Passive icing detection method for plate-shaped structure based on environmental noise

Abstract

The invention provides a plate-shaped structure passive icing detection method based on environmental noise, and belongs to the field of structure health monitoring, and the method comprises the following steps: S1, completing sensor arrangement according to a sensor arrangement strategy; s2, acquiring a real-time passive guided wave signal of the plate-shaped structure; s3, according to the passive guided wave signal, a complex wavenumber trajectory of the plate-shaped structure is obtained, and the complex wavenumber trajectory comprises an attenuation coefficient curve and a frequency dispersion curve; and S4, according to the complex wavenumber trajectory, performing icing early warning and predicting the thickness of an ice layer on the plate-shaped structure. According to the method, the icing condition of the structure is detected and early warned by identifying the complex wave number track, namely the complex wave number space. In the identified complex wave number space, the real part and the imaginary part of the wave number correspond to a frequency dispersion curve and an attenuation coefficient curve respectively, and ice accumulation on the thin plate structure causes the two curves to shift. The identification of the excursion of the two quantities can estimate the icing thickness and realize icing early warning, which has potential application value in passive monitoring of aircraft icing.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of structural health monitoring, and particularly relates to a plate structure passive icing detection method based on environmental noise. BACKGROUND

[0002] Structural health monitoring technology is a core means to ensure aviation safety and reduce maintenance costs. Ultrasonic guided wave technology has become one of the important technologies for defect detection of large and complex structures such as wings and fuselage thin-walled components, because it has the ability of long-distance propagation and large-scale detection. Ultrasonic guided waves can realize the identification of cracks, corrosion, icing and other damages or state abnormalities through the changes of reflection, refraction or attenuation characteristics during propagation in the structure, and have the advantages of high detection efficiency and no need for full-coverage sensor arrangement.

[0003] In the prior art, ultrasonic guided wave monitoring methods are mainly divided into active and passive types: Active ultrasonic guided wave method: a specific frequency guided wave signal is injected into the structure through artificial excitation devices such as piezoelectric transducers and electromagnetic ultrasonic transducers, and then the echo signal is collected by receiving sensors for analysis. This method has been widely used in the monitoring of industrial pipelines, bridges and other ground structures. For example, the active guided wave system based on phased array technology can realize high-precision positioning of pipeline corrosion, and the detection range can reach dozens of meters. However, in the field of aviation, the active method has significant limitations: the weight of traditional excitation devices usually exceeds 500g per channel, and the power consumption is as high as several watts. However, the weight and power consumption of on-board devices are strictly limited in modern commercial aircraft, such as the weight of each square meter of structure should be less than 100g, and the average power consumption should be less than 1W, which makes it difficult to integrate the active ultrasonic guided wave system into lightweight key structures such as wings and tail wings, and cannot meet the needs of online real-time monitoring.

[0004] Passive ultrasonic guided wave method: without artificial excitation, natural noise such as airflow disturbance, mechanical vibration and engine noise in the structure environment is directly used as a probe source, and the change of guided wave characteristics formed by the propagation of noise in the structure is analyzed to realize monitoring. During flight, the aircraft structure is continuously excited by airflow-induced vibration, engine vibration and component movement. The frequency range of airflow-induced vibration turbulence noise is generally 10-1000Hz, and the frequency range of engine vibration is 20-2000Hz. These environmental noises can form stable passive guided wave sources, providing natural conditions for real-time monitoring.

[0005] For the special state monitoring of the thin-walled structure of an aircraft, such as icing state, existing research shows that the icing of the structure will change the physical parameters such as the stiffness and density of the material, and then change the complex wavenumber characteristics of the guided wave propagation, and the icing thickness and distribution state can be quantitatively characterized by the shift of the complex wavenumber trajectory. As the core parameter for describing the propagation speed and attenuation characteristics of the guided wave, the accurate estimation of the complex wavenumber is the key to realize the icing state monitoring. However, the traditional wavenumber estimation methods such as spatial Fourier transform and multiple signal classification algorithm have strict assumptions: the accurate position of the noise source and the spectral characteristics of the excitation signal need to be known, and the full-field response data of the structure need to be collected, usually dozens or even hundreds of sensors need to be arranged. However, in actual aviation applications, the weight and space limit the sensor arrangement, and it is difficult to arrange the number of sensors required by the traditional method, so the traditional method is difficult to effectively identify the complex wavenumber space, which greatly limits the engineering application of passive ultrasonic guided wave technology in aviation structure health monitoring.

