Terahertz Non-destructive Testing Method and System for Delamination Defects in Wind Turbine Blades

Abstract

This invention discloses a terahertz non-destructive testing method and system for delamination defects in wind turbine blades, belonging to the field of non-destructive intelligent testing technology. The method involves acquiring a three-dimensional geometric model of the wind turbine blade, extracting thickness gradient data from the model, determining the internal material density distribution based on the thickness gradient data, determining scanning parameters based on the internal material density distribution, transmitting a terahertz detection signal into the blade based on the scanning parameters, and acquiring the return signal. A delamination analysis model is constructed, and time-domain analysis is performed on the return signal to determine the signal attenuation location and extract the background noise level. Based on the background noise level, delamination defect features are extracted from the signal attenuation location. The three-dimensional spatial location of the defect is determined based on the delamination defect features, and a three-dimensional mapping map of the internal delamination defects of the blade is generated based on the three-dimensional spatial location of the defect. This invention enables high-reliability non-destructive testing of delamination defects across the entire thickness of wind turbine blades.

Description

Technical Field

[0001] This invention belongs to the field of non-destructive intelligent testing technology, specifically relating to a terahertz non-destructive testing method and system for delamination defects in wind turbine blades. Background Technology

[0002] As the core component for wind energy capture, the internal health of wind turbine blades directly affects the operational safety and lifespan of the entire machine. Therefore, accurate detection of internal defects in the blades is a key link in ensuring the stable development of the wind power industry. Terahertz non-destructive testing technology has shown great potential in this field due to its excellent penetration of non-metallic composite materials. However, when applying this technology to actual blade inspection, it faces a series of unique challenges brought about by the complex structure of the blades themselves.

[0003] Current terahertz detection methods are mostly designed for materials with uniform thickness or simple structure. They have significant limitations when dealing with large components such as wind turbine blades. The thickness of the blade is not constant, but exhibits a significant gradient change from the root to the tip and from the leading edge to the trailing edge. At the same time, the internal glass fiber or carbon fiber composite material layup structure also causes a non-uniform distribution of material density in the depth direction. This dual gradient characteristic of geometry and material properties makes it difficult for detection systems using fixed parameters to adapt. If a single focal length and fixed power are used for scanning, the signal may be too strong and distorted in thinner areas, while in thicker or denser areas, the signal may be excessively attenuated, making it impossible to detect deeper information, resulting in incomplete detection results or misjudgments.

[0004] The core technical challenges arising from this focus on the two interrelated attributes of the detection system: "adaptability" and "penetration depth resolution." Due to the continuous spatial variation of blade thickness and material density, the detection system must be able to dynamically adapt to different scanning depth conditions. However, insufficient penetration depth resolution directly leads to the inability to clearly distinguish whether signal attenuation originates from the material's normal gradient or internal defects. For example, in the thick-walled region at the blade root, the weak signal generated by a deeply buried layered defect is easily submerged by the material's strong absorption and scattering effects on terahertz waves. Conventional methods cannot effectively extract and identify it in this complex context.

[0005] Therefore, how to construct a detection method that can adaptively adjust to follow the blade shape and material gradient, and clearly distinguish defect signals from background noise at different depth levels, has become a key issue in realizing non-destructive testing of layered defects across the full thickness of wind turbine blades with high reliability. Summary of the Invention

[0006] The purpose of this invention is to overcome the problem of not being able to achieve full-thickness, high-reliability non-destructive testing of layered defects in wind turbine blades, and to propose a terahertz non-destructive testing method and system for layered defects in wind turbine blades.

[0007] To achieve the above objectives, the present invention adopts the following technical solution: In a first aspect, the present invention provides a terahertz nondestructive testing method for delamination defects in wind turbine blades, comprising the following steps: Obtain a three-dimensional geometric model of the wind turbine blade, extract thickness gradient data from the three-dimensional geometric model, and determine the internal material density distribution of the blade based on the thickness gradient data; The scanning parameters are determined based on the density distribution of the material inside the blade. Terahertz detection signals are then transmitted into the blade based on the scanning parameters, and the return signals of the detection signals are collected. A hierarchical analysis model is constructed to perform time-domain analysis on the returned signal, determine the signal attenuation location, and extract the background noise level; Based on the background noise level, layered defect features are extracted from the signal at the signal attenuation location. The three-dimensional spatial location of the defect is determined based on the characteristics of the layered defect, and a three-dimensional mapping map of the layered defect inside the blade is generated based on the three-dimensional spatial location of the defect.

[0008] Furthermore, the scanning parameters include the center frequency and the scanning step size; The scanning parameters are determined based on the density distribution of the material inside the blade, including: Based on the density values ​​of different regions of the blade, the center frequency of each region is adaptively selected within a preset frequency range. Based on the thickness gradient modulus of different regions of the blade, the scanning step distance of each region is adaptively determined within a preset step distance range.

