A method and system for detecting the aging state of a bushing based on terahertz imaging

Abstract

The application relates to the technical field of terahertz nondestructive testing, and discloses a casing aging state detection method and system based on terahertz imaging. The method comprises the following steps: collecting terahertz time-domain transmission signals at different positions of a casing to be detected; performing filter back-projection processing on the terahertz time-domain transmission signals at different positions to obtain an aging detection image of the casing to be detected, wherein a composite filter is introduced in the filter back-projection processing, and the composite filter is designed to adopt a three-section frequency response structure to perform frequency section regulation and control on the terahertz time-domain transmission signals; performing texture feature extraction and hierarchical identification on the aging detection image to obtain an aging state detection result of the casing to be detected. The application comprehensively utilizes the non-ionization, high penetration and millimeter-level resolution characteristics of terahertz waves, combines signal processing and intelligent identification algorithms, and realizes accurate and efficient detection of the aging state of the casing.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of terahertz nondestructive testing, in particular to a bushing aging state detection method and system based on terahertz imaging. BACKGROUND

[0002] As a key component for ensuring the insulation performance of equipment in the power system, the aging state of the bushing is directly related to the safe and stable operation of the power equipment. With the development of the power system towards high voltage and large capacity, the bushing is subjected to the combined action of electric, thermal, mechanical stress and environmental factors for a long time, which is prone to insulation aging, local defects and other problems. If it is not detected and evaluated in time, it may cause equipment failure or even large-scale power outage accidents. Therefore, it is of great significance to accurately and efficiently detect the aging state of the bushing for early warning of potential risks and prolonging the service life of the equipment.

[0003] At present, a variety of bushing aging state detection technologies have been developed, but all have limitations. For example, the infrared imaging technology relies on the difference in surface thermal radiation of the bushing to judge internal defects, but its detection depth is limited, and the thermal signal of internal deep aging defects is severely attenuated after transmission, which is difficult to effectively identify. At the same time, environmental temperature, humidity, solar radiation and other external factors can easily interfere with the surface thermal radiation characteristics, resulting in large deviation of the detection results. For example, the ultra-high frequency method realizes detection by capturing ultra-high frequency electromagnetic waves generated by partial discharge, but due to the influence of the structure of the bushing, there is complex attenuation and reflection in the propagation of electromagnetic waves, and the signal analysis is difficult, and it is easy to be disturbed by the surrounding electromagnetic environment, resulting in low defect positioning accuracy, high missed detection or false detection rate. For example, the oil chromatography method needs to extract oil samples for analysis of dissolved gas content, which is complicated and cannot realize real-time monitoring. For early and slight aging, the change of gas in the oil is weak, and it is difficult to quickly and accurately reflect the aging state, which cannot meet the early warning demand. The dielectric response technology evaluates aging based on the polarization and loss characteristics of insulating materials, but the measurement results are easily disturbed by environmental factors such as temperature and humidity, and the precision and repeatability are poor; for local aging, delamination and other complex defects, it is difficult to comprehensively and accurately evaluate the characteristic signals. SUMMARY

[0004] In order to solve the problems of poor non-contact, low recognition accuracy and inability to quantify and grade in the traditional bushing aging state detection technology, the present application provides a bushing aging state detection method and system based on terahertz imaging.

[0005] In a first aspect, the embodiments of the present application provide a bushing aging state detection method based on terahertz imaging, comprising:

[0006] Collecting terahertz time domain transmission signals at different positions of the bushing to be tested;

[0007] The terahertz time-domain transmission signals at different positions are subjected to filter back projection processing to obtain an aging detection image of the to-be-tested sleeve, wherein a composite filter is introduced in the filter back projection processing, and the composite filter is designed to perform frequency segment regulation and control on the terahertz time-domain transmission signals by adopting a three-section frequency response structure;

[0008] The aging detection image is subjected to texture feature extraction and hierarchical identification to obtain an aging state detection result of the to-be-tested sleeve.

[0009] Preferably, the filter back projection processing of the terahertz time-domain transmission signals at different positions to obtain the aging detection image of the to-be-tested sleeve comprises:

[0010] The terahertz time-domain transmission signals at different positions are subjected to frequency domain conversion to obtain frequency domain data at corresponding positions;

[0011] The frequency domain data at each position are respectively subjected to filter processing by the composite filter to obtain filtered frequency domain data at the corresponding positions;

[0012] The filtered frequency domain data at each position are respectively subjected to back projection reconstruction to obtain the aging detection image of the to-be-tested sleeve.

[0013] Preferably, the filter processing of the frequency domain data at each position by the composite filter to obtain the filtered frequency domain data at the corresponding positions comprises:

[0014] A composite filter is constructed based on a three-section frequency response structure comprising a low-frequency suppression zone, a main passband reinforcement zone and a high-frequency attenuation compensation zone;

[0015] The frequency domain data at each position are respectively subjected to filter processing based on the composite filter to obtain filtered frequency domain data at the corresponding positions.

