A guided wave based method and system for monitoring damage to a wind turbine blade

Abstract

The application discloses a kind of fan blade damage monitoring method and system based on guided wave, it is related to fan blade monitoring technical field, including introducing ultrasonic signal into fan blade, and receiving reflected ultrasonic signal;Signal pretreatment is carried out to ultrasonic signal, after processing, ultrasonic signal is screened, and initial signal is obtained;All kinds of features in initial signal are extracted, and the correlation of each feature with time domain, frequency domain, wavelet transform domain is calculated;All kinds of features in initial signal are divided into time domain set, frequency domain set, wavelet transform domain set;According to time domain set, frequency domain set, wavelet transform domain set and preset neural network model determine fan blade damage.Combining time domain, frequency domain, wavelet transform domain to monitor blade damage, improve monitoring adaptability and precision.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of fan blade monitoring, and more particularly to a fan blade damage monitoring method and system based on guided waves. BACKGROUND

[0002] Fan blade damage monitoring based on guided waves is an advanced technology that uses the characteristics of sound and ultrasonic wave propagation inside the fan blade to detect damage to the blade. This technology combines knowledge from multiple fields such as acoustics, signal processing, sensor technology, and machine learning. By capturing acoustic information inside the blade with sensors, combined with signal processing and analysis, information about the health of the blade can be extracted. This technology requires knowledge of sound wave propagation, signal processing, sensor selection and deployment, and data analysis. Through modeling, experimental verification, and machine learning, accurate detection and real-time monitoring of blade damage can be achieved.

[0003] In the prior art, only one aspect of the feature is used for judgment and monitoring, and multiple aspects of the feature cannot be combined for monitoring, resulting in poor monitoring adaptability and low accuracy.

[0004] Therefore, how to improve the monitoring adaptability and accuracy is a technical problem to be solved at present. SUMMARY

[0005] The present application provides a fan blade damage monitoring method based on guided waves to solve the technical problem of poor monitoring adaptability and low accuracy in the prior art. The method comprises:

[0006] Introducing an ultrasonic wave signal into the fan blade and receiving the reflected ultrasonic wave signal;

[0007] Signal preprocessing is performed on the ultrasonic wave signal, and the filtered ultrasonic wave signal is obtained as an initial signal;

[0008] Extract all types of features in the initial signal, and calculate the correlation of each feature with the time domain, frequency domain, and wavelet transform domain, respectively;

[0009] Divide all types of features in the initial signal into time domain sets, frequency domain sets, and wavelet transform domain sets;

[0010] Determine the fan blade damage according to the time domain sets, frequency domain sets, wavelet transform domain sets, and a preset neural network model.

[0011] In some embodiments of the present application, signal preprocessing is performed on the ultrasonic wave signal, including:

[0012] Preprocessing includes filtering, denoising, and gain;

[0013] The gain includes:

[0014] obtaining an amplitude range of the ultrasonic signal, determining a product coefficient according to a size relationship and a difference value between the amplitude range of the ultrasonic signal and a target amplitude range;

[0015] amplifying or reducing the ultrasonic signal according to the product coefficient to obtain a signal after preliminary gain;

[0016] performing nonlinear adjustment according to a change of the signal after preliminary gain, so that different parts of the signal can obtain different degrees of gain or compression.

[0017] In some embodiments of the present application, the screened ultrasonic signal after processing is obtained as an initial signal, including:

[0018] obtaining all attributes of the ultrasonic signal, screening an attribute capable of representing fluctuation from all attributes, denoted as a fluctuation attribute;

[0019] calculating a fluctuation amount of each fluctuation attribute, and obtaining a total fluctuation amount;

[0020] eliminating the ultrasonic signal corresponding to the total fluctuation amount exceeding a fluctuation threshold;

[0021] performing gap filling according to adjacent signal segments of the eliminated ultrasonic signal, so as to obtain a complete ultrasonic signal, denoted as an initial signal.

[0022] In some embodiments of the present application, all kinds of features in the initial signal are extracted, and the correlation of each feature with the time domain, the frequency domain and the wavelet transform domain is calculated, including:

[0023] calculating the correlation degree of each feature with the time domain index, the frequency domain index and the wavelet transform domain index;

[0024] the features with the correlation degree exceeding the corresponding first threshold are denoted as correlation features.

