method of any one of claims 1 to 7, characterized in that, It comprises: A microwave detection array module is arranged along the root of the wind turbine blade, each microwave detection needle contains a microwave transmitter and receiver, and each microwave detection needle is arranged in a ring shape with a fixed interval θ; A signal processing module is used to control the microwave detection array to emit a linear frequency modulation microwave signal to the surface of the glass fiber reinforced plastic composite material in a static state of the blade, receive the reflected echo, and record the initial phase of each detection needle under the corresponding state; In the running state of the blade, the microwave echo signals received by each detection needle are collected in real time; A reference model establishment module is used to calculate the electromagnetic wave propagation speed according to the dielectric constant of the blade material, and establish a non-vibration reference model containing the reference arrival time and the initial phase based on the geometric parameters of the installation position of the detection needle; A feature extraction module is used to compare the microwave echo signals of each detection needle with the reference signal model under the non-vibration state, construct the analysis signal, calculate the phase difference and arrival time offset, and perform undersampling processing to obtain the undersampling signal; A signal reconstruction module is used to perform sparse reconstruction on the undersampling signal, and solve the vibration displacement equation through the reconstruction result to obtain the vibration displacement information of the blade; An alarm triggering module is used to construct a bidirectional vibration model according to the vibration displacement information of the blade, analyze the circumferential vibration component and the axial vibration component, and trigger the corresponding multi-level alarm mechanism when the vibration in any direction exceeds the preset threshold value.

9. A computing device, comprising: It comprises: One or more processors; A storage device is used to store one or more programs, when the one or more programs are executed by the one or more processors, the one or more processors implement the method of any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer readable storage medium stores a program, which is executed by the processor to implement the method of any one of claims 1 to 7.

Images