[0009] Furthermore, after acquiring the return signal from the detection signal, the return signal is corrected, including: Extract the frequency domain features of the returned signal and detect signal distortion regions; signal distortion regions are areas where high-frequency components exceed a preset threshold. Analyze the impact of scanning parameters on the signal distortion area to determine the parameter adjustment range; Adjust the scanning parameters according to the parameter adjustment range, re-acquire the return signal in the signal distortion area, and obtain the corrected return signal.

[0010] Furthermore, the areas of signal distortion are detected, including: Perform a windowed Fourier transform on the returned signal to obtain the frequency domain amplitude spectrum; Calculate the ratio of high-frequency energy to low-frequency energy in the frequency domain amplitude spectrum; When the ratio exceeds the preset distortion detection threshold, it is determined that there is signal distortion at the scan point.

[0011] Furthermore, the influence of scanning parameters on the signal distortion region is analyzed, and the parameter adjustment range is determined, including: The overdrive factor is obtained by calculating the ratio of the peak amplitude of the returned signal in the time domain in the signal distortion region to the maximum amplitude threshold of the receiver's linear operating region. The upper limit for adjusting the transmit power is determined based on the overdrive factor; Adjust the scanning parameters according to the parameter adjustment range, including: The adjusted transmit power is calculated based on the overdrive factor and the desired peak signal amplitude. The returned signal in the signal distortion area was reacquired using the adjusted transmit power.

[0012] Furthermore, a hierarchical analysis model is constructed to perform time-domain analysis on the returned signal, including: Identify abnormal fluctuation segments in the returned signal to obtain potential over-attenuation regions; The signal in the potential over-attenuation region is decomposed, the local change trend is extracted, and the specific over-attenuation segment is identified. Background noise components are separated from specific over-attenuated segments to obtain background noise distribution characteristics; The signal of the excessively attenuated segment is repaired based on the background noise distribution characteristics to obtain optimized signal data.

[0013] Furthermore, based on the background noise level, layered defect features are extracted from the signal at the signal attenuation location, including: Compare the signal amplitude at the signal attenuation location with the background noise level; Signal features whose amplitude exceeds a preset multiple of the background noise level are extracted as layered defect features.

[0014] Secondly, the present invention provides a terahertz non-destructive testing system for delamination defects in wind turbine blades, comprising: The density distribution determination module is used to acquire a three-dimensional geometric model of the wind turbine blade, extract thickness gradient data from the three-dimensional geometric model, and determine the internal material density distribution of the blade based on the thickness gradient data. The return signal acquisition module is used to determine the scanning parameters based on the density distribution of the material inside the blade, transmit terahertz detection signals into the blade based on the scanning parameters, and acquire the return signals of the detection signals. The background noise extraction module is used to build a hierarchical analysis model, perform time-domain analysis on the returned signal, determine the signal attenuation location, and extract the background noise level. The layered defect extraction module is used to extract layered defect features from the signal at signal attenuation locations based on the background noise level. The blade defect mapping module is used to determine the three-dimensional spatial location of defects based on the characteristics of layered defects, and to generate a three-dimensional mapping map of the layered defects inside the blade based on the three-dimensional spatial location of the defects.

[0015] Thirdly, the present invention provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement a terahertz non-destructive testing method for delamination defects in wind turbine blades.

[0016] Fourthly, the present invention provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements a terahertz non-destructive testing method for delamination defects in wind turbine blades.

[0017] Compared with the prior art, the present invention has the following beneficial technical effects: This invention proposes a terahertz non-destructive testing method for layered defects in wind turbine blades. Based on the three-dimensional geometric model and thickness gradient of the wind turbine blade, it pre-calculates the non-uniform distribution of material density and dynamically generates optimal scanning parameters (such as center frequency, scanning step size, and transmission power) for different depth regions. This enables the terahertz detection system to adapt to changes in thickness and density gradients from the root to the tip and from the leading edge to the trailing edge of the blade, effectively solving the core problems of fixed-parameter detection methods, such as signal oversaturation in thin-walled regions and insufficient signal penetration in thick-walled or high-density regions. It achieves full-thickness, consistent, and high signal-to-noise ratio detection of complex gradient components. By constructing a layered analysis model and performing refined time-domain analysis on the returned signal sequence, it can accurately separate normal signal attenuation caused by material gradients from abnormal signal attenuation caused by layered defects. Combined with the extraction and evaluation of background noise levels and targeted signal repair strategies, it significantly improves the ability to identify deep and weak defect signals under strong background noise and complex material responses, reducing the false negative and false positive rates. This invention introduces a real-time detection mechanism for signal distortion and an adaptive parameter adjustment mechanism. By analyzing the frequency domain to identify signal distortion areas and quantitatively analyzing the impact of parameters, key parameters such as transmit power are iteratively optimized to ensure that signal quality remains within the optimal linear range. Simultaneously, advanced signal decomposition, noise separation, and deep learning-assisted repair techniques are employed in the layered analysis stage to effectively overcome signal quality degradation caused by environmental interference and equipment nonlinearity, ensuring the reliability of the original data. From 3D model import, adaptive parameter calculation, data acquisition, signal processing and analysis to the final defect mapping generation, this invention forms a complete and automated detection process. By generating an intuitive internal defect mapping map of the blade, the location, size, distribution density, and depth information of layered defects can be clearly displayed, greatly facilitating the interpretation and evaluation of detection results and subsequent maintenance decisions, thus improving the intelligence level of wind turbine blade operation and maintenance.