[0025] In some embodiments of the present application, all kinds of features in the initial signal are divided into a time domain set, a frequency domain set and a wavelet transform domain set, including:

[0026] the correlation features with the correlation degree exceeding the corresponding second threshold are divided into the corresponding time domain set, the frequency domain set and the wavelet transform domain set;

[0027] the correlation features with the correlation degree exceeding the corresponding first threshold and not exceeding the second threshold are denoted as intermediate features;

[0028] if the correlation degree of the intermediate feature is greater than an average threshold, the intermediate feature is divided into the corresponding time domain set, the frequency domain set and the wavelet transform domain set;

[0029] otherwise, no division is performed;

[0030] The average threshold is an average of the first threshold and the second threshold.

[0031] In some embodiments of the present application, the fan blade damage is determined according to the time domain set, the frequency domain set, the wavelet transform domain set, and a preset neural network model, including:

[0032] The first fan blade damage is determined according to the time domain set and a first neural network model;

[0033] The second fan blade damage is determined according to the frequency domain set and a second neural network model;

[0034] The third fan blade damage is determined according to the wavelet transform domain set and a third neural network model;

[0035] The fourth fan blade damage is determined according to the time domain set, the frequency domain set, the wavelet transform domain set, and a fourth neural network model;

[0036] The respective influence weights corresponding to the time domain set, the frequency domain set, and the wavelet transform domain set are determined based on the time domain set, the frequency domain set, and the wavelet transform domain set, respectively;

[0037] The fan blade damage is determined according to the influence weights, the first fan blade damage, the second fan blade damage, the third fan blade damage, and the fourth fan blade damage.

[0038] In some embodiments of the present application, the fan blade damage is determined according to the influence weights, the first fan blade damage, the second fan blade damage, the third fan blade damage, and the fourth fan blade damage, including:

[0039] If the influence weights of the time domain set, the frequency domain set, and the wavelet transform domain set are all within the corresponding weight intervals, the fourth fan blade damage is taken as the fan blade damage;

[0040] Otherwise, the weight excess amounts of the influence weights of the time domain set, the frequency domain set, and the wavelet transform domain set exceeding the corresponding weight intervals are calculated;

[0041] The fan blade damage is determined according to the blade damage corresponding to the maximum weight excess amount and the fourth fan blade damage.

[0042] Correspondingly, the present application also provides a fan blade damage monitoring system based on guided waves, which comprises:

[0043] A first module is configured to introduce an ultrasonic signal into a fan blade and receive a reflected ultrasonic signal;

[0044] A second module is configured to perform signal preprocessing on the ultrasonic signal, screen the ultrasonic signal after the preprocessing, and obtain an initial signal;

[0045] The third module is configured to extract all kinds of features in the initial signal and calculate the correlation of each feature with the time domain, the frequency domain and the wavelet transform domain respectively.

[0046] The fourth module is configured to divide all kinds of features in the initial signal into the time domain set, the frequency domain set and the wavelet transform domain set.

[0047] The fifth module is configured to determine the fan blade damage according to the time domain set, the frequency domain set, the wavelet transform domain set and the preset neural network model.

[0048] By applying the above technical solutions, the ultrasonic signal is introduced into the fan blade, and the reflected ultrasonic signal is received; the ultrasonic signal is preprocessed, and the ultrasonic signal is screened after processing to obtain an initial signal; all kinds of features in the initial signal are extracted, and the correlation of each feature with the time domain, the frequency domain and the wavelet transform domain is calculated; all kinds of features in the initial signal are divided into the time domain set, the frequency domain set and the wavelet transform domain set; and the fan blade damage is determined according to the time domain set, the frequency domain set, the wavelet transform domain set and the preset neural network model. The time domain, the frequency domain and the wavelet transform domain are combined to monitor the blade damage, and the monitoring adaptability and accuracy are improved. BRIEF DESCRIPTION OF DRAWINGS

[0049] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0050] Figure 1 A flowchart of a fan blade damage monitoring method based on guided waves according to an embodiment of the present application is shown;

[0051] Figure 2 A structural diagram of a fan blade damage monitoring system based on guided waves according to an embodiment of the present application is shown. DETAILED DESCRIPTION

[0052] The technical solutions in the embodiments of the present application will be described clearly and completely in combination with the drawings in the embodiments of the present application. Obviously, the described embodiments are only some embodiments of the present application, not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0053] An embodiment of the present application provides a fan blade damage monitoring method based on guided waves, as shown in Figure 1 The method comprises the following steps:

[0054] Step S101, introduce ultrasonic signals into the fan blade, and receive the reflected ultrasonic signals.

