Quality detection method and device of carbon fiber shadowless medical line and computer equipment

Abstract

The invention relates to the technical field of quality detection, and discloses a quality detection method and device for a carbon fiber shadowless medical line and computer equipment. The method comprises the following steps: installing electrical connection terminals on an inner core conductor and a shielding wire of the carbon fiber shadowless medical wire, and fixing the carbon fiber shadowless medical wire on a test platform to obtain a sample to be tested; testing the conductivity of the to-be-tested sample to obtain a conductivity test data set, and testing the mechanical property of the to-be-tested sample to obtain a mechanical property test data set; performing feature extraction on the conductivity test data set to obtain conductivity feature parameters, and performing feature extraction on the mechanical performance test data set to obtain mechanical feature parameters; and inputting the conductive characteristic parameters and the mechanical characteristic parameters into a double-layer stacking quality detection model for quality detection, and outputting a quality detection result. According to the invention, the comprehensiveness and accuracy of quality detection of the carbon fiber shadowless medical line are effectively improved.

Description

Technical Field

[0001] The present invention relates to the technical field of quality inspection, and in particular to a quality inspection method, device and computer equipment for a carbon fiber shadowless medical wire. Background Art

[0002] As a key component of X-ray equipment, the quality of carbon fiber shadowless medical wire directly affects the accuracy and reliability of medical imaging diagnosis. Traditional carbon fiber shadowless medical wire quality inspection mainly relies on manual experience judgment, which is difficult to achieve standardization and normalization, and the inspection efficiency is low, which is prone to missed inspections and misjudgments.

[0003] In recent years, with the continuous development of medical equipment technology, the performance requirements of carbon fiber shadowless medical wires have been increasing, and the detection of their conductive and mechanical properties has become increasingly important. However, existing detection methods often separate conductive and mechanical properties and test them separately, ignoring the mutual influence and correlation between the two, making it difficult to ensure the comprehensiveness and accuracy of the test results. Summary of the invention

[0004] The present invention provides a method, a device and a computer equipment for quality inspection of a carbon fiber shadowless medical thread, which effectively improves the comprehensiveness and accuracy of quality inspection of the carbon fiber shadowless medical thread.

[0005] In a first aspect, the present invention provides a quality inspection method for a carbon fiber shadowless medical line, the quality inspection method for a carbon fiber shadowless medical line comprising:

[0006] Install electrical connection terminals on the inner core conductor and shielding wire of the carbon fiber shadowless medical wire, and fix the carbon fiber shadowless medical wire on a test platform to obtain a sample to be tested;

[0007] Performing a conductivity test on the sample to be tested to obtain a conductivity test data set, and performing a mechanical property test on the sample to be tested to obtain a mechanical property test data set;

[0008] Performing feature extraction on the conductive performance test data set to obtain conductive feature parameters, and performing feature extraction on the mechanical performance test data set to obtain mechanical feature parameters;

[0009] The conductive characteristic parameters and the mechanical characteristic parameters are input into a double-layer stack quality detection model for quality detection, and a quality detection result is output.

[0010] In a second aspect, the present invention provides a quality detection device for a carbon fiber shadowless medical line, the quality detection device for a carbon fiber shadowless medical line comprising:

[0011] An installation module is used to install electrical connection terminals on the inner core conductor and shielding wire of the carbon fiber shadowless medical wire, and fix the carbon fiber shadowless medical wire on a test platform to obtain a sample to be tested;

[0012] A testing module, used to perform a conductivity test on the sample to be tested to obtain a conductivity test data set, and to perform a mechanical property test on the sample to be tested to obtain a mechanical property test data set;

[0013] A feature extraction module, used to perform feature extraction on the conductive performance test data set to obtain conductive feature parameters, and to perform feature extraction on the mechanical performance test data set to obtain mechanical feature parameters;

[0014] The quality detection module is used to input the conductive characteristic parameters and the mechanical characteristic parameters into a double-layer stacking quality detection model for quality detection and output a quality detection result.

[0015] The third aspect of the present invention provides a quality inspection device for a carbon fiber shadowless medical line, comprising: a memory and at least one processor, wherein the memory stores instructions; the at least one processor calls the instructions in the memory so that the quality inspection device for the carbon fiber shadowless medical line performs the above-mentioned quality inspection method for the carbon fiber shadowless medical line.

[0016] In the technical solution provided by the present invention, a dual-branch feature extraction structure is designed to carry out targeted processing on the conductive properties and mechanical properties respectively, and the feature interaction layer is used to realize the deep fusion of the two types of performance characteristics, which effectively improves the comprehensiveness and accuracy of quality detection. A feature weighting method combining information entropy weight and coefficient of variation weight is adopted to reasonably allocate the importance of different feature parameters, overcoming the limitations of the traditional single weight method. The feature interaction layer design of the introduction of attention mechanism and residual connection enhances the information interaction between conductive features and mechanical features and improves the feature expression ability. Three parallel feature extraction branches are designed, and multi-scale feature extraction is realized through convolution kernels and pooling operations of different scales, which enhances the model's perception of features of different scales. Through the channel attention mechanism in the feature fusion branch, the importance of different feature channels is adaptively adjusted, which improves the accuracy and reliability of quality detection. The conductive performance test and mechanical performance test are standardized and automated, which significantly improves the detection efficiency and reduces the interference of human factors. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0018] Figure 1 A schematic flow chart of a quality inspection method for a carbon fiber shadowless medical line provided in an embodiment of the present application;

[0019] Figure 2 A schematic block diagram of the structure of a quality detection device for a carbon fiber shadowless medical line provided in an embodiment of the present application;

[0020] Figure 3 A schematic block diagram of the structure of a quality inspection device for a carbon fiber shadowless medical wire provided in an embodiment of the present application. DETAILED DESCRIPTION

[0021] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0022] The flowcharts shown in the accompanying drawings are only examples and do not necessarily include all the contents and operations / steps, nor must they be executed in the order described. For example, some operations / steps may also be decomposed, combined or partially merged, so the actual execution order may change based on actual conditions.

[0023] It should also be understood that the terms used in this application specification are only for the purpose of describing specific embodiments and are not intended to limit the application. As used in this application specification and the appended claims, unless the context clearly indicates otherwise, the singular forms "a", "an" and "the" are intended to include plural forms.

[0024] It should be further understood that the term “and / or” used in the specification and appended claims refers to any combination and all possible combinations of one or more of the associated listed items, and includes these combinations.