[0018] Furthermore, the method of this invention closely integrates with the actual manufacturing process and structural characteristics of wind turbine blades (such as variable thickness design and composite material layup), and the proposed model and algorithm both consider the feasibility and efficiency of engineering applications. Through adaptive planning of the scanning step and optimization of the signal processing algorithm, the method ensures detection accuracy while also meeting the efficiency requirements of large blade inspection, and has significant value for engineering promotion. Attached Figure Description

[0019] The accompanying drawings described herein are for illustrative purposes only and are not intended to limit the scope of the invention in any way. Furthermore, the shapes and proportions of the components in the drawings are merely schematic to aid in understanding the invention and do not specifically limit the shapes and proportions of the components. In the drawings: Figure 1 This is a flowchart of the terahertz non-destructive testing method for delamination defects in wind turbine blades according to the present invention.

[0020] Figure 2 This is a simplified structural diagram of the terahertz non-destructive testing system for delamination defects in wind turbine blades according to the present invention.

[0021] Figure 3 This is an electronic device diagram of the terahertz non-destructive testing method for delamination defects in wind turbine blades according to the present invention.

[0022] Figure 4 This is a schematic diagram of the terahertz non-destructive testing method for delamination defects in wind turbine blades according to an embodiment of the present invention. Detailed Implementation

[0023] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0024] Example 1 See Figure 1 A terahertz non-destructive testing method for delamination defects in wind turbine blades includes the following steps: A three-dimensional geometric model of a wind turbine blade is obtained, and thickness gradient data is extracted from the model to determine the internal material density distribution. Scanning parameters are determined based on this data, and a terahertz detection signal is emitted into the blade's interior according to these parameters. The returned signal is then collected. A layered analysis model is constructed, and the returned signal is analyzed in the time domain to determine the signal attenuation location and extract the background noise level. Based on the background noise level, layered defect features are extracted from the signal attenuation location. The three-dimensional spatial location of the defect is determined based on these features, and a three-dimensional mapping map of the layered defects inside the blade is generated.

[0025] This invention obtains a three-dimensional geometric model of a wind turbine blade, extracts thickness gradient data, and generates a variation distribution map based on the thickness gradient data. Based on the variation distribution map, the non-uniform density distribution of the internal composite material is calculated, leading to scanning parameters at different depths. Initial parameters for the blade region are obtained based on the scanning parameters. A terahertz wave transmitter is used to emit a detection signal into the blade's interior, and the returned signal sequence is collected. A layered analysis model is constructed, and time-domain analysis is performed on the returned signal sequence to determine attenuation locations and extract background noise levels. Layered defect characteristics are obtained based on the background noise levels, resulting in defect location distribution. An internal defect mapping map of the blade is generated based on the defect location distribution. This invention ensures detection accuracy while also meeting the efficiency requirements of large blade inspection, making it valuable for engineering application.

[0026] Specifically, such as Figure 4 As shown in the figure, this embodiment provides a terahertz non-destructive testing method for delamination defects in wind turbine blades.

[0027] S1. Obtain the three-dimensional geometric model of the wind turbine blade, extract the thickness gradient data from the three-dimensional geometric model, and obtain the variation distribution map based on the thickness gradient data.

[0028] As a large composite material structure, the accurate acquisition of the three-dimensional geometric model of wind turbine blades is fundamental for subsequent non-destructive testing. First, the blade is digitized in its entirety using a high-precision 3D laser scanner or structured light scanning system to acquire point cloud data. The point density must be below 0.5 mm to ensure detail capture. The point cloud data is then denoised, registered, and fused to reconstruct a solid model. This solid model not only includes the blade's external aerodynamic shape but also, through pre-embedded blade design data (such as CAD drawings) or sampling calibration using an ultrasonic thickness gauge, interpolates the internal layered structure, thus establishing a layered three-dimensional geometric model encompassing the skin, main beam, web, core material, and bonding areas.

[0029] In this embodiment, a point cloud dataset is set. ,in, NThis represents the total number of point clouds; a watertight triangular mesh surface is generated using Poisson surface reconstruction or the Delaunay triangulation algorithm. In the formula, V Represents a set of vertices; E Represents an edge set; F This represents a set of triangular facets.

[0030] Thickness gradient data were extracted using a slice analysis method, establishing a three-dimensional mesh coordinate system along the blade's spanwise (length direction), chordwise (width direction), and normal (thickness direction). At each mesh node, the theoretical thickness value was obtained by calculating the normal distance between the inner and outer surfaces of the model. Due to the variable thickness design of the composite blade, significant thickness variations occurred, particularly at the root, main beam cap region, leading edge, and trailing edge. The extracted thickness data formed a three-dimensional matrix. T ( x , y , z ),in x , y For planar coordinates, z The depth direction is represented by this matrix. Based on this matrix, the thickness change rate is calculated using a gradient operator. Gradient data can reveal regions of abrupt thickness changes, which are often defect-prone areas, such as missing layers, resin accumulation, or core material joints.