[0016] Preferably, the construction of the composite filter based on the three-section frequency response structure comprising the low-frequency suppression zone, the main passband reinforcement zone and the high-frequency attenuation compensation zone comprises:

[0017] The low-frequency suppression zone is constructed based on a low-frequency truncation mechanism designed to set signal components lower than a noise critical frequency to zero;

[0018] The main passband reinforcement zone is constructed based on an adaptive coherent compensation window function designed to perform artifact suppression on signal components between the noise critical frequency and a main passband cutoff frequency;

[0019] The high-frequency attenuation compensation region is constructed based on an exponential attenuation compensation function designed for noise suppression and edge preservation of signal components higher than the main passband cutoff frequency.

[0020] The low-frequency suppression region, the main passband reinforcement region and the high-frequency attenuation compensation region are integrated to obtain a composite filter.

[0021] Preferably, the texture feature extraction and hierarchical identification of the aging detection image are performed to obtain the aging state detection result of the to-be-tested sleeve, including:

[0022] The texture feature extraction is performed on the aging detection image to obtain a texture feature parameter.

[0023] The hierarchical identification is performed on the texture feature parameter to obtain the aging state detection result of the to-be-tested sleeve.

[0024] Preferably, the texture feature extraction performed on the aging detection image to obtain a texture feature parameter includes:

[0025] The continuous wavelet transform is performed on the aging detection image to obtain a texture feature parameter.

[0026] In the continuous wavelet transform process, an adaptive weighting mechanism is introduced, which is designed to dynamically select an optimal wavelet basis function through dictionary learning, introduce a scale weighting function for response enhancement of a target frequency band, and perform directional texture enhancement and edge preservation filtering on a wavelet energy map after the transform. The texture feature parameter includes a texture contrast, a spectral entropy and a wavelet detail coefficient variance.

[0027] Preferably, the hierarchical identification performed on the texture feature parameter to obtain the aging state detection result of the to-be-tested sleeve includes:

[0028] The texture feature parameter is input into a pre-constructed aging hierarchical model for grade prediction to obtain an aging grade of the to-be-tested sleeve, wherein the aging hierarchical model is constructed based on a support vector machine.

[0029] In a second aspect, an embodiment of the present application provides a sleeve aging state detection system based on terahertz imaging, including:

[0030] A terahertz wave imaging device is configured to collect terahertz time-domain transmission signals at different positions of a to-be-tested sleeve, and perform filter back-projection processing on the terahertz time-domain transmission signals at different positions to obtain an aging detection image of the to-be-tested sleeve. In the filter back-projection processing, a composite filter is introduced, which is designed to use a three-section frequency response structure to regulate and control the terahertz time-domain transmission signals in different frequency bands.

[0031] An aging state evaluation module is configured to perform texture feature extraction and hierarchical recognition on the aging detection image to obtain an aging state detection result of the sleeve under test.

[0032] Preferably, the terahertz wave imaging device comprises a terahertz emission end, a terahertz detection end, a vector network analyzer and a data processing terminal.

[0033] The terahertz emission end is configured to emit modulated terahertz waves to penetrate the sleeve under test.

[0034] The terahertz detection end is configured to collect terahertz time-domain transmission signals at different positions of the sleeve under test.

[0035] The vector network analyzer is configured to simultaneously measure the terahertz waves and the terahertz time-domain transmission signals, and calculate S parameters based on the vector relationship between the terahertz waves and the terahertz time-domain transmission signals.

[0036] The data processing terminal is configured to perform filter back-projection processing on the terahertz time-domain transmission signals at different positions to obtain an aging detection image of the sleeve under test.

[0037] Preferably, the terahertz wave imaging device further comprises a mechanical scanning device.

[0038] The mechanical scanning device comprises a roller and a guide rail, the roller is configured to support and drive the sleeve under test to rotate 360°, and the guide rail is configured to drive the terahertz emission end and the terahertz detection end to translate along the axial direction of the sleeve under test.

[0039] Compared with the prior art, the method and system for detecting the aging state of a bushing based on terahertz imaging have the beneficial effects that: the terahertz imaging technology is adopted to deeply penetrate the insulating medium of the bushing, directly obtain internal structure information, overcome the limitation that traditional infrared technology can only detect the surface, and does not need to contact the bushing to be detected, thereby avoiding potential damage to the bushing in the detection process; a composite filter with a three-section frequency response structure is introduced in the filtered back-projection processing to accurately control the frequency bands of the terahertz time-domain transmission signal, effectively suppress the interference artifacts caused by the coherence of terahertz waves, improve the signal-to-noise ratio of the image and retain key edge details, and provide a high-quality aging detection image for subsequent accurate identification; based on texture feature extraction and hierarchical identification, subtle aging features can be captured, and the automatic grading of the aging state is realized, thereby breaking through the limitation of relying on manual experience for judgment, having higher detection sensitivity, accuracy and standardization, and being suitable for engineering batch detection and online monitoring scenes. In general, the present application comprehensively utilizes the non-ionizing, high-penetration and millimeter-level resolution characteristics of terahertz waves, combines signal processing and intelligent identification algorithms, and realizes accurate and efficient detection of the aging state of the bushing. BRIEF DESCRIPTION OF DRAWINGS

[0040] Figure 1 is a flowchart of a method for detecting the aging state of a bushing based on terahertz imaging according to an embodiment of the present application;

[0041] Figure 2 is a structural schematic diagram of a terahertz imaging device according to an embodiment of the present application;

[0042] Figure 3 is another structural schematic diagram of a terahertz imaging device according to an embodiment of the present application;