[0025] In conjunction with the accompanying drawings, some embodiments of the present application are described in detail below. In the absence of conflict, the following embodiments and features in the embodiments can be combined with each other.

[0026] See also Figure 1 , Figure 1A flow chart of a quality inspection method for a carbon fiber shadowless medical line provided in an embodiment of the present application is shown in FIG. Figure 1 As shown, the quality inspection method of the carbon fiber shadowless medical wire provided in the embodiment of the present application includes steps S100 to S400.

[0027] Step S100, installing electrical connection terminals on the inner core conductor and shielding wire of the carbon fiber shadowless medical wire, and fixing the carbon fiber shadowless medical wire on a test platform to obtain a sample to be tested;

[0028] It is understandable that the execution subject of the present invention may be a quality detection device for carbon fiber shadowless medical wire, or a terminal or a server, which is not limited here. The embodiment of the present invention is described by taking a server as the execution subject as an example.

[0029] Specifically, the inner core conductor of the carbon fiber shadowless medical line is connected to the first connection part, and the shielding wire is connected to the second connection part to form an initial connection structure. In this process, the physical contact between the inner core conductor and the first connection part is ensured to be reliable to reduce the contact resistance, and the shielding wire is tightly combined with the second connection part to ensure the electrical continuity of the shielding layer and the electromagnetic interference shielding effect. The first connection part is gold-plated to improve its conductivity and oxidation resistance so that it can maintain a stable electrical connection during long-term use, and the thickness and uniformity of the gold-plated layer need to be strictly controlled to ensure that the diameter of the first connection part can match the diameter of the inner core conductor to form a high-quality inner core conductor connection structure. At the same time, the second connection part is treated with a silver-plated copper mesh structure so that it can accurately match the shielding wire braided layer of the carbon fiber shadowless medical line to optimize the electromagnetic shielding effect, improve the stability of high-frequency signal transmission, and form a shielding wire connection structure. The inner core conductor connection structure and the shielding wire connection structure are fixed on the test platform, and the temperature and relative humidity of the test platform are set to ensure that environmental factors do not have unnecessary effects on the test results. During this process, environmental parameters such as temperature stability and humidity uniformity are precisely controlled to reduce resistance drift or impedance instability caused by environmental changes. The test platform has the ability to adjust the environment to maintain the required standard test conditions and obtain an environmentally adjusted test platform. This platform can improve the repeatability of the test and effectively simulate the performance of the carbon fiber shadowless medical line in the actual use environment to ensure that the obtained data has higher credibility. After completing the environmental adjustment, a four-wire resistance measurement system is installed on the test platform to build a complete resistance measurement structure. Compared with the traditional two-wire measurement method, the four-wire measurement system can effectively eliminate the influence of contact resistance and wire resistance and obtain more accurate resistance data. The measurement system includes a high-precision current source and voltage measurement equipment, and the stability of all connection points needs to be ensured to prevent the generation of measurement errors. A high-frequency impedance analyzer and a network analyzer are installed on the environmentally adjusted test platform to test the high-frequency electrical characteristics of the carbon fiber shadowless medical line. In order to ensure the accuracy of the test, the measurement ports of the high-frequency impedance analyzer and the network analyzer are connected to the resistance measurement structure using standard RF cables to ensure the stability of the signal path and avoid additional attenuation or noise interference. At the same time, the network analyzer measures S parameters, especially insertion loss (S21) and reflection loss (S11), to evaluate the high-frequency transmission performance of the carbon fiber shadowless medical cable. In order to simulate the performance changes of the carbon fiber shadowless medical cable in long-term use, the temperature of the aging test box is set, and the sample is subjected to aging test to obtain the sample to be tested.

[0030] Step S200, conducting a conductivity test on the sample to be tested to obtain a conductivity test data set, and conducting a mechanical property test on the sample to be tested to obtain a mechanical property test data set;

[0031] Specifically, during the conductivity test, the four-wire resistance measurement technology is used to eliminate the influence of contact resistance and wire resistance, and obtain high-precision temperature-resistance correspondence data. During the measurement process, the sample is placed in different temperature environments, its resistance value is measured and recorded, and the relationship between temperature and resistance is established for subsequent analysis of its conductive stability and environmental adaptability. In order to evaluate the signal transmission capability of the shielded wire, the transmission loss is measured, and the transmission loss parameter S21 and crosstalk parameter S41 of the shielded wire are measured using a high-frequency impedance analyzer to obtain the average transmission loss data, thereby evaluating the shielding layer's ability to suppress external signals and signal integrity. Electromagnetic shielding tests are performed on the samples to be tested to evaluate their anti-electromagnetic interference capabilities. During the test, a network analyzer is used to perform a frequency sweep test, and its electromagnetic shielding effectiveness is measured to obtain frequency-attenuation curve data. By analyzing the attenuation at different frequencies, it is determined whether the shielding effect of the carbon fiber shadowless medical line in high-frequency application scenarios meets the requirements. In order to analyze the long-term conductivity stability of carbon fiber shadowless medical wire, the samples to be tested were placed in an aging test box and aged under high temperature, high humidity, oxidation and other environmental conditions. The conductivity parameters were measured regularly during the aging process to obtain the aging conductivity data. The measured temperature-resistance correspondence data, transmission loss average data, frequency-attenuation curve data and aging conductivity data were integrated to form a complete conductivity test data set. Mechanical properties tests were performed on the samples to evaluate their durability and structural reliability. Tensile tests were performed on the samples to be tested at a preset loading rate to determine their stress-strain characteristics under tension, and stress-strain data were recorded to analyze their tensile strength and ductility. According to the preset bending radius, the samples were subjected to reciprocating bending tests to simulate repeated bending in actual applications, and the number of bending times was measured to evaluate their flexibility and durability. In order to analyze the structural integrity and mechanical stress resistance of the samples, the samples were subjected to torsion tests at a preset torsion speed, and the number of torsion turns was recorded to determine their stability and critical point of failure under torsion load. The samples to be tested are subjected to alternating bending fatigue tests to evaluate their reliability under long-term repeated loads. During the test, alternating bending loads are continuously applied and load-displacement data are recorded to analyze the material fatigue characteristics and fracture trends of the samples under long-term use. The stress-strain data, bending number data, torsion circle data, and load-displacement data are integrated to form a mechanical performance test data set for subsequent quality assessment and product optimization.