[0031] For any point q The thickness of the blade skin is T ( q Its thickness is along the normal direction at that point. n ( q Distance to the corresponding backplate inside the blade: In the formula, Indicates along - n ( q Direction and internal backplate curvature M inner The first intersection point. From this, the discretized thickness field can be obtained. T ( u , v ),in,( u , v () represents the coordinates of the blade's unfolding parameters.

[0032] The thickness gradient characterizes the intensity and direction of thickness changes; on a discrete mesh, the vertex... v i The gradient at point is expressed as: ; In the formula, A i Represents vertices v iThe area of ​​Voronoi; Represents vertices v i The set of 1-neighborhood vertices; e ij Indicates by v i point to v j The edge vectors.

[0033] Variation distribution plots are used to visualize gradient data, employing the thickness gradient magnitude: ,in, This represents the variation distribution map. A pseudo-color mapping technique is typically used to map the thickness gradient magnitude to color bands, generating a two-dimensional contour map or a three-dimensional rendering. High gradient regions (represented in red) correspond to areas of rapid thickness change, while low gradient regions (represented in blue) correspond to uniform areas. The variation distribution map not only guides inspection path planning but also, by comparing design tolerances, can preliminarily identify areas where manufacturing deviations exceed limits. To achieve high precision, it is necessary to combine the blade's ply sequence information and independently model the orientation and thickness of each ply (usually glass fiber / carbon fiber prepreg), thereby distinguishing between legitimate gradients caused by design flaws and abnormal gradients caused by defects in the variation distribution map.

[0034] S2. Calculate the non-uniform density distribution of the internal composite material based on the variation distribution map, and obtain the scanning parameters at different depths based on the non-uniform density distribution.

[0035] The density inhomogeneity of composite materials directly affects the propagation characteristics of terahertz waves. Based on the distribution map, regions with large thickness gradients often correspond to areas of density variation. (Density distribution) ρ ( x , y , z Correlate with thickness gradient through an empirical model: ρ = ρ 0+ k | T |, of which ρ 0 represents the density of the matrix resin. k The material-related coupling coefficient is determined through sampling experiments. For glass fiber reinforced epoxy resin composites, the fiber volume fraction... V f The change is the main cause of density non-uniformity, as explained by the mixing law. ρ = V f ρ f +(1 V f ) ρm Calculation, where ρ f and ρ m These represent fiber density and matrix density, respectively. The high gradient regions in the variation distribution map suggest... V f There may be localized areas with excessively high levels (such as fiber accumulation) or excessively low levels (such as increased porosity).

[0036] After the non-uniform density distribution calculation is completed, it is used to optimize the terahertz scanning parameters. The penetration depth and resolution of terahertz waves in materials are significantly affected by frequency. Center frequency f c The choice of frequency requires a trade-off: high frequencies (e.g., 2-3 THz) offer high lateral resolution (up to 50 μm) but weak penetration; low frequencies (0.1-1 THz) offer strong penetration (up to several centimeters) but lower resolution. Based on density distribution, an adaptive frequency selection algorithm is used: for high-density areas (e.g., main beam caps), a lower frequency (e.g., 0.3 THz) is selected to ensure penetration; for low-density areas (e.g., core material), a higher frequency (e.g., 1.5 THz) is selected to improve defect detection sensitivity. Scanning step Δ s Similarly, adjustments are made based on the density gradient, using a fine step size (e.g., 0.5 mm) in high-gradient regions and a coarse step size (e.g., 2 mm) in uniform regions to balance detection efficiency and coverage. Additionally, the transmit power... P t According to depth z Dynamic adjustments are made to local density, following the Beer-Lambert law. P r(z) = P t e α(ρ)z ,in, P r(z) The attenuation coefficient represents the received terahertz wave power at propagation depth z. α ( ρ The signal-to-noise ratio (SNR) of the echo signal is determined by a lookup table calibrated in the pre-experiment, ensuring that it is greater than a preset threshold (e.g., 20dB).

[0037] In this embodiment, the center frequency f c The selection methods include: ; In the formula, f low ~0.2THz; f high ~1.5THz; f mid~0.8THz; , , This represents the threshold value, determined experimentally.

[0038] Scan step Δ s The methods for determining this include: ; In the formula, λ eff Indicates the effective wavelength; K ( r ) represents a function relating a local region to the gradient magnitude, used to ensure denser sampling in regions where features change rapidly.

[0039] S3. Based on the scanning parameters, the initial parameters of the blade region are obtained. Based on the initial parameters, a terahertz wave transmitter is used to transmit a detection signal into the blade and collect the return signal sequence.