[0043] Figure 4 is a flowchart of filtered back-projection processing according to an embodiment of the present application;

[0044] Figure 5 is a flowchart of texture feature extraction and hierarchical identification according to an embodiment of the present application;

[0045] Figure 6 is a structural schematic diagram of a system for detecting the aging state of a bushing based on terahertz imaging according to an embodiment of the present application;

[0046] REFERENCE SIGNS:

[0047] 1, terahertz wave imaging device; 2, aging state evaluation module; 11, terahertz transmitting end; 12, terahertz detecting end; 13, vector network analyzer; 14, data processing terminal; 15, roller; 16, guide rail. DETAILED DESCRIPTION

[0048] The specific embodiments of the present application will be further described in detail below with reference to the accompanying drawings and examples. The following examples are used to illustrate the present application, but are not intended to limit the scope of the present application.

[0049] In the description of the present application, it should be noted that unless otherwise defined, all technical and scientific terms used in the present application have the same meaning as commonly understood by one of ordinary skill in the art. The terms used in the specification of the present application are only for the purpose of describing specific embodiments and are not intended to limit the present application. For those of ordinary skill in the art, the specific meanings of the above terms in the present application can be understood according to the specific circumstances.

[0050] Terahertz waves are between millimeter waves and far infrared waves, have good penetration and non-ionizing properties, and can effectively penetrate the insulating material of the bushing to non-destructively detect the internal structure and aging defects. Therefore, the present application combines terahertz imaging technology to realize precise and efficient evaluation of the aging state of the bushing, thereby ensuring the safe and stable operation of the power system.

[0051] As shown in Figure 1 , it is a flowchart of a bushing aging state detection method based on terahertz imaging according to an embodiment of the present application. Referring to Figure 1 , a bushing aging state detection method based on terahertz imaging according to an embodiment of the present application includes the following steps:

[0052] S1, collecting terahertz time domain transmission signals at different positions of the to-be-tested bushing;

[0053] S2, filtering back-projection processing the terahertz time domain transmission signals at different positions to obtain an aging detection image of the to-be-tested bushing;

[0054] In the filtering back-projection processing process, a composite filter is introduced, which is designed to use a three-section frequency response structure to control the terahertz time domain transmission signals in different frequency bands.

[0055] S3, texture feature extraction and hierarchical identification are performed on the aging detection image to obtain an aging state detection result of the to-be-tested bushing.

[0056] It should be noted that steps S1-S2 are implemented by a terahertz imaging device according to an embodiment of the present application. That is, the terahertz wave imaging device is used to collect terahertz time domain transmission signals at different positions of the to-be-tested bushing, and to filter back-projection process the terahertz time domain transmission signals at different positions to obtain an aging detection image of the to-be-tested bushing. In the filtering back-projection processing process, a composite filter is introduced, which is designed to use a three-section frequency response structure to control the terahertz time domain transmission signals in different frequency bands.

[0057] As shown in Figure 2The diagram shown is a structural schematic of a terahertz imaging device according to an embodiment of the present invention. (Refer to...) Figure 2 This invention provides a terahertz imaging device, comprising a terahertz transmitter 11, a terahertz detector 12, a vector network analyzer 13, and a data processing terminal 14.

[0058] The terahertz transmitter is used to emit modulated terahertz waves that penetrate the bushing under test; that is, the terahertz transmitter emits continuous or pulsed terahertz waves to penetrate the bushing under test. Specifically, in this embodiment, the terahertz transmitter is a 390 GHz terahertz wave source. Figure 2 The direction of the middle arrow indicates the direction of terahertz wave propagation.

[0059] The terahertz detector is used to acquire terahertz time-domain transmission signals at different locations on the sleeve under test. It can be understood that step S1 is actually implemented by the terahertz detector.

[0060] The vector network analyzer is used to simultaneously measure terahertz waves and terahertz time-domain transmission signals, and calculates S-parameters based on the vector relationship between the terahertz waves and the terahertz time-domain transmission signals. Specifically, the vector network analyzer is connected to both the terahertz transmitter and the terahertz detector. The terahertz transmitter is connected to the vector network analyzer and, together with a lens, forms a collimated beam system to collimate the emitted wave into a parallel beam, uniformly illuminating the surface of the sleeve under test. The terahertz detector is connected to the vector network analyzer and, together with a lens, focuses the scattered wave to improve the signal-to-noise ratio. In this embodiment, all lenses are made of polytetrafluoroethylene (PTFE).

[0061] The data processing terminal is used to filter and back-project the terahertz time-domain transmission signals at different locations to obtain the aging detection image of the sleeve under test. It can be understood that step S2 is actually implemented by the data processing terminal. The data processing terminal can be, but is not limited to, a computer.

[0062] like Figure 3 As shown, this is another structural schematic diagram of a terahertz imaging device according to an embodiment of the present invention. (Refer to...) Figure 3 This invention provides a terahertz imaging device, which further includes a mechanical scanning device. The mechanical scanning device includes a roller 15 and a guide rail 16.

[0063] The rollers support the sleeve under test and drive it to rotate 360°. The guide rails drive the terahertz transmitter and detector to translate along the axis of the sleeve under test to cover the entire length and cross-section of the sleeve. In this embodiment, the rotation step angle is set to 2° and the translation step size is set to 1mm.