[0055] In this embodiment, the guided wave-based fan blade damage monitoring is a method that uses ultrasonic guided wave technology to detect the structural health of fan blades. This technology uses the blade itself as a medium for guided wave propagation, and analyzes the signal changes during propagation to determine whether the blade has damage, fatigue or other structural problems.

[0056] Step S102, signal preprocessing of ultrasonic signals, after processing, screening ultrasonic signals, getting initial signals.

[0057] In this embodiment, the signal preprocessing step may include the following content:

[0058] Sampling and data acquisition: Obtain the original signal from the sensor, usually ultrasonic signal, for subsequent processing and analysis.

[0059] Filtering: Use filters to filter signals to remove high-frequency noise and interference while retaining useful signal components. Common filtering methods include low-pass, high-pass, band-pass and band-stop filtering.

[0060] De-noising: Improve signal quality by removing unnecessary noise in the signal. De-noising methods can include median filtering, wavelet de-noising, etc.

[0061] Gain adjustment: Depending on the changes in signal strength, gain adjustment may be needed to ensure that the signal range is within the appropriate scale.

[0062] Time-to-frequency domain conversion: Fourier transform and other methods can be used to convert signals from time domain to frequency domain to analyze the frequency components of the signal.

[0063] Envelope extraction: Extract the envelope from the signal to capture the transient characteristics and changes of the signal.

[0064] Sampling rate adjustment: In some cases, it may be necessary to adjust the sampling rate of the signal to meet the needs of subsequent analysis.

[0065] Phase correction: If signal acquisition is performed between multiple sensors, phase correction may be needed to ensure time synchronization.

[0066] Data alignment: If signals collected by multiple sensors need to be compared and analyzed, data alignment may be needed to ensure consistency in time.

[0067] Outlier processing: Due to uncertainties in the acquisition process, outliers may occur. Outlier processing is needed to avoid interference with subsequent analysis.

[0068] Data calibration: If the signal needs to be associated with actual physical quantities, data calibration may be required to map signal values to specific physical quantities.

[0069] Different signal preprocessing steps may be required for different applications and situations. The goal of preprocessing is to improve signal quality, reduce noise and interference, and enable more accurate analysis and diagnosis.

[0070] In some embodiments of the present application, signal preprocessing is performed on the ultrasonic signal, including:

[0071] Preprocessing includes filtering, denoising and gain;

[0072] Gain includes:

[0073] Obtain the amplitude range of the ultrasonic signal, and determine the multiplication coefficient according to the size relationship and difference value between the amplitude range of the ultrasonic signal and the target amplitude range;

[0074] According to the multiplication coefficient, the ultrasonic signal is amplified or reduced to obtain a signal after preliminary gain;

[0075] According to the change of the signal after preliminary gain, nonlinear adjustment is performed to make different parts of the signal obtain different degrees of gain or compression.

[0076] In this embodiment, the following are some steps to adjust the gain of the signal:

[0077] Observe the signal range: First, observe the amplitude range of the original signal. If the amplitude of the signal changes greatly, it may cause some signals to be too strong or too weak, making it difficult to analyze the details.

[0078] Select gain adjustment strategy: According to the characteristics and needs of the signal, select appropriate gain adjustment strategy. Gain can be adjusted linearly or nonlinearly.

[0079] Linear gain adjustment: Linear gain adjustment is to amplify or reduce the signal by multiplying a fixed coefficient. If the amplitude of the signal changes greatly, you can try to amplify the signal by a suitable multiple.

[0080] Dynamic gain adjustment: Dynamic gain adjustment is to adjust the gain according to the local characteristics of the signal. For example, in the frequency domain, different gain adjustments can be made to signals in different frequency ranges.

[0081] Nonlinear adjustment will adjust according to the characteristics and changes of the signal. This allows different parts of the signal to obtain different degrees of gain or compression, so that the details of the signal can be better displayed in different areas.