[0040] By scanning parameters and parameter sets, the initial parameter configuration of the blade region is obtained, completing the pre-detection preparation work and determining the baseline data for regional positioning. Based on the baseline data for regional positioning, the transmission direction and intensity of the terahertz wave transmitter are adjusted, and the detection signal is sent to the internal structure of the blade to obtain a preliminary signal sequence. For the preliminary signal sequence, signal processing tools are used for noise reduction to obtain a clear return signal for subsequent analysis. If the intensity of the return signal is lower than a preset threshold, the parameters of the transmitter are adjusted, and the detection signal is retransmitted to the internal structure of the blade to obtain a new signal sequence. Based on the new signal sequence and the internal structural characteristics of the blade region, key signal features are extracted to determine the structural distribution data inside the blade. Using the structural distribution data, the propagation path of the terahertz wave inside the blade is analyzed to determine the completeness of the signal acquisition and generate the final detection result. If there are incomplete signal acquisitions in the detection result, the scanning parameters are readjusted based on the regional positioning data to obtain supplementary return signals and improve the final result.

[0041] When acquiring the initial parameter configuration for the blade region, a pre-set set of scanning parameters, combined with the specific geometry and material properties of the blade, can be used to determine baseline data before detection. Assuming the thickness of the blade region ranges from 2.5 to 5.0 mm, the initial parameter configuration allocates different scanning intensities and frequencies based on the thickness difference to ensure signal penetration. This configuration provides a reliable starting point for subsequent detection.

[0042] When adjusting the emission direction and intensity of a terahertz wave transmitter, the emission angle can be adjusted to be perpendicular to the blade surface based on the reference data for regional positioning, thereby reducing signal reflection loss. Assuming the tilt angle of a certain area of ​​the blade is 15 degrees, the transmitter's direction is adjusted accordingly, while the intensity is increased to 1.2 times the reference value to ensure the signal can effectively penetrate the internal structure. This adjustment method helps to improve the coverage of the detection signal.

[0043] For denoising the initial signal sequence, this embodiment employs a time-domain based filtering method to separate background noise and spurious signals from the returned signal. Assuming the initial signal contains 50 Hz interference noise, signal processing tools can effectively filter out this interference, resulting in a clear returned signal. This processing approach provides a more accurate data foundation for subsequent analysis.

[0044] When extracting key signal features, the propagation time and amplitude variations of the signal are analyzed in conjunction with the characteristics of the blade's internal structure. For example, if a significant delay is observed in the signal at a certain layer within the blade's structure, it may indicate a change in thickness or material delamination. Extracting this feature can provide a preliminary assessment of the internal structural distribution. This analytical approach helps to gain a more accurate understanding of the blade's internal structure.

[0045] When analyzing the propagation path of terahertz waves inside a blade, the integrity of the acquisition is determined by the refraction and reflection characteristics of the signal. If the signal undergoes multiple reflections in a certain area, it indicates the possible existence of undetected internal gaps or defects; in this case, the area can be marked as a region to be further investigated. This path analysis method effectively improves the comprehensiveness of the detection results.

[0046] After obtaining the returned signal sequence, it needs to be corrected. In this embodiment, Fourier transform is used to extract frequency domain features to obtain preliminary frequency distribution data. For the extracted frequency domain features, if high-frequency components exceed a preset threshold, it is determined that a signal distortion region exists, and the distribution information of the distortion region is obtained. Based on the distribution information of the distortion region, the influence of the scanning parameters on the signal is analyzed to determine the range of parameters that need to be adjusted. By adjusting the scanning parameters item by item to reduce the power output, the adjusted parameter configuration data is obtained. Using the adjusted parameter configuration data, the signal sequence is resampled to obtain the corrected signal sequence. For the corrected signal sequence, Fourier transform is performed again to extract new frequency domain features, and it is determined whether the high-frequency components still exceed the preset threshold. If the high-frequency components in the new frequency domain features still exceed the preset threshold, the parameter configuration data is fine-tuned to obtain the final signal processing result.

[0047] Specifically, for the original time-domain returned signal sequence s raw [ n ],n =0,1,... C -1, where, C This represents the total number of samples in the time-domain returned signal sequence, with a sampling interval of Δ. t Methods for determining the presence of signal distortion regions include: Step 1: Windowing and Fourier Transform.

[0048] .

[0049] Then, a discrete Fourier transform is performed to obtain the frequency domain representation: .

[0050] The corresponding frequency axis is f k = k / ( C Δ t );[ S raw [ k []] represents the amplitude spectrum of the original time-domain returned signal.

[0051] Step 2: High-frequency distortion detection.

[0052] In an ideal scenario, the energy of a terahertz pulse is primarily concentrated at its center frequency. f c Within a limited bandwidth, signal distortion (especially harmonic distortion caused by receiver saturation or transmitter nonlinearity) can generate anomalous energy in the high-frequency region (far above the system's inherent bandwidth).

[0053] Distortion assessment methods include: ; in, ; In the formula, F s Indicates the sampling frequency; f high , f cutoff These represent the distortion detection start frequency and the theoretical highest effective frequency for normal system operation, respectively.

[0054] when R high / low Exceeding the distortion detection threshold γ When this occurs, it is determined that there is signal distortion at that scan point.