[0032] Step S300, performing feature extraction on the conductive performance test data set to obtain conductive feature parameters, and performing feature extraction on the mechanical performance test data set to obtain mechanical feature parameters;

[0033] Specifically, feature extraction is performed on the conductive performance test data set and the mechanical performance test data set to obtain conductive feature parameters and mechanical feature parameters. In the process of conductive performance feature extraction, linear regression is performed on the temperature-resistance correspondence data, and the resistance temperature coefficient is obtained by calculating the slope of the resistance-temperature curve. This parameter characterizes the conductive stability of the carbon fiber shadowless medical line under different temperature environments. At the same time, in order to measure the amplitude of its resistance change, the maximum and minimum value ratio of the temperature-resistance correspondence data is calculated to obtain the resistance stability coefficient, and quantify the fluctuation of the resistance value when the temperature changes. In order to analyze the signal transmission capability of the shielded line, the transmission loss average value data is arithmetic averaged to obtain the transmission loss mean, which reflects the attenuation degree of signal energy during transmission. At the same time, the crosstalk parameter S41 is numerically calculated to obtain the crosstalk noise ratio, which measures the external interference intensity of the carbon fiber shadowless medical line during signal transmission. The frequency-attenuation curve data is integrated and averaged to obtain the electromagnetic shielding attenuation coefficient, which reflects the shielding effectiveness of the carbon fiber shadowless medical line in different frequency ranges, ensuring that it can effectively suppress electromagnetic interference during the use of medical equipment. At the same time, in order to evaluate its long-term stability, the change rate of the aging conductive performance data is calculated to obtain the conductive performance aging rate, so as to analyze the resistance change trend of the sample in the aging test. By integrating the resistance temperature coefficient, resistance stability coefficient, transmission loss mean, crosstalk noise ratio, electromagnetic shielding attenuation coefficient and conductive performance aging rate, a conductive characteristic parameter set is formed. In the process of mechanical performance feature extraction, the slope of the linear segment of the stress-strain data is calculated to obtain the Young's modulus, which reflects the rigidity and elastic modulus of the carbon fiber shadowless medical line and provides its deformation characteristics under mechanical load. At the same time, in order to evaluate its mechanical strength, the maximum value of the stress-strain data is extracted to obtain the breaking strength and breaking elongation, which measure the tensile strength and deformation capacity of the carbon fiber shadowless medical line under extreme stress conditions. In order to analyze the bending durability, the bending number data is standardized to obtain the bending life coefficient, which is used to characterize the durability of the carbon fiber shadowless medical line under repeated bending. At the same time, the torsion circle data is standardized to obtain the torsion strength coefficient, which quantifies its structural stability and anti-destruction ability under torsion stress. In fatigue life analysis, the cumulative damage of load-displacement data is calculated to obtain fatigue life parameters, which can predict the service life of carbon fiber shadowless medical wires under long-term alternating loads. By integrating Young's modulus, breaking strength, breaking elongation, bending life coefficient, torsional strength coefficient and fatigue life parameters, a set of mechanical characteristic parameters is formed.

[0034] Step S400: input the conductive characteristic parameters and the mechanical characteristic parameters into a double-layer stack quality inspection model for quality inspection, and output the quality inspection result.

[0035] Specifically, the information entropy weight and the coefficient of variation weight of the conductive feature parameters are calculated to quantify the importance of each feature in the overall conductive performance evaluation, and the conductive feature comprehensive weight coefficient is obtained. At the same time, the mechanical feature parameters are calculated in the same way to evaluate the importance of each mechanical feature to the overall mechanical performance and obtain the mechanical feature comprehensive weight coefficient. According to the conductive feature comprehensive weight coefficient, the conductive feature parameters are weighted and combined to construct a conductive feature fusion vector. At the same time, according to the mechanical feature comprehensive weight coefficient, the mechanical feature parameters are weighted and combined to form a mechanical feature fusion vector. The conductive feature fusion vector is input into the conductive feature branch of the first neural network in the double-layer stacking quality detection model. The branch consists of a fully connected layer, a batch normalization layer, and a ReLU activation function layer, so that the input feature data is normalized in the neural network, and the nonlinear expression ability is enhanced by the ReLU activation function, so as to learn more complex feature representations. At the same time, the mechanical feature fusion vector is input into the mechanical feature branch of the first neural network, which also contains a fully connected layer, a batch normalization layer, and a ReLU activation function layer to ensure that the mechanical features are fully characterized in the network and the most discriminative information is extracted. After this stage of processing, the conductive features and mechanical features are mapped to high-dimensional space, and the conductive feature representation vector and the mechanical feature representation vector are generated. In the first neural network, a feature interaction layer is introduced to enhance the synergistic relationship between the conductive features and the mechanical features. In the feature interaction layer, the conductive feature representation vector and the mechanical feature representation vector are weighted by the attention mechanism and residual connection processing to generate an interactive feature vector. The attention mechanism dynamically adjusts the feature weights by calculating the correlation between different features, so that the key features receive greater attention in the decision-making process, and the addition of residual connections ensures that the model will not suffer performance degradation due to gradient vanishing or gradient explosion during deep learning, so that information flows smoothly in the neural network. The interactive feature vector and the conductive feature representation vector and the mechanical feature representation vector are reorganized to construct a multi-scale feature vector. Using information representation methods of different scales, feature information of different levels is extracted, so that the model can learn macroscopic features, such as overall resistance change trend, mechanical strength stability, etc., and can also capture microscopic details, such as high-frequency transmission loss changes, microstructural strain, etc. The multi-scale feature vector is input into the second neural network in the double-layer stacking quality inspection model for quality inspection analysis. The second neural network includes three parallel feature extraction branches and one feature fusion branch. The three feature extraction branches are used to extract conductive features, mechanical features, and interactive features from different angles, respectively, so that the model comprehensively considers the feature information of different modes, while the feature fusion branch integrates the outputs of each branch to ultimately form a comprehensive quality evaluation index. Through this process, the model accurately evaluates the overall quality of the carbon fiber shadowless medical line and ultimately outputs the quality inspection results to determine whether it meets the medical use standards.