[0006] Ultrasonic guided waves are a powerful tool for structural health monitoring (SHM) due to their advantages of high efficiency and wide range monitoring. However, the active ultrasonic guided wave method usually requires heavy and high-power wave excitation equipment, which is difficult to meet the strict requirements of lightweight aviation structures, and therefore cannot be used for online monitoring of aviation structures with strict requirements for lightweight design. In contrast, the passive detection method uses environmental noise as a probe, which is more suitable for SHM of aircraft. Especially during flight, environmental noise has a significant impact on aircraft structures due to airflow induction or mechanical vibration. The icing state of the thin-walled structure can be characterized by the complex wavenumber trajectory. However, the wavenumber estimation method usually assumes that the source position and excitation signal are known, and the full-field response on the structure is measured, therefore, how to effectively identify the complex wavenumber space from the passive environmental noise measured by a small number of sensors is still a pending problem. SUMMARY

[0007] In order to solve the above problems of the prior art, the present application provides a plate structure passive icing detection method based on environmental noise. The arrangement of passive sensors is guided by a non-uniform linear array design strategy, environmental noise signals are collected in real time, and the impulse response function of the structure is retrieved, which is matched with the theoretical model to determine the complex wavenumber trajectory of the structure, i.e. the relationship between the complex wavenumber and the wave frequency, and the icing state of the structure is determined based on the identified complex wavenumber trajectory.

[0008] Specifically, the present application discloses a plate structure passive icing detection method based on environmental noise, which comprises the following steps: S1, according to the sensor arrangement strategy, complete the sensor arrangement, specifically including the following sub-steps: S11, set the minimum distance d between adjacent sensors and the distance D between the first and last sensors; S12. Given the number of sensors M, there are a total of For sensors, and The maximum number of sensors indivual value, For the The sensor and The distance between the sensors, For the sensor coordinates, set and ,exist Iteratively add sensors on the right , m = 3,···, M, where is an integer multiple of d, ; S2, obtaining real-time passive guided wave signals of the plate structure; S3. Obtain the complex wave number trajectory of the plate structure based on the passive waveguide signal. The complex wave number trajectory includes the attenuation coefficient curve and the dispersion curve. The complex wave number is calculated using the following formula: ; in, is the complex wave number The objective function matrix, Frequency is the structural impulse response function measured by the sensor and processed by coda wave elimination, is the theoretical structural impulse response function of the sensor pair, is the impulse response vector of the measurement structure constructed by different sensor pairs, Theoretical structural impulse response vectors constructed for different sensor pairs; S4. Based on the complex wave number trajectory, an ice early warning is performed and the thickness of the ice layer on the plate structure is predicted, which includes the following sub-steps: S41, performing icing warning by attenuation coefficient curve offset and Bootstrap-based hypothesis test, and then proceeding to step S42 after the icing warning; S42. Predicting the thickness of ice on plate-like structures using dispersion curve migration.

[0009] Preferably, step S3 specifically includes the following sub-steps: S31, calculating the cross-correlation function between all sensor pairs; S32, performing a whitening operation on the cross-correlation function; S33, performing tail wave elimination on the whitened time domain structure impulse response; S34. Calculate the complex wave number using the inverse wave correlation method formula and obtain the complex wave number trajectory.

[0010] Preferably, the calculation of the cross-correlation function in step S31 is specifically: The cross-correlation processing is performed on the passive guided wave signals measured at positions and , and a cross-correlation signal is obtained, which is processed by the following formula: ; Wherein, represents the cross-correlation signal, represents a preset time interval, and the cross-correlation signal is obtained by moving a discrete time signal step by step and repeatedly calculating the correlation between the two signals.

[0011] Preferably, the elimination of the tail wave in step S33 is specifically: the tail wave component is eliminated by applying a window function to the time domain impulse response function estimation between the two sensors after whitening in step S32: ; Wherein, is a step function window: ; Wherein, is the direct wave travel time of the sensor pair; After eliminating the tail wave, the frequency domain structural impulse response function is obtained by Fourier transform: ; Preferably, The main peak corresponding to the direct wave travel time of the sensor pair is determined by identifying the whitened cross-correlation.