[0064] To facilitate understanding, the implementation process of step S1 will be explained below with reference to the terahertz imaging device of this embodiment:

[0065] S1, collecting terahertz time-domain transmission signals at different positions of the to-be-tested casing;

[0066] Specifically, the terahertz emission end emits terahertz waves in a modulated form under the excitation of the vector network analyzer, and after collimation by the polytetrafluoroethylene lens, the terahertz waves uniformly irradiate the surface of the to-be-tested casing. At the same time, the roller in the mechanical scanning device supports the to-be-tested casing and drives it to rotate by 360° at a step angle of 2°, and the guide rail drives the terahertz emission end and the detection end to translate along the axis of the to-be-tested casing at a step length of 1 mm, so as to realize scanning coverage of the full length and full cross section of the casing through the cooperative movement of rotation and translation.

[0067] Further, the terahertz detection end focuses the scattered waves that have penetrated the casing in cooperation with the polytetrafluoroethylene lens, and under the cooperative action of the vector network analyzer, terahertz time-domain transmission signals at different scanning positions are collected, and the full-range signal collection is completed.

[0068] It should be noted that before step S1, the surface of the to-be-tested casing needs to be pretreated, including removing oil stains, rust and oxide layers, so as to ensure that the collected transmission signals can truly reflect the internal structure information of the casing and provide reliable raw data for subsequent processing.

[0069] S2, filtering back-projection processing is performed on the terahertz time-domain transmission signals at different positions to obtain an aging detection image of the to-be-tested casing;

[0070] As shown in Figure 4 , it is a flowchart of step S2 of the embodiment of the present application. Referring to Figure 4 , step S2 includes:

[0071] S201, frequency domain conversion is performed on the terahertz time-domain transmission signals at different positions to obtain frequency domain data at the corresponding positions;

[0072] The frequency domain conversion is performed on the terahertz time-domain transmission signals at different positions by using fast Fourier transform to obtain frequency domain data at the corresponding positions. Through fast Fourier transform, the amplitude and phase information of the transmission signals at different frequencies can be obtained.

[0073] S202, a composite filter is used to filter the frequency domain data at each position respectively to obtain filtered frequency domain data at the corresponding positions;

[0074] Considering the coherence, attenuation characteristics and frequency band characteristics of terahertz waves, the filtering function is customized designed in this step, that is, the composite filter is designed to have a three-section frequency response structure to regulate and control the terahertz time-domain transmission signals in different frequency bands, so as to solve the special problems in the terahertz wave field.

[0075] Specifically, step S202 includes:

[0076] 1) Constructing a composite filter based on a three-section frequency response structure containing a low-frequency suppression zone, a main passband enhancement zone, and a high-frequency attenuation compensation zone;

[0077] The composite filter achieves effective suppression of interference artifacts, preservation of structural edge details, and flexible control of high-frequency noise by setting the low-frequency suppression zone, the main passband enhancement zone, and the high-frequency attenuation compensation zone.

[0078] Specifically, step 1) includes:

[0079] 11) Constructing a low-frequency suppression zone based on a low-frequency cutoff mechanism;

[0080] The low-frequency cutoff mechanism is designed to set the signal components below the noise critical frequency to zero.

[0081] 12) Constructing a main passband enhancement zone based on an adaptive coherent compensation window function;

[0082] The adaptive coherent compensation window function is designed to suppress artifacts for signal components between the noise critical frequency and the main passband cutoff frequency.

[0083] Since terahertz waves have high coherence, when there is micro-interference in a layered structure or edge, it will introduce quasi-periodic artifacts in the frequency spectrum. Therefore, an adaptive coherent compensation window function is introduced, as follows:

[0084]

[0085] wherein, W(f) represents the adaptive coherent compensation window function, f represents the frequency of the terahertz signal, fc represents the noise critical frequency, fcutoff represents the main passband cutoff frequency.

[0086] It should be noted that the window function has the following characteristics: the beginning and end gradually approach 0, which can suppress truncated oscillations; the middle section enhances the main information frequency band; it can naturally suppress periodic artifacts, especially effectively handle the light ring interference stripes formed by the edges of tomographic imaging.

[0087] 13) Constructing a high-frequency attenuation compensation zone based on an exponential attenuation compensation function;

[0088] The exponential attenuation compensation function is designed to suppress noise and preserve edges for signal components above the main passband cutoff frequency.

[0089] When terahertz waves penetrate highly absorbent materials (such as water-containing layers), their high-frequency components are severely attenuated, causing edge blurring. In order to preserve certain edge details, a high-frequency tail exponential compensation attenuation mechanism, i.e., an exponential attenuation compensation function, is introduced, as follows:

[0090]

[0091] wherein, denotes an exponential decay compensation function, denotes a compensation coefficient, denotes a transition slope control parameter.

[0092] It should be noted that through this mechanism, part of the structural details of the casing under test can be retained while suppressing noise.

[0093] 14) The low-frequency suppression zone, the main passband reinforcement zone and the high-frequency decay compensation zone are integrated to obtain a composite filter.

[0094] Specifically, the composite filter is characterized by the following formula:

[0095]

[0096] wherein, denotes a frequency response function of the composite filter.