[0036] The conductive feature representation vector is self-attention calculated, and the self-attention mechanism is used to capture the long-range dependencies and key feature weights within the conductive feature to obtain the conductive feature attention weight matrix. At the same time, the mechanical feature representation vector is self-attention calculated to mine the feature correlation within the mechanical feature and generate the mechanical feature attention weight matrix. Through this process, the model can automatically pay attention to the important information in the conductive and mechanical features, and assign corresponding weights to different features to enhance the accuracy of feature expression. The conductive feature representation vector and the mechanical feature representation vector are cross-attention calculated to establish an interactive relationship between the two and obtain the cross-feature attention weight matrix. The cross-attention calculation captures the mutual information correlation between the conductive feature and the mechanical feature, so that the model can automatically learn the coupling mode between the two types of features, ensuring that the mutual influence of the conductive performance and the mechanical performance is comprehensively considered in the quality inspection process. Based on the cross-feature attention weight matrix, a weighted summation operation is performed to obtain a weighted cross-feature vector, which contains the individual information of the conductive and mechanical features and integrates the interactive characteristics between the two, so that the model can make full use of the key information of different modalities in the decision-making process. In order to optimize the feature expression, the conductive feature representation vector is subjected to 1×1 convolution to reduce the feature dimension and retain local information to obtain the conductive feature residual vector. At the same time, the mechanical feature representation vector is subjected to the same 1×1 convolution to obtain the mechanical feature residual vector. The convolution operation enhances the local representation ability of the feature and reduces the amount of calculation, so that the model can efficiently extract useful information. The conductive feature attention weight matrix is ​​multiplied with the conductive feature residual vector to obtain the weighted conductive feature vector to ensure that the weights assigned by the attention mechanism can effectively adjust the expression of the original feature information. At the same time, the mechanical feature attention weight matrix is ​​multiplied with the mechanical feature residual vector to obtain the weighted mechanical feature vector, thereby optimizing the attention weighting effect of the mechanical feature. The weighted conductive feature vector, the weighted mechanical feature vector and the weighted cross feature vector are feature spliced ​​to form a spliced ​​feature vector. The spliced ​​feature vector is transformed by a fully connected layer to learn the high-dimensional feature expression, and the stability and generalization ability of the feature are improved by normalization to ensure that the model can better adapt to the characteristics of different carbon fiber shadowless medical wire samples, and finally generate an interactive feature vector.

[0037] The multi-scale feature vector is input into the first feature extraction branch of the second neural network, which uses a 3×3 convolution operation to extract local spatial features and combines the maximum pooling process to reduce the feature dimension and enhance the robustness of the model to obtain the first feature map. At the same time, the multi-scale feature vector is input into the second feature extraction branch of the second neural network, which uses a 5×5 convolution operation to capture a wider range of local features and combines the average pooling process to smooth the feature map and reduce noise to generate the second feature map. The multi-scale feature vector is input into the third feature extraction branch, which uses a 7×7 convolution to obtain a wider range of feature patterns and combines the adaptive pooling process to ensure the stability of the features under different input sizes to obtain the third feature map. Through these three feature extraction branches of different scales, the model can capture different levels of information from local details to global patterns. Channel attention calculation is performed on the first feature map to learn the importance of its different channels and generate the first channel weight vector. At the same time, the same channel attention calculation is performed on the second feature map to obtain the second channel weight vector. The third feature map is calculated by channel attention to extract the third channel weight vector. The channel attention mechanism can adaptively adjust the weight of each channel, so that the network pays more attention to key features, while reducing redundant information and improving the effectiveness and discrimination ability of features. The first channel weight vector is weighted with the first feature map to highlight important channel information and obtain a weighted first feature map. Similarly, the second channel weight vector is weighted with the second feature map to optimize feature expression and generate a weighted second feature map. The third channel weight vector is weighted with the third feature map to obtain a weighted third feature map. The weighted first feature map, the weighted second feature map, and the weighted third feature map are input into the feature fusion branch to integrate multi-scale information. In the feature fusion branch, the fully connected layer is processed so that features of different scales can be fused in the same feature space to obtain a fused feature vector. In this process, the fully connected layer effectively integrates the feature information of different branches and improves the discrimination ability of features through nonlinear transformation, so that the final fused feature vector retains local detail information and has the ability to express global patterns. The fused feature vector is input into the Softmax classification layer to calculate the probability distribution of quality categories, and the quality of the carbon fiber shadowless medical line is classified and judged according to the probability distribution. Softmax processing can normalize the output of the model into probability values, so that the degree of belonging of each sample in different quality categories can be clearly expressed, and the quality category probability distribution is compared with the preset quality classification threshold to determine the final quality test result. Through this process, it is possible to identify whether the carbon fiber shadowless medical line meets the quality standards and give a specific quality grade.

[0038] In the embodiment of the present invention, by designing a dual-branch feature extraction structure, the conductive properties and mechanical properties are processed in a targeted manner respectively, and the feature interaction layer is used to realize the deep fusion of the two types of performance features, which effectively improves the comprehensiveness and accuracy of quality detection. A feature weighting method combining information entropy weight and coefficient of variation weight is adopted to reasonably allocate the importance of different feature parameters, overcoming the limitations of the traditional single weight method. The feature interaction layer design of the introduction of attention mechanism and residual connection enhances the information interaction between conductive features and mechanical features and improves the feature expression ability. Three parallel feature extraction branches are designed, and multi-scale feature extraction is realized through convolution kernels and pooling operations of different scales, which enhances the model's perception of features of different scales. Through the channel attention mechanism in the feature fusion branch, the importance of different feature channels is adaptively adjusted, which improves the accuracy and reliability of quality detection. The conductive performance test and mechanical performance test are standardized and automated, which significantly improves the detection efficiency and reduces the interference of human factors.

[0039] In a specific embodiment, the process of executing step S100 may specifically include the following steps:

[0040] Connecting the inner core conductor of the carbon fiber shadowless medical wire to the first connection part, and connecting the shielding wire of the carbon fiber shadowless medical wire to the second connection part to obtain an initial connection structure;

[0041] The first connection part is subjected to gold plating treatment so that the diameter of the first connection part matches the diameter of the inner core conductor to obtain an inner core conductor connection structure, and the second connection part is subjected to silver-plated copper mesh structure treatment so that the second connection part matches the braided layer of the shielding wire to obtain a shielding wire connection structure;

[0042] The inner core conductor connection structure and the shielding wire connection structure are fixed on the test platform, the temperature and relative humidity of the test platform are set to obtain the test platform after environmental adjustment, and a four-wire resistance measurement system is installed on the test platform after environmental adjustment to obtain a resistance measurement structure;

[0043] A high-frequency impedance analyzer and a network analyzer are installed on the test platform after environmental adjustment, the measurement ports of the high-frequency impedance analyzer and the network analyzer are connected to the resistance measurement structure through standard radio frequency cables, and the temperature of the aging test box is set to obtain the sample to be tested.