[0055] Step 3: Obtain the distribution of distorted regions.

[0056] Perform the above distortion detection on all scanning points, mark all points where signal distortion exists, and form a distortion area distribution map.

[0057] The process of analyzing the impact of scanning parameters on the signal based on the distribution information of the distorted region, and determining the range of parameters that need to be adjusted, includes: Step 1: Analysis of the degree of influence of parameters.

[0058] Distortion is usually related to transmit power P t Excessive direct correlation can cause the receiver to enter the nonlinear region or saturate. Analysis of the distortion region is needed. D ( x , y Distribution with )=1. Spatial correlation analysis: If the distortion area is concentrated in the spatial location corresponding to the high reflectivity interface (such as the front surface, large delamination defects), it strongly indicates that the saturation is caused by excessive signal strength. Density model correlation analysis: Compare the distortion area with the low density area calculated in step S2. ρ ( x , y Do they overlap? In low-density regions, the attenuation coefficient... α Small, same P t This will result in stronger echoes and make it easier to saturate.

[0059] This embodiment quantifies the degree of impact by estimating the extent of excessive power using an overdrive factor: ; In the formula, This indicates the peak amplitude in the time domain at that point; A sat This represents the maximum amplitude threshold of the receiver's linear operating region; O >1 indicates saturation.

[0060] Step 2: Determine the range of adjustment parameters.

[0061] In this embodiment, the core parameter that needs to be adjusted is the transmission power. P t The adjustment objective is to make the overdrive factor ≤ 1 for all points. This embodiment achieves this by reducing the attenuation factor.

[0062] The new upper limit for the decay factor includes: .

[0063] The adjusted new transmit power is: ; In the formula, β This represents the adjusted attenuation factor.

[0064] Meanwhile, in order to avoid insufficient signal-to-noise ratio for deep weak signals after power reduction, this embodiment adjusts the transmission power as follows: ; In the formula, This indicates the transmission power before adjustment; This indicates the adjusted transmission power; A target This indicates the desired peak amplitude of the signal.

[0065] Step 3: Iterative Adjustment. Using the adjusted parameter configuration data, rescan the distorted region and its surroundings to obtain a new time-domain signal sequence. Perform the above steps on the new time-domain signal sequence.

[0066] S4. Construct a hierarchical analysis model, perform time-domain analysis on the returned signal sequence based on the hierarchical analysis model, obtain the attenuation position and extract the background noise level.

[0067] The hierarchical analysis model performs time-domain analysis on the returned signal sequence to obtain the attenuation position and extract the background noise level. This process includes: S41: Obtain the corrected return signal sequence, perform preliminary processing on the return signal sequence using a pre-established deep learning model, identify abnormal fluctuation segments in the time domain, and obtain potential over-attenuation regions.

[0068] One-dimensional convolutional neural networks or long short-term memory networks are used as sequence segmentation models. The sliding window signal segments of the input corrected return signal sequence are assigned a label (0, 1), where 1 indicates that there are abnormal fluctuations, i.e., potential over-attenuation regions.

[0069] S42: For potential over-attenuation regions, time-domain analysis is used to decompose the signal characteristics within each region, extract the local change trend of the signal, and determine the specific over-attenuation segment.

[0070] In this embodiment, singular value analysis is used to decompose the trend of the signal in the potential excessive attenuation region. First, a trajectory matrix X is constructed, and singular value decomposition is performed. Before selection r Reconstructing the principal components yields the main trend components. T [ m ].

[0071] Calculate the residual e[m]=sRk[m]-T[m] between the corrected return signal sequence and the main trend component, and calculate the energy of the residual: , mThis indicates the amount of signal within a potential over-attenuation region. When the residual energy exceeds the region baseline energy threshold, the region is identified as a specific over-attenuation segment, and the localization mask is updated. M d [ n ].

[0072] S43: From a specific over-attenuation segment, the background noise component is separated using signal processing techniques (such as wavelet threshold denoising or adaptive filters) to obtain the distribution characteristics of the background noise (noise level and distribution test) and determine the level of noise.

[0073] S44: If the background noise level exceeds the preset threshold, further feature extraction is performed on the signal sequence of this segment to obtain the specific frequency domain distribution information of the noise (noise spectrum profile and signal-to-noise ratio spectrum).

[0074] S45: Based on the frequency domain distribution information, a deep learning model is used to conduct a refined assessment of the noise level and determine the specific degree of interference of noise on signal quality.

[0075] This embodiment uses a fully connected neural network. The distribution characteristics extracted by S3 and S4 are input, and the output is a predicted noise interference level score. The noise interference level score is marked by experts based on the interpretability of the signal, where 0 means no interference and 1 means completely unusable.

[0076] S46: Based on the evaluation results of the interference level and the output of the time domain analysis, the signal sequence of the excessively attenuated segment is repaired in a targeted manner to obtain optimized signal data.