[0097] 2) The frequency domain data at each position are respectively filtered based on the composite filter to obtain filtered frequency domain data at the corresponding position.

[0098] Based on the three-section frequency response structure of the composite filter, the frequency domain data at each position are processed one by one in the frequency band to obtain the filtered frequency domain data at the corresponding position.

[0099] S203, the filtered frequency domain data at each position are respectively back-projected and reconstructed to obtain an aging detection image of the casing under test.

[0100] For the filtered frequency domain data at each position, the inverse fast Fourier transform is first used to convert the data to the spatial domain to obtain the spatial domain signal at the corresponding position, and then the spatial domain projection data under different detection angles are superimposed in the spatial domain, that is, the spatial domain signals corresponding to each angle are sequentially added. The result obtained after superimposing is an aging detection image that can reflect the aging condition of the casing under test.

[0101] S3, texture feature extraction and hierarchical recognition are performed on the aging detection image to obtain an aging state detection result of the casing under test.

[0102] After the image reconstruction is completed, in order to realize intelligent recognition and grade determination of the aging state of the casing under test, a multi-scale spectral texture feature extraction method is introduced in the aging state evaluation stage, and an automatic recognition process is constructed in combination with a machine learning model.

[0103] For example, Figure 5As shown, it is the flowchart of step S3 of the embodiment of the present application. Referring to Figure 5 , step S3 includes:

[0104] S301, texture feature extraction is performed on the aging detection image to obtain texture feature parameters.

[0105] The aging detection image is subjected to continuous wavelet transform to obtain the texture feature parameters.

[0106] It should be noted that the traditional continuous wavelet transform has certain limitations in processing the aging detection image. For example, the fixed wavelet basis function lacks pertinence and is difficult to adapt to the complex texture characteristics of different aging areas; the single scale processing method cannot take into account both the subtle aging characteristics and the overall texture trend, which may lead to insufficient precision of feature extraction; and the transformed image is easily disturbed by noise, which may mask the texture differences between the aging area and the normal area and affect the distinguishability of the feature parameters. Therefore, the self-adaptive weighting mechanism is introduced in this step to improve the continuous wavelet transform, so as to enhance the sensitivity and extraction accuracy of the aging characteristics.

[0107] Specifically, the self-adaptive weighting mechanism is designed to dynamically select the optimal wavelet basis function that adapts to the texture characteristics of different aging areas through dictionary learning, introduce a scale weighting function to enhance the response of the target frequency band related to the aging characteristics, and perform directional texture enhancement and edge preservation filtering on the wavelet energy map after transformation, so as to finally extract the texture contrast, spectral entropy and wavelet detail coefficient variance from the processed image as the texture feature parameters.

[0108] Further, the continuous wavelet transform with the self-adaptive weighting mechanism is specifically described as follows:

[0109] 1) Adaptive selection of wavelet basis function

[0110] The dictionary learning idea is introduced to dynamically select the optimal wavelet function according to the spectral characteristics of typical defect textures in the sample; and the optimal wavelet basis for the current window is selected according to the maximum energy concentration degree or the minimum reconstruction error principle.

[0111] Specifically, for the texture complexity of different areas in the aging detection image, a candidate set of wavelet bases covering various waveform characteristics is first defined. When processing the image, the continuous wavelet transform energy concentration degree of the candidate wavelet basis and the signal in each local window (corresponding to the local features of the aging texture) is calculated, and the wavelet basis that best adapts to the texture characteristics of the current window is dynamically selected by maximizing the energy concentration degree or minimizing the reconstruction error.

[0112] For example, for the texture area of fine cracks on the surface of the casing, a wavelet basis with more dense wave oscillation is preferred to ensure accurate capture of high-frequency details; while for the large-area uniform aging area, a wavelet basis with better smoothness is selected to improve the stability of overall feature extraction.

[0113] 2) Scale weighting mechanism;

[0114] In order to strengthen the target frequency band corresponding to the aging feature, a scale weighting function is introduced as follows:

[0115]

[0116] wherein, denotes the scale weighting function, denotes the wavelet transform scale factor, denotes the target defect dominant scale, denotes the weight bandwidth. In the continuous wavelet transform, the scale and the translation of the wavelet coefficient After that, the wavelet coefficient is weighted by the scale weighting function, so that the transform response of the target scale (such as the 0.2THz~0.8THz frequency band corresponding to the aging feature) is significantly improved, while the noise interference of irrelevant frequency bands is suppressed, and the energy distribution of the aging texture in the time-frequency domain is more prominent, providing a more targeted signal basis for subsequent feature extraction.

[0117] 3) Texture enhancement and artifact reconstruction.

[0118] After completing the weighted transformation to obtain the wavelet energy spectrum, it is necessary to suppress background noise and strip artifacts, and enhance the defect texture edge.

[0119] Specifically, structure tensor guided filtering or nonlinear edge preserving denoising (such as anisotropic diffusion equation) is introduced to retain and enhance the real aging texture edge in the energy spectrum, while smoothing the noise and artifact area.

[0120] Through the above processing, the signal-to-noise ratio of the energy spectrum is improved, and the texture details of the aging area (such as crack direction and aging layer boundary) are clearer. Finally, the optimized multi-scale texture image matrix is output, providing high-quality data support for subsequent extraction of texture contrast, spectral entropy, and wavelet detail coefficient variance, so that the aging feature is more easily quantified and identified in the image.