[0044] Specifically, the inner core conductor of the carbon fiber shadowless medical line is connected to the first connection part, and the shielding wire is connected to the second connection part to form an initial connection structure. The inner core conductor is the main conductive part for signal transmission, while the shielding wire is used to reduce external electromagnetic interference and ensure signal stability. The first connection part is gold-plated to make its surface have excellent conductivity and anti-oxidation ability, while ensuring that the diameter d of the first connection part is 1The diameter d of the inner core conductor c Match, that is, satisfy the relationship:

[0045] d 1 =d c ;

[0046] Among them, d 1 represents the final diameter of the gold-plated connection, while d c Represents the diameter of the inner core conductor. Through this matching operation, the contact resistance of the connection part is ensured to be minimum, thereby improving the stability of current transmission. For the connection of the shielded wire, a silver-plated copper mesh structure is used to enhance the conductivity and ensure that the second connection part is accurately matched with the braided layer of the shielded wire. If the braided layer diameter of the shielded wire is d s , then the diameter d of the second connecting part 2 Need to meet:

[0047] d 2 =d s ;

[0048] At the same time, the thickness t of the silver-plated copper mesh needs to be controlled within a reasonable range to ensure the shielding effectiveness S E Achieve the expected goal, S E The calculation is as follows:

[0049]

[0050] Among them, E i Represents the external electromagnetic field strength, E t Indicates the electromagnetic field strength after passing through the shielding layer. By properly controlling the silver plating thickness, ensure that the shielding layer effectively reduces high-frequency signal interference, thereby improving signal integrity. After completing the connection between the inner core conductor and the shielding wire, fix the inner core conductor connection structure and the shielding wire connection structure to the test platform to ensure that the sample will not deform or have poor contact during the test. At the same time, set the temperature T of the test platform p and relative humidity H p , to simulate the use environment of medical equipment and ensure the reliability of test data. Install a four-wire resistance measurement system to measure the conductive properties of carbon fiber shadowless medical wire. The resistance calculation formula measured by the four-wire method is as follows:

[0051]

[0052] Among them, R represents the measured resistance, V represents the measured voltage, and I represents the measured current. Compared with the traditional two-wire measurement method, the four-wire method can effectively eliminate the influence of wire and contact resistance, thereby improving the measurement accuracy. A high-frequency impedance analyzer and a network analyzer are installed on the environmentally conditioned test platform to analyze the characteristics of the carbon fiber shadowless medical line during high-frequency signal transmission. The high-frequency impedance analyzer is used to measure the AC impedance Z(f) of the sample, and its calculation formula is:

[0053] Z(f)=R+jX;

[0054] Where R represents the AC resistance component, X represents the reactance component, and j is an imaginary unit. By analyzing the change of impedance with frequency f, the high-frequency performance of the carbon fiber shadowless medical cable is evaluated. At the same time, the network analyzer is used to measure the scattering parameters (S parameters), especially the insertion loss S 21 and reflection loss S 11 Insertion loss describes the energy lost by the signal after passing through the sample, and the calculation formula is

[0055]

[0056] Among them, P in is the input signal power, P out is the output signal power. Reflection loss is used to measure the return loss of the signal, and its calculation formula is:

[0057] S 11 =20log(|Γ|);

[0058] Where Γ represents the reflection coefficient and is defined as:

[0059]

[0060] Among them, Z L is the load impedance of the carbon fiber shadowless medical cable, Z 0 is the system characteristic impedance, usually 50Ω. In order to analyze the long-term stability of the carbon fiber shadowless medical cable, the temperature T of the aging test chamber is set. a , and conduct aging tests on samples to simulate the performance changes of cables during long-term use due to factors such as temperature, humidity, and mechanical stress. The aging temperature is set between specific temperatures, and a long-term constant temperature and humidity experiment is conducted to monitor the changes in its conductive properties over time t. The resistance change rate ΔR is calculated by the following formula:

[0061]

[0062] Among them, R t Represents the resistance at a certain moment after aging, R 0Represents the initial resistance value. If ΔR exceeds the set threshold, it means that the conductivity of the sample has significantly deteriorated under the aging environment, thus affecting its reliability in actual medical applications.

[0063] In a specific embodiment, the process of executing step S200 may specifically include the following steps:

[0064] Perform four-wire resistance measurement on the sample to be tested to obtain temperature-resistance correspondence data;

[0065] The transmission loss of the sample to be tested is measured, and the transmission loss parameter S21 and the crosstalk parameter S41 of the shielded line are measured using a high-frequency impedance analyzer to obtain the average transmission loss data;

[0066] Conduct electromagnetic shielding test on the sample to be tested, use network analyzer to perform frequency sweep test and electromagnetic shielding effectiveness measurement, and obtain frequency-attenuation curve data;

[0067] The sample to be tested is placed in an aging test chamber for aging treatment and conductivity parameter measurement to obtain aging conductivity data, and the temperature-resistance correspondence data, transmission loss average data, frequency-attenuation curve data and aging conductivity data are combined into a conductivity test data set;

[0068] Perform a tensile test on the sample to be tested according to a preset loading rate, perform a reciprocating bending test on the sample to be tested according to a preset bending radius, and perform a torsion test on the sample to be tested according to a preset torsion speed to obtain stress-strain data, bending number data, and torsion circle number data;

[0069] The sample under test is subjected to an alternating bending fatigue test, and the load-displacement data is recorded. The stress-strain data, bending number data, torsion circle number data, and load-displacement data are combined into a mechanical property test data set.

[0070] Specifically, the conductive properties of the carbon fiber shadowless medical wire were measured, including four-wire resistance measurement, high-frequency transmission loss measurement, electromagnetic shielding test and aging test. Four-wire resistance measurement is a method for accurately measuring low resistance. It eliminates the influence of contact resistance and wire resistance on the measurement results by introducing probes for current and voltage measurement respectively, and obtains high-precision temperature-resistance correspondence data. The resistance calculation formula measured by the four-wire method is as follows:

[0071]

[0072] Where R represents the measured resistance, V is the measured voltage, and I is the measured current through the sample. In order to study the relationship between the change of resistance and temperature, measurements are performed under different temperature conditions, and a mathematical relationship between temperature T and resistance R is established, such as a linear relationship:

[0073] R(T)=R 0 (1+α(TT 0 ));

[0074] Among them, R 0 is the initial temperature T 0 The resistance value under the condition of α is the temperature coefficient of resistance, which is used to characterize the resistance stability of the carbon fiber shadowless medical cable when the temperature changes. In order to evaluate the signal transmission capability of the carbon fiber shadowless medical cable, a high-frequency impedance analyzer is used to measure the transmission loss parameter S of the shielded cable. 21 and crosstalk parameter S 41 , and calculate the average value of transmission loss. The definition of transmission loss is as follows:

[0075]

[0076] Among them, P in is the input signal power, P out is the output signal power. Crosstalk parameter S 41 Reflects the shielded cable's ability to resist interference from external signals. The calculation formula is:

[0077]

[0078] Among them, V induced is the induced voltage due to crosstalk, V input is the input signal voltage. By measuring S at different frequencies 21 and S 41 , calculate the average value of transmission loss, and thus evaluate the shielding effectiveness. In order to analyze the electromagnetic compatibility of carbon fiber shadowless medical wire, a network analyzer was used to perform a frequency sweep test and measure its electromagnetic shielding effectiveness S E , and its calculation formula is:

[0079]

[0080] Among them, E incident is the incident electromagnetic wave field strength, E transmitted It is the electromagnetic wave field strength after transmission. By measuring the frequency-attenuation curve data, it is determined whether the shielding performance of the carbon fiber shadowless medical cable in different frequency ranges meets the requirements of medical equipment. The carbon fiber shadowless medical cable is placed in an aging test chamber and the temperature is set to T a and humidity H a , and regularly measure its conductive performance parameters, such as the resistance change rate ΔR calculated as follows:

[0081]

[0082] Among them, R t Represents the resistance at a certain moment after aging, R0 Represents the initial resistance value. The temperature-resistance correspondence data, transmission loss average data, frequency-attenuation curve data, and aging conductivity data are combined into a complete conductivity test data set. In the mechanical performance test section, a tensile test is performed at a preset loading rate v to obtain stress-strain data. The tensile stress σ is calculated as follows:

[0083]

[0084] Where F is the applied tensile force and A is the cross-sectional area of ​​the sample. The strain ε is calculated as follows:

[0085]

[0086] Where ΔL is the elongation, L 0 is the initial length. In order to evaluate the flexibility, a reciprocating bending test is performed according to the preset bending radius r, and the number of bends N is recorded. b Until failure occurs, perform torsion test at preset torsion speed ω and measure the number of torsion turns N t , its torque M is calculated as follows:

[0087] M = G·θ;

[0088] Where G is the shear modulus and θ is the torsion angle. In order to analyze the fatigue life of the carbon fiber shadowless medical line, an alternating bending fatigue test was performed and the load-displacement data were recorded. Fatigue life N f The calculation is as follows:

[0089]

[0090] Among them, σ max is the maximum stress, σ 0 is the material parameter, and b is the material fatigue index. The stress-strain data, bending number data, torsion circle number data, and load-displacement data are integrated to form a mechanical performance test data set.

[0091] In a specific embodiment, the process of executing step S300 may specifically include the following steps:

[0092] Perform linear regression on the temperature-resistance correspondence data in the conductive performance test data set, calculate the slope of the resistance-temperature curve, obtain the resistance temperature coefficient, and calculate the maximum and minimum value ratio of the temperature-resistance correspondence data to obtain the resistance stability coefficient;

[0093] Perform arithmetic averaging on the transmission loss average value data in the conductive performance test data set to obtain the transmission loss mean value, and perform numerical calculation on the crosstalk parameter S41 to obtain the crosstalk-to-noise ratio;

[0094] The frequency-attenuation curve data in the conductive performance test data set are integrated and averaged to obtain the electromagnetic shielding attenuation coefficient, and the change rate of the aged conductive performance data is calculated to obtain the conductive performance aging rate, and the resistance temperature coefficient, resistance stability coefficient, transmission loss mean, crosstalk noise ratio, electromagnetic shielding attenuation coefficient and conductive performance aging rate are combined into conductive characteristic parameters;

[0095] The slope of the linear segment of the stress-strain data in the mechanical properties test data set is calculated to obtain the Young's modulus, and the maximum value of the stress-strain data is extracted to obtain the breaking strength and breaking elongation;

[0096] The bending times data in the mechanical performance test data set is standardized to obtain the bending life coefficient, and the torsion turns data is standardized to obtain the torsion strength coefficient;

[0097] The cumulative damage of the load-displacement data is calculated to obtain the fatigue life parameters, and Young's modulus, fracture strength, fracture elongation, bending life coefficient, torsional strength coefficient and fatigue life parameters are combined into mechanical characteristic parameters.

[0098] Specifically, linear regression is performed on the temperature-resistance correspondence data to obtain the slope of the resistance-temperature curve, thereby calculating the resistance temperature coefficient. Assuming that there is a linear relationship between resistance R and temperature T, its mathematical expression is expressed as:

[0099] R(T)=R 0 +αT;

[0100] Where R(T) represents the resistance value at temperature T, R 0 is the initial resistance, α represents the temperature coefficient of resistance, and this parameter is solved by the least squares method:

[0101]

[0102] In order to measure the stability of the resistor, the ratio of the maximum and minimum values ​​in the temperature-resistance correspondence data is calculated, that is, the resistance stability coefficient S R :

[0103]

[0104] Among them, R max and R min Represent the maximum and minimum resistance values ​​measured respectively. Perform arithmetic averaging on the transmission loss measurement data in the conductive performance test data set to obtain the transmission loss mean value. The calculation formula is as follows:

[0105]

[0106] Where N is the total number of measurement points, S21,i is the i-th group of measurement data. In order to analyze the crosstalk situation, calculate the crosstalk parameter S 41 The numerical value is calculated as follows:

[0107]

[0108] In order to determine the shielding effectiveness of carbon fiber shadowless medical wire, the electromagnetic shielding attenuation coefficient S is calculated. E , which is calculated as:

[0109]

[0110] Among them, f 1 and f 2 is the measurement frequency range, E incident and E transmitted are the incident and transmitted electromagnetic wave field strengths respectively. Calculate the conductivity aging rate ΔR:

[0111]

[0112] Among them, R t Represents the resistance after aging, R 0 Represents the initial resistance value. In terms of mechanical properties, the slope of the linear segment of the stress-strain curve is calculated to obtain the Young's modulus E:

[0113]

[0114] Where Δσ is the stress increment and Δε is the strain increment. Fracture strength σ f The maximum stress is calculated as:

[0115] σ f =max(σ);

[0116] Elongation at breakε f The maximum strain is calculated as:

[0117] ε f =max(ε);

[0118] In order to analyze the bending performance, the bending number data is standardized to obtain the bending life coefficient S b :

[0119]

[0120] Similarly, the torsional strength coefficient S t Calculated by normalizing the torsion circle data:

[0121]

[0122] Calculate fatigue life parameter Nf , used to evaluate the durability of carbon fiber shadowless medical wires under alternating stress, the calculation formula is:

[0123]

[0124] Among them, σ max is the maximum stress, σ 0 and b are material related parameters. Young's modulus, fracture strength, fracture elongation, bending life coefficient, torsional strength coefficient and fatigue life parameters are combined into mechanical characteristic parameters.