[0077] For the degree of interference in the prediction Perform graded repair: for mild interference ( For interference <0.3, spectral subtraction is used for signal enhancement. For moderate interference (0.3 ≤ ... ≤0.7), train an autoencoder to learn the mapping from noisy signals to clean signals. Under heavy interference ( (≥0.7), using signals from neighboring normal regions, alternative signals are generated through interpolation or prediction based on neural networks.

[0078] S5. Based on the background noise level, obtain the layered defect features, and then based on the layered defect features, obtain the defect location distribution.

[0079] Initial data is obtained from the original returned signal sequence. Preprocessing techniques are used to denoise the corrected returned signal sequence, resulting in processed signal data. Based on the processed signal data, background noise is compared with a preset threshold. If the background noise is below the threshold, layered defect information is extracted from the processed signal data to determine preliminary defect features. A deeper analysis of these preliminary defect features is conducted, using a support vector machine (SVM) algorithm to classify the features, obtaining classified defect category information and confidence levels. Based on the classified defect category information, a location mapping operation is performed for each defect category to obtain the specific distribution of defects in the signal sequence and determine the detailed coordinates of the defect locations. By integrating and analyzing the detailed coordinates of the defect locations, the correspondence between defect locations and the signal sequence is obtained, revealing the overall pattern of defect distribution. Statistical analysis tools are used to quantify the distribution patterns, yielding the final defect distribution results. Based on the final defect distribution results, a corresponding defect location distribution map is generated, visualizing the defects in the signal sequence.

[0080] In this embodiment, the peak detection algorithm is used to locate the positions of all candidate defect echoes on the processed signal data. For each candidate defect echo, a set of preliminary defect features is extracted: ;in, A i Indicates the peak echo amplitude; t i Indicates the time of arrival of the echo; FWHM i This represents the full width at half maximum (FWHM) of the echo pulse; R i,i-1 Indicates the amplitude ratio compared to the previous echo; S i Indicated by t i The short-term energy centered on this.

[0081] For echoes classified as true defects, their temporal location is combined with the spatial location of the scanning probe to obtain the three-dimensional spatial coordinates of the defects. Cluster analysis is then performed on these clusters to obtain defect clusters. Statistical measures such as surface defect density, volumetric defect density, depth distribution histogram, and spatial autocorrelation analysis are used to obtain the defect distribution results. The defect location distribution map includes three-dimensional visualization, two-dimensional heatmaps, statistical tables, and detailed defect attributes, comprehensively presenting the visual morphology of defects within the signal sequence.

[0082] S6. Generate a defect mapping map inside the blade based on the defect location distribution.

[0083] By collecting scanning data of the blade's interior, raw image information related to its internal structure is obtained. Layered slicing technology is used for preliminary image segmentation to obtain multi-layered structural data of the blade's interior. For this multi-layered structural data, layers are divided according to the full thickness range, and detection layer information for each layer is obtained. The data from each layer is integrated through overlay processing to determine the correlation distribution between layers. Based on the correlation distribution between layers, the distribution of defect locations in different layers is analyzed. Anomalies are marked using a preset threshold to identify potential defect locations. If the marked defect locations show continuity across multiple layers, a specific distribution model of defects within the blade is constructed using layered defect feature extraction technology, obtaining preliminary mapping results of layered defects. For these preliminary mapping results, a convolutional neural network algorithm is applied to perform in-depth analysis of the correspondence between defect locations and internal structures, obtaining more accurate layered defect distribution data. Based on the accurate layered defect distribution data, a defect mapping map of the blade's interior is generated. This mapping map is then rendered using visualization technology to determine the final non-destructive testing results.

[0084] Example 2 See Figure 2 A terahertz non-destructive testing system for delamination defects in wind turbine blades, comprising: The density distribution determination module is used to acquire a three-dimensional geometric model of the wind turbine blade, extract thickness gradient data from the three-dimensional geometric model, and determine the internal material density distribution of the blade based on the thickness gradient data. The return signal acquisition module is used to determine the scanning parameters based on the density distribution of the material inside the blade, transmit terahertz detection signals into the blade based on the scanning parameters, and acquire the return signals of the detection signals. The background noise extraction module is used to build a hierarchical analysis model, perform time-domain analysis on the returned signal, determine the signal attenuation location, and extract the background noise level. The layered defect extraction module is used to extract layered defect features from the signal at signal attenuation locations based on the background noise level. The blade defect mapping module is used to determine the three-dimensional spatial location of defects based on the characteristics of layered defects, and to generate a three-dimensional mapping map of the layered defects inside the blade based on the three-dimensional spatial location of the defects.

[0085] Example 3 See Figure 3 An electronic device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements a terahertz non-destructive testing method for delamination defects in wind turbine blades.

[0086] Example 4 A computer-readable storage medium storing a computer program, which, when executed by a processor, implements a terahertz non-destructive testing method for delamination defects in wind turbine blades.

[0087] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, read-only optical discs, optical storage, etc.) containing computer-usable program code.

[0088] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0089] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0090] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0091] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the protection scope of the present invention.