[0121] S302, classifying and identifying the texture feature parameters to obtain the aging state detection result of the casing to be measured.

[0122] The texture feature parameters are input into the pre-constructed aging classification model for grade prediction to obtain the aging grade of the casing to be measured. The aging classification model is constructed based on a support vector machine.

[0123] Specifically, a batch of casing samples with clear aging degree are collected, and the aging detection image of each casing sample is obtained based on step S2. Texture feature parameters are extracted from these images, that is, according to step S301, the texture contrast is calculated by calculating the gray level co-occurrence matrix, the spectral entropy is calculated by frequency spectrum transformation, and the wavelet detail coefficient variance is obtained by discrete wavelet decomposition, forming a feature vector composed of the three types of indicators. At the same time, according to the actual aging condition of the sample, the aging state is labeled. For example, A represents normal, B represents mild aging, C represents moderate aging, and D represents severe aging, thereby constructing a training data set containing feature vectors and corresponding aging state labels, and providing a training data set for subsequent model training.

[0124] Further, a support vector machine is used to build an aging grading model. Specifically, the multi-class support vector machine (multi-class kernel SVM) is used in this embodiment, the feature vector prepared in the early stage is used as the model input, and the labeled aging state label (A, B, C, D) is used as the target label, so that the model learns the correlation between the texture feature parameters and the aging state.

[0125] It should be noted that in the model training process, the cross-validation method is used to evaluate the generalization ability of the model, and the classification boundary is continuously optimized to enable the model to more accurately distinguish different aging states, thereby completing the construction of the aging grading model. The model can output the corresponding aging grade prediction result based on the input texture feature parameters.

[0126] Further, in the actual aging detection of the casing to be detected, the steps include:

[0127] 1) Image reconstruction and feature extraction;

[0128] First, the aging detection image of the detection area of the casing to be detected is reconstructed by filter back projection processing, and then the aging detection image is subjected to feature extraction, that is, the texture contrast is calculated by calculating the gray level co-occurrence matrix, the spectral entropy is obtained by frequency spectrum transformation, and the wavelet detail coefficient variance is obtained by discrete wavelet decomposition, thereby constructing the texture feature vector corresponding to the detection area.

[0129] 2) Model prediction;

[0130] The constructed texture feature vector is input into the pre-constructed and trained aging grading model, and the model analyzes the input features according to the learned correlation, and outputs the aging grade label of the corresponding detection area of the casing to be detected, such as A (normal), B (mild aging), C (moderate aging), and D (severe aging).

[0131] 3) Result presentation.

[0132] In the detection software interface, the color marking is presented according to the aging level. For example, the A class is marked with green, the B class is marked with yellow, the C class is marked with orange, and the D class is marked with red, so that the aging state detection result of the to-be-detected bushing is intuitively displayed, and the aging state detection of the to-be-detected bushing is completed.

[0133] The method for detecting the aging state of a bushing based on terahertz imaging adopts terahertz imaging technology, can deeply penetrate the insulating medium of the bushing, directly obtains internal structure information, overcomes the limitation that traditional infrared technology can only detect the surface, and does not need to contact the to-be-detected bushing, thereby avoiding potential damage to the bushing in the detection process. A composite filter with a three-section frequency response structure is introduced in the filtered back-projection processing to accurately control the frequency bands of the terahertz time-domain transmission signal, effectively suppresses the interference artifacts caused by the coherence of terahertz waves, improves the signal-to-noise ratio of the image and preserves key edge details, and provides a high-quality aging detection image for subsequent accurate identification. Based on texture feature extraction and hierarchical identification, subtle aging features can be captured, and automatic grading of the aging state is realized, thereby breaking through the limitation of relying on manual experience for judgment, having higher detection sensitivity, accuracy and standardization, and being suitable for engineering batch detection and online monitoring scenarios. In general, the present application comprehensively utilizes the non-ionizing, high-penetration and millimeter-level resolution characteristics of terahertz waves, combines signal processing and intelligent identification algorithms, and realizes accurate and efficient detection of the aging state of the bushing.

[0134] As shown in Figure 6 , which is a structural schematic diagram of the system for detecting the aging state of a bushing based on terahertz imaging. Referring to Figure 6 , the system for detecting the aging state of a bushing based on terahertz imaging comprises:

[0135] The terahertz wave imaging device 1 is used to collect terahertz time-domain transmission signals at different positions of the to-be-detected bushing, and perform filtered back-projection processing on the terahertz time-domain transmission signals at different positions to obtain an aging detection image of the to-be-detected bushing. In the filtered back-projection processing, a composite filter is introduced, and the composite filter is designed to control the frequency bands of the terahertz time-domain transmission signal.

[0136] The aging state evaluation module 2 is used to extract texture features and perform hierarchical identification on the aging detection image to obtain an aging state detection result of the to-be-detected bushing.

[0137] Specifically, the terahertz wave imaging device comprises a terahertz emission end, a terahertz detection end, a vector network analyzer and a data processing terminal.

[0138] The terahertz emission end is used to emit modulated terahertz waves to penetrate the to-be-detected bushing.