[0012] Preferably, step S34 specifically includes the following sub-steps: S341, the expression of the theoretical structural impulse response function of the sensor pair in the plate structure is: ; Wherein, is the complex wave number to be calculated, , is the mass per unit area of the plate, is the first kind of Hankel function, is , is the real part, is the imaginary part; S342, according to the above equation, two vectors are constructed for all different sensor pairs: ; ; S343, given a frequency , the complex wave number The normalized inner product of the two vectors is calculated to obtain S344, repeat step S343, give a plurality of frequencies , obtain the complex wave number The objective function matrix is obtained, and the complex wave number trajectory is obtained.

[0013] Preferably, the method for drawing the dispersion curve or the attenuation coefficient curve in step S34 is: drawing the objective function matrix as The color map of Or The color map of The dispersion curve or the attenuation coefficient curve can be intuitively identified, or by selecting the Or The dispersion curve or the attenuation coefficient curve corresponding to the maximum value of the objective function at each frequency

[0014] Preferably, the ice warning in step S41 is based on the Bootstrap-based hypothesis test by the attenuation coefficient curve offset, and the specific steps are as follows: S411, assuming that the plate structure has no ice, i.e. And set the alternative hypothesis as ; S412, given a measurement of environmental noise, the attenuation coefficient at different frequencies is defined as the imaginary part of the complex wave number obtained in step S3 The test statistic of the test is defined as the average attenuation coefficient in the given frequency range: ; Where, The test statistic is The number of selected frequencies; S413, calculate the average test statistic: ; Where, S is the number of passive experiments, and the superscript * indicates that the data or the test statistic comes from Bootstrap resampling, and the resampling process is repeated B times to obtain the Sort In ascending order, that is ; S414, for any β∈(0,1), the Bootstrap test statistic of the 100β percentile is : ; Where [·] represents rounding to the nearest integer operation; ​S415, implement the hypothesis test of icing early warning, given any new data of unknown icing conditions, derive the test statistic alpha, for any significance level beta e (0, 1), if beta satisfies the following formula, accept the null hypothesis H0: ; Otherwise, reject H0 and accept the alternative hypothesis H1, trigger the icing alarm.

[0015] Preferably, the calculation method of the average test statistic in step S413 is specifically: derive the probability distribution of the test statistic alpha under the null hypothesis H0, assume that the passive experiment is repeated S times under the no-icing condition, derive a test statistic for each experiment result, denoted as , then randomly sample S samples from with replacement, denoted as , and calculate the average test statistic using the average value: Preferably, step S42 is specifically: record the dispersion curve as , and calculate the ice thickness by minimizing the distance between and the theoretical wavenumber solution of the double-layer ice-plate structure wave model .

[0016] Compared with the prior art, the beneficial effects of the present application are as follows: 1. The present application can minimize the number of measurement points and maximize the samples in the wavenumber domain by optimizing the sensor arrangement, which is crucial for the potential application of passive SHM of aircraft, can ensure the accuracy of measurement with fewer measurement points, and has important application prospects in engineering practice.

[0017] 2. The present application extends the detection method to passive application scenarios, in the case where the position and signal of the random environmental noise source are unknown, the impulse response function of the structure is retrieved from the environmental noise and matched with the theoretical model to estimate the complex wavenumber trajectory, the complex wavenumber trajectory is used to complete the icing early warning and ice thickness calculation, the calculation is simple and the precision is high, and the application range is wide.

[0018] 3. The present application detects and warns the icing condition of the structure by identifying the complex wavenumber trajectory, i.e. the complex wavenumber space, in the identified complex wavenumber space, the real part and the imaginary part of the wavenumber correspond to the dispersion curve and the attenuation coefficient curve respectively, and the ice accumulation on the thin plate structure will cause the two curves to shift, thereby accurately performing icing early warning and ice thickness identification. In engineering practice, identifying the shift of the two quantities can estimate the icing thickness and realize icing early warning, which has potential application value in passive monitoring of aircraft icing and can be applied in various application scenarios. BRIEF DESCRIPTION OF DRAWINGS