Claims

1. A terahertz non-destructive testing method for delamination defects in wind turbine blades, characterized in that, Includes the following steps: A three-dimensional geometric model of a wind turbine blade is obtained, thickness gradient data is extracted from the three-dimensional geometric model, and the internal material density distribution of the blade is determined based on the thickness gradient data. The scanning parameters are determined based on the material density distribution inside the blade, and a terahertz detection signal is emitted into the blade based on the scanning parameters, and the return signal of the detection signal is collected. A hierarchical analysis model is constructed to perform time-domain analysis on the returned signal, determine the signal attenuation location, and extract the background noise level; Based on the background noise level, layered defect features are extracted from the signal at the signal attenuation location; The three-dimensional spatial location of the defect is determined based on the layered defect characteristics, and a three-dimensional mapping map of the layered defects inside the blade is generated based on the three-dimensional spatial location of the defect.

2. The terahertz non-destructive testing method for delamination defects in wind turbine blades according to claim 1, characterized in that, The scanning parameters include the center frequency and the scanning step size; The step of determining the scanning parameters based on the internal material density distribution of the blade includes: Based on the density values ​​of different regions of the blade, the center frequency of each region is adaptively selected within a preset frequency range. Based on the thickness gradient modulus of different regions of the blade, the scanning step distance of each region is adaptively determined within a preset step distance range.

3. The terahertz non-destructive testing method for delamination defects in wind turbine blades according to claim 1, characterized in that, After acquiring the return signal of the detection signal, the return signal is corrected, including: Extract the frequency domain features of the returned signal and detect signal distortion regions; the signal distortion regions are areas where high-frequency components exceed a preset threshold. Analyze the impact of scanning parameters on the signal distortion region to determine the parameter adjustment range; Adjust the scanning parameters according to the parameter adjustment range, re-acquire the return signal of the signal distortion area, and obtain the corrected return signal.

4. The terahertz non-destructive testing method for delamination defects in wind turbine blades according to claim 3, characterized in that, The region of distortion in the detected signal includes: Perform a windowed Fourier transform on the returned signal to obtain the frequency domain amplitude spectrum; Calculate the ratio of high-frequency energy to low-frequency energy in the frequency domain amplitude spectrum; When the ratio exceeds the preset distortion detection threshold, it is determined that there is signal distortion at the scan point.

5. The terahertz non-destructive testing method for delamination defects in wind turbine blades according to claim 3, characterized in that, The analysis of the influence of the scanning parameters on the signal distortion region, and the determination of the parameter adjustment range, include: The overdrive factor is obtained by calculating the ratio of the peak amplitude of the returned signal in the time domain to the maximum amplitude threshold of the receiver's linear operating region in the signal distortion region. The upper limit for adjusting the transmit power is determined based on the overdrive factor; The step of adjusting the scanning parameters according to the parameter adjustment range includes: The adjusted transmit power is calculated based on the overdrive factor and the desired peak signal amplitude. The returned signal from the signal distortion area is reacquired using the adjusted transmit power.

6. The terahertz non-destructive testing method for delamination defects in wind turbine blades according to claim 1, characterized in that, The construction of the hierarchical analysis model to perform time-domain analysis on the returned signal includes: Identify abnormal fluctuation segments in the returned signal to obtain potential over-attenuation regions; The signal in the potential over-attenuation region is decomposed, local variation trends are extracted, and specific over-attenuation segments are identified. Background noise components are separated from the specific excessive attenuation segments to obtain background noise distribution characteristics; The signal of the excessively attenuated segment is repaired based on the background noise distribution characteristics to obtain optimized signal data.

7. The terahertz non-destructive testing method for delamination defects in wind turbine blades according to claim 1, characterized in that, The step of extracting layered defect features from the signal at the signal attenuation location based on the background noise level includes: The signal amplitude at the signal attenuation location is compared with the background noise level; Signal features whose amplitude exceeds a preset multiple of the background noise level are extracted as layered defect features.

8. A terahertz non-destructive testing system for delamination defects in wind turbine blades, characterized in that, include: The density distribution determination module is used to acquire a three-dimensional geometric model of a wind turbine blade, extract thickness gradient data from the three-dimensional geometric model, and determine the internal material density distribution of the blade based on the thickness gradient data. The return signal acquisition module is used to determine the scanning parameters based on the material density distribution inside the blade, transmit a terahertz detection signal into the blade based on the scanning parameters, and acquire the return signal of the detection signal. The background noise extraction module is used to construct a hierarchical analysis model, perform time-domain analysis on the returned signal, determine the signal attenuation location, and extract the background noise level. The layered defect extraction module is used to extract layered defect features from the signal at the signal attenuation location based on the background noise level. The blade defect mapping module is used to determine the three-dimensional spatial location of the defect based on the layered defect characteristics, and to generate a three-dimensional mapping map of the layered defects inside the blade based on the three-dimensional spatial location of the defect.

9. An electronic device, characterized in that, The method includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the terahertz nondestructive testing method for delamination defects in wind turbine blades as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the terahertz non-destructive testing method for delamination defects in wind turbine blades as described in any one of claims 1-7.

Images