[0082] Step S103, extract all kinds of features in the initial signal, and calculate the correlation of each feature with the time domain, frequency domain, and wavelet transform domain respectively.

[0083] In this embodiment, generally, the features are as follows:

[0084] Time domain features:

[0085] Mean: the average value of the signal.

[0086] Variance: the degree of dispersion of signal values.

[0087] Peak: the maximum value of the signal.

[0088] Peak-to-peak value: the difference between the maximum and minimum values of the signal.

[0089] Pulse count: the number of times the signal crosses a certain threshold, used to detect mutations.

[0090] Autocorrelation coefficient: the correlation between the signal and its lagged version.

[0091] Frequency domain features:

[0092] Peak frequency: the frequency with the highest energy in the signal spectrum.

[0093] Spectrum width: the width of the frequency spectrum, reflecting the range of frequency distribution of the signal.

[0094] Energy distribution: the energy distribution in different frequency ranges.

[0095] Spectrum kurtosis: an index reflecting the shape of the spectrum, which can be used to identify different vibration modes.

[0096] Wavelet transform:

[0097] Wavelet coefficients: coefficients generated by wavelet transform, which can reflect the energy distribution of the signal in different frequency ranges.

[0098] Wavelet energy: the energy of the signal at each scale in the wavelet transform.

[0099] In some embodiments of the present application, the processed ultrasonic signals are screened to obtain the initial signals, including:

[0100] Obtain all attributes of the ultrasonic signals, and select the attributes that can represent fluctuations from all attributes, denoted as fluctuation attributes.

[0101] Calculate the fluctuation amount of each fluctuation attribute and obtain the total fluctuation amount.

[0102] Remove the ultrasonic signals whose total fluctuation amount exceeds the fluctuation threshold.

[0103] According to the adjacent signal segments of the removed ultrasonic signals, gaps are filled to obtain a complete ultrasonic signal, denoted as an initial signal.

[0104] In step S104, all kinds of features in the initial signal are classified into a time domain set, a frequency domain set and a wavelet transform domain set.

[0105] In some embodiments of the present application, all kinds of features in the initial signal are extracted, and the correlation of each feature with the time domain, the frequency domain and the wavelet transform domain is calculated, including:

[0106] The correlation of each feature with the time domain index, the frequency domain index and the wavelet transform domain index is calculated.

[0107] Features with a correlation greater than a corresponding first threshold value are denoted as associated features.

[0108] In some embodiments of the present application, all kinds of features in the initial signal are classified into a time domain set, a frequency domain set and a wavelet transform domain set, including:

[0109] Associated features with a correlation greater than a corresponding second threshold value are classified into a corresponding time domain set, a frequency domain set and a wavelet transform domain set.

[0110] Associated features with a correlation greater than a corresponding first threshold value and less than a second threshold value are denoted as intermediate features.

[0111] If the correlation of an intermediate feature is greater than an average threshold value, the intermediate feature is classified into a corresponding time domain set, a frequency domain set and a wavelet transform domain set.

[0112] Otherwise, no classification is performed.

[0113] The average threshold value is the average of the first threshold value and the second threshold value.

[0114] In step S105, the wind turbine blade damage is determined according to the time domain set, the frequency domain set, the wavelet transform domain set and a preset neural network model.

[0115] In some embodiments of the present application, the wind turbine blade damage is determined according to the time domain set, the frequency domain set, the wavelet transform domain set and a preset neural network model, including:

[0116] A first wind turbine blade damage is determined according to the time domain set and a first neural network model.

[0117] A second wind turbine blade damage is determined according to the frequency domain set and a second neural network model.

[0118] A third wind turbine blade damage is determined according to the wavelet transform domain set and a third neural network model.

[0119] The fourth fan blade damage is determined according to the time domain set, the frequency domain set, the wavelet transform domain set and the fourth neural network model;

[0120] The respective influence weights are determined according to the time domain set, the frequency domain set and the wavelet transform domain set respectively;

[0121] The fan blade damage is determined according to the influence weight, the first fan blade damage, the second fan blade damage, the third fan blade damage and the fourth fan blade damage.

[0122] In the embodiment, the first neural network model is a model trained according to the time domain set, and the others are the same.