[0019] Figure 1 Schematic diagram of the method flow of the present invention; Figure 2 Schematic diagram of the power spectrum density of the vibration signal measured by the sensor in an embodiment of the present invention; Figure 3 Schematic diagram of cross-correlation functions calculated for sensor pairs with different spacings in an embodiment of the present invention; Figure 4 In the embodiment of the present invention, under ice-free conditions, The objective function is related to the real wave number and frequency Schematic diagram of the results; Figure 5 In the embodiment of the present invention, under the condition of 1mm ice thickness, The objective function is related to the real wave number and frequency Schematic diagram of the results; Figure 6 For the present invention The objective function is related to the real wave number and frequency Schematic diagram of the results; Figure 7 The embodiment of the present invention is directed to Figures 4-6 ,extract Frequency corresponding to the maximum value of the objective function - real wave number Schematic diagram of the results; Figure 8 Schematic diagram of the variation of the attenuation coefficient with frequency based on the maximization of the IWC objective function under the conditions of no ice, 1 mm ice thickness, and 3 mm ice thickness in an embodiment of the present invention; Figure 9 Schematic diagram of ice thickness estimation results when the ice thickness is 0 mm in an embodiment of the present invention; Figure 10 Schematic diagram of ice thickness estimation results when the ice thickness is 1 mm in an embodiment of the present invention; Figure 11 Schematic diagram of ice thickness estimation results when the ice thickness is 3 mm in an embodiment of the present invention; Figure 12 It is assumed that there is no ice in the embodiment of the present invention. Below is a diagram of the test statistic histogram and the 95% quantile threshold for B = 1000 resampling times. DETAILED DESCRIPTION

[0020] Hereinafter, embodiments of the present invention will be described with reference to the accompanying drawings.

[0021] Specifically, the present invention discloses a passive icing detection method for plate structures based on ambient noise, such as Figure 1 As shown, it includes the following steps: S1, complete sensor arrangement according to sensor arrangement strategy, specifically including the following sub-steps: S11, set the minimum distance d between adjacent sensors and the distance D between the first and last sensors.

[0022] S12, given the number of sensors M, there are for sensors, and for sensors, at most values, is the distance between the first sensor and the first sensor, is the coordinate of the sensor, set and , and add sensors iteratively on the right side , m = 3, ···, M, where is an integer multiple of d, ; In this embodiment, the adjacent sensors are arranged non-uniformly, and the arrangement strategy described above can minimize the number of measurement points and maximize the samples in the wave number domain, which can ensure the accuracy of ice detection on the basis of as few sensors as possible. In specific embodiments, the environmental noise measured by a small number of sensors is used to passively identify the complex wave number trajectory of the thin plate structure, and is used for passive ice detection of the plate structure, realizing ice alarm and ice thickness estimation.

[0023] S2, obtain the real-time passive guided wave signal of the plate structure. In specific embodiments, the M sensors are used to obtain the passive guided wave signal generated by the environmental noise in the plate structure in real time. The signal measured by the first sensor is represented as , and T is the total measurement time.

[0024] In this embodiment, the passive guided wave signal can be represented as a discrete time signal by continuous signal sampling; where represents the sampling time interval, n = 1, … N, and N is a positive integer. Figure 2 The power spectral density of the vibration signal measured by the sensor is shown in the schematic diagram.

[0025] S3, obtain the complex wave number trajectory of the plate structure according to the passive guided wave signal, and the complex wave number trajectory includes the attenuation coefficient curve and the dispersion curve. The complex wave number is calculated by the following formula: where is the objective function matrix, is the frequency​ is a measured structural impulse response function of a sensor pair after tail wave elimination processing, is a theoretical structural impulse response function of a sensor pair. is a measured structural impulse response vector constructed by different sensor pairs, is a theoretical structural impulse response vector constructed by different sensor pairs.

[0026] Step S3 specifically comprises the following sub-steps: S31, calculating the cross-correlation function between all sensor pairs, the specific steps of calculating the cross-correlation function are as follows: correlating the passive guided wave signals measured at positions and , obtaining a cross-correlation signal, and processing by the following formula: ; wherein, represents the cross-correlation signal, represents a preset time interval, and the cross-correlation signal is obtained by moving a discrete time signal step by step and repeatedly calculating the correlation between the two signals. Figure 3 The cross-correlation functions calculated by sensor pairs with different spacings are shown.

[0027] S32, performing whitening operation on the cross-correlation function, and performing inverse Fourier transform to obtain the whitened time-domain impulse response function estimate between the sensor pairs: .