[0125] Before the conductive characteristic parameters and mechanical characteristic parameters are input into the double-layer stacking quality detection model, the method also includes: segmenting the conductive characteristic parameters according to the working cycle of the X-ray equipment to obtain a periodic conductive characteristic sequence, performing a two-dimensional convolution operation on the periodic conductive characteristic sequence to obtain a conductive characteristic periodic pattern; performing long-term trend decomposition on the conductive characteristic parameters, extracting the aging characteristic change law, and obtaining the conductive characteristic long-term trend through full connection layer processing; performing short-term trend modeling based on the most recent N sampling points of the conductive characteristic parameters to obtain the conductive characteristic short-term trend; segmenting the mechanical characteristic parameters according to the stress loading cycle to obtain a periodic mechanical characteristic sequence, performing a two-dimensional convolution operation on the periodic mechanical characteristic sequence to obtain a mechanical characteristic periodic pattern; The long-term trend of mechanical characteristic parameters is decomposed to extract the variation law of fatigue characteristics, and the long-term trend of mechanical characteristics is obtained through full connection layer processing; short-term trend modeling is performed based on the latest M sampling points of mechanical characteristic parameters to obtain the short-term trend of mechanical characteristics; the conductive characteristic periodic pattern, the conductive characteristic long-term trend and the conductive characteristic short-term trend are adaptively fused to obtain optimized conductive characteristic parameters, and the mechanical characteristic periodic pattern, the mechanical characteristic long-term trend and the mechanical characteristic short-term trend are adaptively fused to obtain optimized mechanical characteristic parameters; characteristic feedback noise is generated based on the optimized conductive characteristic parameters and the optimized mechanical characteristic parameters, and noise compensation is performed on the optimized conductive characteristic parameters and the optimized mechanical characteristic parameters to obtain compensated conductive characteristic parameters and compensated mechanical characteristic parameters.

[0126] In a specific embodiment, the process of executing step S400 may specifically include the following steps:

[0127] The information entropy weight and the coefficient of variation weight are calculated for the conductive characteristic parameters to obtain the comprehensive weight coefficient of the conductive characteristic, and the information entropy weight and the coefficient of variation weight are calculated for the mechanical characteristic parameters to obtain the comprehensive weight coefficient of the mechanical characteristic;

[0128] Conductive feature parameters are weighted combined according to the comprehensive weight coefficient of the conductive feature to construct a conductive feature fusion vector, and mechanical feature parameters are weighted combined according to the comprehensive weight coefficient of the mechanical feature to construct a mechanical feature fusion vector;

[0129] Input the conductive feature fusion vector into the conductive feature branch of the first neural network in the double-layer stacking quality detection model, the conductive feature branch includes a fully connected layer, a batch normalization layer, and a ReLU activation function layer, to obtain a conductive feature representation vector; input the mechanical feature fusion vector into the mechanical feature branch of the first neural network in the double-layer stacking quality detection model, the mechanical feature branch includes a fully connected layer, a batch normalization layer, and a ReLU activation function layer, to obtain a mechanical feature representation vector;

[0130] In the double-layer stacking quality inspection model, a feature interaction layer is set in the first neural network, and the conductive feature representation vector and the mechanical feature representation vector are weighted by the attention mechanism and residual connection is processed to obtain an interaction feature vector;

[0131] Performing feature reorganization on the interaction feature vector, the conductive feature representation vector, and the mechanical feature representation vector to obtain a multi-scale feature vector;

[0132] The multi-scale feature vector is input into the second neural network in the double-layer stacking quality inspection model for quality inspection analysis. The second neural network includes three parallel feature extraction branches and one feature fusion branch, and outputs the quality inspection results.

[0133] Specifically, the information entropy weight calculation and the coefficient of variation weight calculation are performed on the conductive characteristic parameters to obtain the conductive characteristic comprehensive weight coefficient. The information entropy weight calculation is based on the information content of the feature, and its importance is determined by calculating the normalized entropy of each feature. Suppose the conductive characteristic parameter matrix is ​​X = [x ij ], where x ij represents the jth eigenvalue of the i-th sample, then the information entropy of the j-th feature is j Calculated by the following formula:

[0134]

[0135] in, is the normalization coefficient, p ij is the normalized probability distribution, defined as:

[0136]

[0137] The smaller the entropy value, the more important the feature is, so the information entropy weight w j Calculated as:

[0138]

[0139] At the same time, in order to enhance the feature discrimination, the coefficient of variation weight is calculated and defined as:

[0140]

[0141] Among them, σ j is the standard deviation of the jth feature, is the mean of the feature, and the final coefficient of variation weight is:

[0142]

[0143] Considering the information entropy weight and the variation coefficient weight, the conductive feature comprehensive weight coefficient W j Calculated as:

[0144] W j =λw j +(1-λ)v j ;

[0145] Among them, λ is an adjustment parameter used to control the contribution of the two weights. The same method is used to calculate the comprehensive weight coefficient of the mechanical feature. After obtaining the weight coefficient, the conductive feature parameters are weighted combined to construct the conductive feature fusion vector:

[0146]

[0147] Among them, x j Represents the normalized conductive characteristic parameter. Similarly, the mechanical characteristic fusion vector is calculated as follows:

[0148]

[0149] Among them, y j Represents the normalized mechanical characteristic parameter. The conductive feature fusion vector is input into the conductive feature branch of the first neural network of the double-layer stack quality detection model, which includes a fully connected layer, a batch normalization layer, and a ReLU activation function layer. Its mathematical expression is:

[0150] F ′ elec =ReLU(BN(W elec F elec +b elec ));

[0151] Among them, W elec is the weight matrix of the fully connected layer, b elec is the bias term, BN represents the batch normalization operation, and ReLU represents the activation function. Similarly, the mechanical feature representation vector is calculated as follows:

[0152] F ′ mech =ReLU(BN(W mech F mech +b mech ));