[0123] In some embodiments of the application, the fan blade damage is determined according to the influence weight, the first fan blade damage, the second fan blade damage, the third fan blade damage and the fourth fan blade damage, including:

[0124] If the influence weights of the time domain set, the frequency domain set and the wavelet transform domain set are all within the corresponding weight intervals, the fourth fan blade damage is taken as the fan blade damage;

[0125] Otherwise, the weight excess amounts of the influence weights of the time domain set, the frequency domain set and the wavelet transform domain set exceeding the corresponding weight intervals are calculated;

[0126] The fan blade damage is determined according to the blade damage corresponding to the maximum weight excess amount and the fourth fan blade damage.

[0127] In the embodiment, the fourth neural network model is a model trained according to the time domain set, the frequency domain set and the wavelet transform domain set, but the accuracy is higher when the weights are within the normal intervals. Otherwise, the fourth fan blade damage is adjusted.

[0128] In the embodiment, the fan blade damage is determined according to the blade damage corresponding to the maximum weight excess amount and the fourth fan blade damage, and an adjustment amount is determined according to the difference between the blade damage corresponding to the maximum weight excess amount and the fourth fan blade damage, fourth fan blade damage + adjustment amount = fan blade damage.

[0129] By applying the above technical solutions, the ultrasonic signal is introduced into the fan blade, and the reflected ultrasonic signal is received; the ultrasonic signal is pre-processed, the filtered ultrasonic signal is obtained after processing, and the initial signal is obtained; all kinds of features in the initial signal are extracted, and the correlation of each feature with the time domain, the frequency domain and the wavelet transform domain is calculated; all kinds of features in the initial signal are divided into the time domain set, the frequency domain set and the wavelet transform domain set; and the fan blade damage is determined according to the time domain set, the frequency domain set, the wavelet transform domain set and the preset neural network model. The time domain, the frequency domain and the wavelet transform domain are combined to monitor the blade damage, and the monitoring adaptability and accuracy are improved.

[0130] Through the description of the above embodiments, those skilled in the art can clearly understand that the present application can be implemented by hardware, or by means of software and necessary general hardware platform. Based on such understanding, the technical solutions of the present application can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD-ROM, a U disk, a mobile hard disk, etc.), and includes a plurality of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in various embodiments of the present application.

[0131] Correspondingly, the present application also provides a fan blade damage monitoring system based on guided waves, as shown in Figure 2 The system comprises:

[0132] The first module 201 is configured to introduce the ultrasonic signal into the fan blade, and receive the reflected ultrasonic signal.

[0133] The second module 202 is configured to pre-process the ultrasonic signal, filter the ultrasonic signal after processing, and obtain the initial signal.

[0134] The third module 203 is configured to extract all kinds of features in the initial signal, and calculate the correlation of each feature with the time domain, the frequency domain and the wavelet transform domain.

[0135] The fourth module 204 is configured to divide all kinds of features in the initial signal into the time domain set, the frequency domain set and the wavelet transform domain set.

[0136] The fifth module 205 is configured to determine the fan blade damage according to the time domain set, the frequency domain set, the wavelet transform domain set and the preset neural network model.

[0137] Finally, it should be noted that the above examples are only used to illustrate the technical solutions of the present application, and are not intended to limit the same; although the present application has been described in detail with reference to the foregoing examples, those of ordinary skill in the art will understand that they can still modify the technical solutions described in the foregoing examples, or make equivalent replacements for some of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A guided wave based wind turbine blade damage monitoring method, characterized by, The method comprises: Introducing an ultrasonic signal into a fan blade and receiving a reflected ultrasonic signal; Signal preprocessing is performed on the ultrasonic signal, and the preprocessed ultrasonic signal is screened to obtain an initial signal; All kinds of features in the initial signal are extracted, and the correlation of each feature with the time domain, the frequency domain and the wavelet transform domain is calculated; All kinds of features in the initial signal are divided into a time domain set, a frequency domain set and a wavelet transform domain set; The fan blade damage is determined according to the time domain set, the frequency domain set, the wavelet transform domain set and a preset neural network model; The determination of the fan blade damage according to the time domain set, the frequency domain set, the wavelet transform domain set and the preset neural network model comprises: A first fan blade damage is determined according to the time domain set and a first neural network model; A second fan blade damage is determined according to the frequency domain set and a second neural network model; A third fan blade damage is determined according to the wavelet transform domain set and a third neural network model; A fourth fan blade damage is determined according to the time domain set, the frequency domain set, the wavelet transform domain set and a fourth neural network model; An influence weight corresponding to each of the time domain set, the frequency domain set and the wavelet transform domain set is determined based on the determination of each of the time domain set, the frequency domain set and the wavelet transform domain set; The fan blade damage is determined according to the influence weight, the first fan blade damage, the second fan blade damage, the third fan blade damage and the fourth fan blade damage; The determination of the fan blade damage according to the influence weight, the first fan blade damage, the second fan blade damage, the third fan blade damage and the fourth fan blade damage comprises: If the influence weights of the time domain set, the frequency domain set and the wavelet transform domain set are all within the corresponding weight intervals, the fourth fan blade damage is taken as the fan blade damage; Otherwise, the weight exceeding amount of the influence weights of the time domain set, the frequency domain set and the wavelet transform domain set exceeding the corresponding weight intervals is calculated; The fan blade damage is determined according to the blade damage corresponding to the maximum weight exceeding amount and the fourth fan blade damage.