[0028] wherein, the whitened cross-correlation function: .

[0029] S33, performing tail wave elimination on the whitened time-domain structural impulse response, and after eliminating the tail wave, the Fourier transform of the plate structural impulse response function is represented as: .

[0030] Step S33 specifically eliminates the tail wave by applying a window function to the time-domain impulse response function estimate between the two sensors after whitening in step S32: ; wherein, is a step function window: ; wherein, is the direct wave travel time of the sensor pair; wherein, in specific embodiments, is determined by identifying the main peak of the whitened cross-correlation, and the main peak corresponds to the direct wave travel time of the sensor pair.

[0031] After eliminating the coda wave, the frequency domain structure impulse response function is obtained through Fourier transform: .

[0032] S34. Identify the complex wave number trajectory using the inverse wave correlation method calculation formula.

[0033] Step S34 specifically includes the following sub-steps: S341. The expression of the theoretical structural impulse response function of the sensor pair in the plate structure is: in, , is the complex wave number to be calculated, , is the mass of the plate per unit area, is the Hankel function of the first kind, and i is the imaginary unit. is the real part, is the imaginary part.

[0034] S342. According to the above equation, construct two vectors for all different sensor pairs: S343, given a frequency , complex wave number It is obtained by calculating the normalized inner product of the two vectors mentioned above; S344, then repeat step S343, given multiple frequencies , and the complex wave number is obtained The objective function matrix is ​​obtained to obtain the complex wave number trajectory. Figures 4-6 Shows the different ice thickness conditions The objective function is related to the real wave number and frequency The result, Figure 7 Shows the extraction Figures 4-6 middle Frequency corresponding to the maximum value of the objective function - real wave number Result diagram, Figure 8 A schematic diagram of the variation of the estimated attenuation coefficient with frequency to maximize the IWC objective function under no ice, 1 mm ice, and 3 mm ice conditions is shown. It can be seen from the figure that the attenuation coefficient curve increases exponentially with frequency, and the attenuation rate is generally proportional to the ice thickness.

[0035] In a specific embodiment, the method for drawing the dispersion curve or the attenuation coefficient curve in step S34 is: draw the objective function matrix as about Or Regarding the color map of or the attenuation coefficient curve can be visually identified.

[0036] S4, according to the complex wave number trajectory, icing warning is carried out and the thickness of the ice layer on the plate structure is predicted, which includes the following sub-steps: S41, icing warning is carried out through the attenuation coefficient curve offset and Bootstrap-based hypothesis testing, and the specific steps are as follows: S411, assuming that the plate structure has no ice, i.e. , and setting the alternative hypothesis as ; S412, given one environmental noise measurement, the attenuation coefficient at different frequencies is defined as the imaginary part of the complex wave number obtained in step S3 , and the test statistic of the test is defined as the average attenuation coefficient in the given frequency range: ; wherein is the test statistic, is the number of selected frequencies; S413, the average test statistic is calculated: ; wherein , S is the number of passive experiments, and the superscript * indicates that the data or the test statistic comes from Bootstrap resampling instead of the original data, and the resampling process is repeated B times to obtain the of the B Bootstrap samples of the test statistic under the null hypothesis H0. is sorted in ascending order, i.e. ; S414, for any β ∈ (0, 1), the Bootstrap test statistic of the 100β percentile is: ; ; wherein [·] represents rounding to the nearest integer. Figure 12 The test statistic diagram under the conditions of no ice, ice thickness of 1 mm and ice thickness of 3 mm is shown.

[0037] S415, the hypothesis test for icing warning is realized, given new data under any unknown icing condition, the test statistic α is derived, and for any significance level β ∈ (0, 1), if β satisfies the following formula, the null hypothesis H0 is accepted: ​​ ; Otherwise, reject H0 and accept the alternative hypothesis H1, triggering an ice alarm.

[0038] Preferably, the calculation method of the average test statistic in step S413 is specifically as follows: derive the probability distribution of the test statistic α under the null hypothesis H0, assume that the passive experiment is repeated S times under ice-free conditions, and derive an estimated value of the test statistic for each experimental result, which is recorded as , and then from S samples are randomly selected with replacement, denoted as , estimate the test statistic using the mean: ; S42. Predicting the thickness of ice on plate-like structures using dispersion curve migration.