[0139] The terahertz detection end is used for collecting terahertz time-domain transmission signals at different positions of the to-be-detected bushing;

[0140] The vector network analyzer is used for simultaneously measuring the terahertz wave and the terahertz time-domain transmission signal, and calculating S parameters based on a vector relationship between the terahertz wave and the terahertz time-domain transmission signal;

[0141] The data processing terminal is used for performing filtered back-projection processing on the terahertz time-domain transmission signals at different positions, to obtain an aging detection image of the to-be-detected bushing.

[0142] Further, the terahertz imaging device further comprises a mechanical scanning device.

[0143] The mechanical scanning device comprises a roller and a guide rail, the roller is used for supporting the to-be-detected bushing and driving the to-be-detected bushing to rotate 360 degrees, and the guide rail is used for driving the terahertz emission end and the terahertz detection end to translate along an axial direction of the to-be-detected bushing.

[0144] It should be noted that the aging state evaluation module in the above-mentioned bushing aging state detection system based on terahertz imaging can be realized by software, hardware and a combination thereof. The aging state evaluation module can be embedded in or independent of the processor in the computer device in hardware form, or can be stored in the memory in the computer device in software form, so that the processor calls and executes the corresponding operations of the aging state evaluation module. For specific limitations of a terahertz imaging device, refer to the limitations of a terahertz imaging device in the above, both have the same functions and effects, and will not be repeated here.

[0145] In summary, the bushing aging state detection method and system based on terahertz imaging of the embodiment of the present application adopts terahertz imaging technology, can deeply penetrate the insulating medium of the bushing, directly obtains internal structure information, overcomes the limitation that traditional infrared technology can only detect the surface, and does not need to contact the to-be-detected bushing, avoiding potential damage to the bushing during the detection process; by introducing a composite filter with a three-section frequency response structure in the filtered back-projection processing, the terahertz time-domain transmission signal is precisely controlled in different frequency bands, effectively suppressing the interference artifacts caused by the coherence of the terahertz wave, improving the image signal-to-noise ratio and preserving key edge details, providing a high-quality aging detection image for subsequent accurate identification; based on texture feature extraction and hierarchical identification, subtle aging features can be captured, and automatic grading of the aging state can be realized, breaking through the limitation of relying on manual experience judgment, and having higher detection sensitivity, accuracy and standardization, suitable for engineering batch detection and online monitoring scenarios. In general, the present application comprehensively utilizes the non-ionizing, high-penetration and millimeter-level resolution characteristics of terahertz waves, combines signal processing and intelligent identification algorithms, and realizes accurate and efficient detection of the aging state of the bushing.

[0146] The various embodiments in this specification are described in a progressive manner. For directly identical or similar parts of the embodiments, refer to each other. Each embodiment focuses on its differences from other embodiments. In particular, the system embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions in the method embodiments. It should be noted that the technical features of the above embodiments can be combined arbitrarily. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as the combination of these technical features does not contradict each other, it should be considered within the scope of this specification.

[0147] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and substitutions can be made without departing from the technical principles of the present invention, and these improvements and substitutions should also be considered within the scope of protection of the present invention.

Claims

1. A method for detecting the aging state of a bushing based on terahertz imaging, characterized in that, The method comprises the following steps: Collecting terahertz time-domain transmission signals at different positions of a to-be-tested casing; Performing filtered back-projection processing on the terahertz time-domain transmission signals at different positions to obtain an aging detection image of the to-be-tested casing, wherein a composite filter is introduced in the filtered back-projection processing, and the composite filter is designed to perform frequency segment regulation and control on the terahertz time-domain transmission signals by adopting a three-section frequency response structure; Performing texture feature extraction and hierarchical identification on the aging detection image to obtain an aging state detection result of the to-be-tested casing; The filtered back-projection processing on the terahertz time-domain transmission signals at different positions to obtain the aging detection image of the to-be-tested casing comprises the following steps: Performing frequency domain conversion on the terahertz time-domain transmission signals at different positions to obtain frequency domain data at corresponding positions; Performing filtering processing on the frequency domain data at each position by using the composite filter respectively to obtain filtered frequency domain data at the corresponding positions; Performing back-projection reconstruction on the filtered frequency domain data at each position respectively to obtain the aging detection image of the to-be-tested casing; The filtering processing on the frequency domain data at each position by using the composite filter respectively to obtain filtered frequency domain data at the corresponding positions comprises the following steps: Based on the three-section frequency response structure comprising a low-frequency suppression zone, a main passband reinforcement zone and a high-frequency attenuation compensation zone, a composite filter is constructed; Based on the composite filter, the frequency domain data at each position are filtered respectively to obtain filtered frequency domain data at the corresponding positions; The construction of the composite filter based on the three-section frequency response structure comprising a low-frequency suppression zone, a main passband reinforcement zone and a high-frequency attenuation compensation zone comprises the following steps: Based on a low-frequency truncation mechanism, a low-frequency suppression zone is constructed, and the low-frequency truncation mechanism is designed to set signal components lower than a noise critical frequency to zero; Based on an adaptive coherent compensation window function, a main passband reinforcement zone is constructed, and the adaptive coherent compensation window function is designed to perform artifact suppression on signal components between the noise critical frequency and a main passband cutoff frequency; Based on an exponential attenuation compensation function, a high-frequency attenuation compensation zone is constructed, and the exponential attenuation compensation function is designed to perform noise suppression and edge preservation on signal components higher than the main passband cutoff frequency; The low-frequency suppression zone, the main passband reinforcement zone and the high-frequency attenuation compensation zone are integrated to obtain a composite filter.