2. The guided-wave-based fan blade damage monitoring method of claim 1, wherein, The signal preprocessing of the ultrasonic signal comprises: The preprocessing includes filtering, denoising and gain; The gain includes: An amplitude range of the ultrasonic signal is obtained, and a product coefficient is determined according to the size relationship and the difference between the amplitude range of the ultrasonic signal and a target amplitude range; The ultrasonic signal is amplified or reduced according to the product coefficient to obtain a signal after preliminary gain; The signal after preliminary gain is nonlinearly adjusted according to the change of the signal after preliminary gain, so that different parts of the signal can be obtained with different degrees of gain or compression.

3. The guided-wave-based fan blade damage monitoring method of claim 2, wherein, The screening of the ultrasonic signal after preprocessing to obtain the initial signal comprises: All attributes of the ultrasonic signal are obtained, and attributes capable of representing fluctuations are screened from all the attributes and are denoted as fluctuation attributes; The fluctuation amount of each fluctuation attribute is calculated, and a total fluctuation amount is obtained; The ultrasonic signal corresponding to the total fluctuation amount exceeding a fluctuation threshold is removed; The removed ultrasonic signal is supplemented with adjacent signal segments to obtain a complete ultrasonic signal, which is denoted as the initial signal.

4. The guided-wave based fan blade damage monitoring method of claim 1, wherein, The extraction of all kinds of features in the initial signal and the calculation of the correlation of each feature with the time domain, the frequency domain and the wavelet transform domain comprise: The correlation of each feature with a time domain index, a frequency domain index and a wavelet transform domain index is calculated. The features with the correlation exceeding the corresponding first threshold value are recorded as associated features.

5. The guided-wave based fan blade damage monitoring method of claim 4, wherein, All kinds of features in the initial signal are divided into time domain sets, frequency domain sets and wavelet transform domain sets, including: The associated features with the correlation exceeding the corresponding second threshold value are divided into the corresponding time domain sets, frequency domain sets and wavelet transform domain sets; The associated features with the correlation exceeding the corresponding first threshold value and not exceeding the second threshold value are recorded as intermediate features; If the correlation of the intermediate features is greater than the average threshold value, the intermediate features are divided into the corresponding time domain sets, frequency domain sets and wavelet transform domain sets; Otherwise, no division is performed. The average threshold value is the average of the first threshold value and the second threshold value.

6. A guided wave based wind turbine blade damage monitoring system for use in the guided wave based wind turbine blade damage monitoring method of any of claims 1-5, wherein, The system comprises: A first module for introducing ultrasonic signals into a fan blade and receiving reflected ultrasonic signals; A second module for signal preprocessing of the ultrasonic signals, screening the ultrasonic signals after processing, and obtaining initial signals; A third module for extracting all kinds of features in the initial signals and calculating the correlation of each feature with time domain, frequency domain and wavelet transform domain; A fourth module for dividing all kinds of features in the initial signals into time domain sets, frequency domain sets and wavelet transform domain sets; A fifth module for determining fan blade damage according to the time domain sets, frequency domain sets, wavelet transform domain sets and a preset neural network model.

Images