[0039] Step S42 is specifically as follows: record the dispersion curve as , ice thickness is minimized by Theoretical wave number solution of the guided wave model of double-layer ice-plate structure The distance between them is calculated. The calculation formula is as follows: .

[0040] Figure 9 A schematic diagram of ice thickness estimation when the ice thickness is 0 mm is shown; Figure 10 A schematic diagram of ice thickness estimation when the ice thickness is 1 mm is shown; Figure 11 The figure shows a schematic diagram of ice thickness estimation when the ice thickness is 3 mm. It can be seen from the figure that the ice thickness measured by this method has a small error and high measurement accuracy.

[0041] The present application is described in detail below with reference to a specific embodiment. This embodiment detects icing on an aircraft skin, which includes the following steps: S1. Complete sensor placement according to the sensor placement strategy. This embodiment completes sensor placement according to the above steps. This placement ensures accurate and robust estimation of the dispersion curve within the ambient noise excitation frequency range.

[0042] S2. Acquire real-time passive guided wave signals of the plate-like structure. In this embodiment, on-site vibration measurement is performed on the aircraft skin to acquire real-time passive guided wave signals of the plate-like structure.

[0043] S3. Obtain the complex wave number trajectory of the plate structure based on the passive waveguide signal. The complex wave number trajectory includes the attenuation coefficient curve and the dispersion curve. The complex wave number is calculated using the following formula: ; in, is the objective function matrix, is the frequency, is the measured structural impulse response function of the sensor pair after tail wave elimination processing, is the theoretical structural impulse response function of the sensor pair.

[0044] S4, based on the obtained complex wave number trajectory, ice warning is carried out and the thickness of the ice layer on the plate structure is predicted, which includes the following sub-steps: S41, ice warning is carried out through the attenuation coefficient curve offset and Bootstrap-based hypothesis testing; the ice warning is carried out through the attenuation coefficient curve offset, and in the embodiment, the relevant conditions meet the ice warning, and step S42 is entered.

[0045] S42, on the basis of the ice warning, the thickness of the ice layer on the plate structure is predicted by using the dispersion curve offset, and in the embodiment, the ice thickness e is 2.5 mm obtained by the ice thickness calculation formula.

[0046] The above-described embodiments only describe the preferred embodiments of the present application, and do not limit the scope of the present application. Without departing from the design spirit of the present application, various modifications and improvements to the technical solutions of the present application made by those skilled in the art shall fall within the protection scope determined by the claims of the present application.

Claims

1. A passive icing detection method for plate structures based on ambient noise, characterized by: It includes the following steps: S1. Complete sensor placement according to the sensor placement strategy, which specifically includes the following sub-steps: S11, setting the minimum distance d between adjacent sensors and the distance D between the first and last sensors; S12. Given the number of sensors M, there are a total of For sensors, and The maximum number of sensors indivual value, For the The sensor and The distance between the sensors, For the sensor coordinates, set and ,exist Iteratively add sensors on the right , m = 3,···, M, where is an integer multiple of d, ; S2, obtaining real-time passive guided wave signals of the plate structure; S3. Obtain the complex wave number trajectory of the plate structure based on the passive waveguide signal. The complex wave number trajectory includes the attenuation coefficient curve and the dispersion curve. The complex wave number is calculated using the following formula: ； in, is the complex wave number The objective function matrix, Frequency is the structural impulse response function measured by the sensor and processed by coda wave elimination, is the theoretical structural impulse response function of the sensor pair, is the impulse response vector of the measurement structure constructed by different sensor pairs, Theoretical structural impulse response vectors constructed for different sensor pairs; S4. Based on the complex wave number trajectory, an ice early warning is performed and the thickness of the ice layer on the plate structure is predicted, which includes the following sub-steps: S41, performing icing warning by attenuation coefficient curve offset and Bootstrap-based hypothesis test, and then proceeding to step S42 after the icing warning; S42. Predicting the thickness of ice on plate-like structures using dispersion curve migration.

2. The method for passive ice detection of plate-like structures based on ambient noise according to claim 1, characterized in that: Step S3 specifically includes the following sub-steps: S31, calculating the cross-correlation function between all sensor pairs; S32, performing a whitening operation on the cross-correlation function; S33, performing tail wave elimination on the whitened time domain structure impulse response; S34. Calculate the complex wave number using the inverse wave correlation method formula and obtain the complex wave number trajectory.