2. The method of detecting the aging state of a bushing based on terahertz imaging according to claim 1, characterized in that, The texture feature extraction and hierarchical identification on the aging detection image to obtain the aging state detection result of the to-be-tested casing comprise the following steps: Texture feature parameters are extracted from the aging detection image; The aging state detection result of the to-be-tested casing is obtained by performing hierarchical identification on the texture feature parameters.

3. The method of detecting the aging state of a bushing based on terahertz imaging according to claim 2, characterized by, The texture feature extraction from the aging detection image to obtain texture feature parameters comprises the following steps: Continuous wavelet transformation is performed on the aging detection image to obtain texture feature parameters; Wherein, in the continuous wavelet transform process, an adaptive weighting mechanism is introduced, which is designed to dynamically select the optimal wavelet basis function through dictionary learning, introduce a scale weighting function to respond to the target frequency band for enhancement, and perform directional texture enhancement and edge preservation filtering on the wavelet energy map after transformation. The texture feature parameters include texture contrast, spectral entropy, and wavelet detail coefficient variance.

4. The method of detecting the aging state of a bushing based on terahertz imaging according to claim 2, characterized by, The hierarchical identification of the texture feature parameters obtains the aging state detection result of the test sleeve. The texture feature parameters are input into a pre-constructed aging classification model for grade prediction to obtain the aging grade of the test sleeve, wherein the aging classification model is constructed based on a support vector machine.

5. A system for detecting the aging state of a bushing based on terahertz imaging, characterized by, It includes: A terahertz wave imaging device is used to collect terahertz time-domain transmission signals at different positions of the test sleeve and perform filter back-projection processing on the terahertz time-domain transmission signals at different positions to obtain an aging detection image of the test sleeve. During filter back-projection processing, a composite filter is introduced, which is designed to use a three-section frequency response structure to regulate and control the terahertz time-domain transmission signals in different frequency bands. An aging state evaluation module is used to extract texture features and perform hierarchical identification on the aging detection image to obtain the aging state detection result of the test sleeve. The filter back-projection processing of the terahertz time-domain transmission signals at different positions to obtain the aging detection image of the test sleeve includes: Frequency domain conversion is performed on the terahertz time-domain transmission signals at different positions to obtain frequency domain data at corresponding positions. The composite filter is used to filter the frequency domain data at each position respectively to obtain filtered frequency domain data at the corresponding position. The filtered frequency domain data at each position is back-projected and reconstructed respectively to obtain the aging detection image of the test sleeve. The composite filter is used to filter the frequency domain data at each position respectively to obtain filtered frequency domain data at the corresponding position, which includes: Based on a three-section frequency response structure including a low-frequency suppression zone, a main passband enhancement zone, and a high-frequency attenuation compensation zone, a composite filter is constructed. Based on the composite filter, the frequency domain data at each position are filtered respectively to obtain filtered frequency domain data at the corresponding position. Based on a three-section frequency response structure including a low-frequency suppression zone, a main passband enhancement zone, and a high-frequency attenuation compensation zone, a composite filter is constructed, which includes: Based on a low-frequency cutoff mechanism, a low-frequency suppression zone is constructed, which is designed to set the signal components below the noise critical frequency to zero. Based on an adaptive coherent compensation window function, a main passband enhancement zone is constructed, which is designed to suppress artifacts for signal components between the noise critical frequency and the main passband cutoff frequency. Based on an exponential attenuation compensation function, a high-frequency attenuation compensation zone is constructed, which is designed to suppress noise and preserve edges for signal components above the main passband cutoff frequency. The low-frequency suppression region, the main passband reinforcement region and the high-frequency attenuation compensation region are integrated to obtain a composite filter.

6. The terahertz imaging based jacket aging condition detection system of claim 5, wherein, The terahertz wave imaging device comprises a terahertz emission end, a terahertz detection end, a vector network analyzer and a data processing terminal. The terahertz emission end is used to emit modulated terahertz waves to penetrate a to-be-tested sleeve; The terahertz detection end is used to collect terahertz time-domain transmission signals at different positions of the to-be-tested sleeve; The vector network analyzer is used to simultaneously measure the terahertz waves and the terahertz time-domain transmission signals, and calculate S parameters based on a vector relationship between the terahertz waves and the terahertz time-domain transmission signals; The data processing terminal is used to perform filtered back-projection processing on the terahertz time-domain transmission signals at different positions to obtain an aging detection image of the to-be-tested sleeve.

7. The terahertz imaging based bushing aging condition detection system of claim 6, wherein, The terahertz wave imaging device further comprises a mechanical scanning device; The mechanical scanning device comprises a roller and a guide rail, the roller is used to support the to-be-tested sleeve and drive the to-be-tested sleeve to rotate 360 degrees, and the guide rail is used to drive the terahertz emission end and the terahertz detection end to translate along an axial direction of the to-be-tested sleeve.

Images