3. The method for passive ice detection of plate-like structures based on ambient noise according to claim 2, characterized in that: The cross-correlation function is calculated in step S31 as follows: Right in position and The passive waveguide signal measured at the position is cross-correlated to obtain the cross-correlation signal, which is processed using the following formula: ； in, represents the cross-correlation signal, Represents a preset time interval, and the cross-correlation signal is obtained by stepping a discrete time signal and repeatedly calculating the correlation between the two signals.

4. The method for passive ice detection of plate-like structures based on ambient noise according to claim 2, characterized in that: Step S33 eliminates the coda wave specifically by applying a window function to the time domain impulse response function estimation between the two sensors after whitening in step S32 to eliminate the coda wave component: ； in, is a step function window: ； in, is the travel time of the direct wave of the sensor pair; After eliminating the coda wave, the frequency domain structure impulse response function is obtained through Fourier transform: 。 5. The method for passive ice detection of plate-like structures based on ambient noise according to claim 4, characterized in that: It is determined by identifying the main peak of the whitened cross-correlation, which corresponds to the direct wave travel time of the sensor pair.

6. The method for passive ice detection of plate-like structures based on ambient noise according to claim 2, characterized in that: Step S34 specifically includes the following sub-steps: S341. The expression of the theoretical structural impulse response function of the sensor pair in the plate structure is: ； in, , is the complex wave number to be calculated, , is the mass of the plate per unit area, is the Hankel function of the first kind, is the imaginary unit, is the real part, is the imaginary part; S342. According to the above equation, construct two vectors for all different sensor pairs: ； ； S343, given a frequency , complex wave number It is obtained by calculating the normalized inner product of the two vectors mentioned above; S344, repeat step S343, given multiple frequencies , and the complex wave number is obtained The objective function matrix is ​​obtained to obtain the complex wave number trajectory.

7. The method for passive ice detection of plate-like structures based on ambient noise according to claim 2, characterized in that: The method for drawing the dispersion curve or attenuation coefficient curve in step S34 is: draw the objective function matrix as about or about The color map can be used to visually identify the dispersion curve or attenuation coefficient curve, or by selecting each frequency The maximum value of the objective function corresponds to or Dispersion curve or attenuation coefficient curve.

8. The method for passive ice detection of plate-like structures based on ambient noise according to claim 1, characterized in that: In step S41, an icing warning is performed by using a Bootstrap hypothesis test based on the attenuation coefficient curve offset. The specific steps are as follows: S411. Assume that the plate structure is ice-free, i.e. , and set the alternative hypothesis as ; S412: Given an ambient noise measurement, the attenuation coefficient at different frequencies is defined as the imaginary part of the complex wave number obtained in step S3. , defining the test statistic of the test as the average attenuation coefficient over a given frequency range: ； in, is the test statistic, is the number of selected frequencies; S413. Calculate the average test statistic: ； in, , S is the number of passive experiments, the superscript * indicates that the data or test statistic comes from Bootstrap resampling, the resampling process is repeated B times, and B Bootstrap samples of the test statistic under the null hypothesis H0 are obtained. ,Will Sort in ascending order, i.e. ; S414. For any β∈(0,1), the 100β percentile Bootstrap test statistic for: ； Where [·] indicates rounding to the nearest integer. S415. Implement hypothesis testing for icing warning. Given any new data of unknown icing conditions, derive the test statistic α. For any significance level β∈(0,1), if β satisfies the following formula, then accept the null hypothesis H0: ； Otherwise, reject H0 and accept the alternative hypothesis H1, triggering an ice alarm.

9. The method for passive ice detection of plate-like structures based on ambient noise according to claim 1, characterized in that: The calculation method of the average test statistic in step S413 is as follows: derive the probability distribution of the test statistic α under the null hypothesis H0, assume that the passive experiment is repeated S times under ice-free conditions, and derive a test statistic for each experimental result, which is recorded as , and then from S samples are randomly selected with replacement, denoted as , use the mean to calculate the average test statistic: 。 10. The method for passive ice detection of plate-like structures based on ambient noise according to claim 1, characterized in that: Step S42 is specifically as follows: record the dispersion curve as , ice thickness is minimized by Theoretical wave number solution of the guided wave model of double-layer ice-plate structure Calculate the distance between them.

Images