[0153] A feature interaction layer is set in the first neural network to perform attention mechanism weighted sum and residual connection processing on the conductive feature representation vector and the mechanical feature representation vector to obtain an interaction feature vector. Calculate the attention weight matrices of the conductive features and the mechanical features:

[0154]

[0155] where Q and K represent the query matrix and the key matrix respectively, and softmax ensures weight normalization. Calculate the weighted feature vectors based on the attention weights:

[0156]

[0157] Adopt residual connection:

[0158]

[0159] Construct the interaction feature vector through feature concatenation operation:

[0160]

[0161] Perform feature recombination on the interaction feature vector and the conductive and mechanical feature representation vectors to obtain multi-scale feature vectors:

[0162]

[0163] Input the multi-scale feature vectors into the second neural network of the double-layer stacked quality detection model, which includes three parallel feature extraction branches and one feature fusion branch. The first feature extraction branch uses 3×3 convolution and max pooling, the second uses 5×5 convolution and average pooling, and the third uses 7×7 convolution and adaptive pooling, and calculate respectively:

[0164] F conv3 =AdaptivePool(Conv3x3(F multi ));

[0165] F conv5 =AdaptivePool(Conv5x5(F multi ));

[0166] F conv7 =AdaptivePool(Conv7x7(F multi ));

[0167] Input all the extracted features into the feature fusion branch and fuse them through a fully connected layer:

[0168] F final =ReLU(BN(W final [F conv3 ,Fconv5 ,F conv7 ]+b final ));

[0169] Output quality test results:

[0170] y=Softmax(W out F final +b out ).

[0171] In a specific embodiment, the execution step sets a feature interaction layer in the first neural network in the double-layer stacking quality detection model, performs attention mechanism weighting and residual connection processing on the conductive feature representation vector and the mechanical feature representation vector, and the process of obtaining the interactive feature vector can specifically include the following steps:

[0172] Perform self-attention calculation on the conductive feature representation vector to obtain a conductive feature attention weight matrix, and perform self-attention calculation on the mechanical feature representation vector to obtain a mechanical feature attention weight matrix;

[0173] Performing cross-attention calculation on the conductive feature representation vector and the mechanical feature representation vector to obtain a cross-feature attention weight matrix, and performing a weighted summation operation based on the cross-feature attention weight matrix to obtain a weighted cross-feature vector;

[0174] Performing 1×1 convolution processing on the conductive feature representation vector to obtain a conductive feature residual vector, and performing 1×1 convolution processing on the mechanical feature representation vector to obtain a mechanical feature residual vector;

[0175] Performing a dot multiplication operation on the conductive feature attention weight matrix and the conductive feature residual vector to obtain a weighted conductive feature vector, and performing a dot multiplication operation on the mechanical feature attention weight matrix and the mechanical feature residual vector to obtain a weighted mechanical feature vector;

[0176] The weighted conductive feature vector, the weighted mechanical feature vector and the weighted cross feature vector are feature concatenated to obtain a concatenated feature vector, and the concatenated feature vector is subjected to a fully connected layer transformation and normalization to obtain an interactive feature vector.

[0177] Specifically, self-attention calculation is performed on the conductive feature representation vector and the mechanical feature representation vector to obtain their respective attention weight matrices, and cross-attention calculation is performed on this basis to extract the correlation information between the conductive feature and the mechanical feature. Assume that the conductive feature representation vector is F elec ∈R n×d , where n is the number of samples, d is the feature dimension, and the mechanical feature representation vector is F mech ∈R n×d, in order to calculate the respective self-attention weight matrices, the query matrix Q, key matrix K and value matrix V are constructed, which are defined as follows:

[0178] Q elec =W Q F eleC ,K ELEc =W K F elEC ,V ELEC =W V F elec ;

[0179] Q mech =W Q F mech ,K Mech =W K F mech ,V mech =W V F mECH ;

[0180] Among them, W q ,W k ,W v ∈R D×d is a trainable weight matrix. The attention weight matrix of the conductive and mechanical features is calculated through the self-attention mechanism, and Scaled Dot-Product Attention is used for calculation:

[0181]

[0182] Among them, the softmax operation ensures that the attention weights are normalized so that the sum of all attention values ​​is 1. The conductive feature representation vector and the mechanical feature representation vector are cross-attention calculated to obtain the cross-feature attention weight matrix:

[0183]

[0184] The weight matrix is ​​used to perform a weighted sum operation to obtain the weighted cross eigenvector:

[0185] F croSs =A cross V mech ;

[0186] This step ensures that the conductive feature can focus on the important information in the mechanical feature and adjusts the proportion of information transmission through the weight mechanism. On this basis, in order to enhance the local feature expression ability, the conductive feature representation vector and the mechanical feature representation vector are respectively subjected to 1×1 convolution processing to obtain the conductive feature residual vector and the mechanical feature residual vector, which are defined as follows:

[0187]

[0188] Among them, Conv1x1 represents a 1×1 convolution operation, which does not change the spatial dimension of the feature, but can adjust the channel dimension of the feature and enhance local information. The conductive feature attention weight matrix is ​​dot-multiplied with the conductive feature residual vector to obtain the weighted conductive feature vector:

[0189]

[0190] Similarly, the mechanical feature attention weight matrix is ​​multiplied with the mechanical feature residual vector to obtain the weighted mechanical feature vector:

[0191]

[0192] Among them, ⊙ represents element-by-element multiplication (Hadamard product), ensuring that the weighted adjustment of each feature channel can accurately reflect its attention weight. The weighted conductive feature vector, weighted mechanical feature vector and weighted cross feature vector are concatenated to obtain the concatenated feature vector:

[0193]

[0194] In order to improve the expressiveness of features, the concatenated feature vector is transformed with a fully connected layer to construct the final interactive feature vector, which is calculated as follows:

[0195] F final =ReLU(BN(WF concat + b));

[0196] Among them, W is the weight matrix of the fully connected layer, b is the bias term, BN represents the batch normalization operation, and ReLU is the nonlinear activation function, which ensures that the model can capture complex feature relationships and improve learning ability.

[0197] Among them, the conductive feature parameters and mechanical feature parameters are enhanced to obtain enhanced feature parameters, including: constructing a conductive feature task set for the conductive feature parameters, each task corresponds to the conductive feature distribution under different working conditions, and constructing a mechanical feature task set for the mechanical feature parameters, each task corresponds to the mechanical feature distribution under different stress conditions, to obtain a multi-task feature set; performing shared feature extraction on the multi-task feature set, constructing a dual-path feature extraction network including a conductive feature encoder and a mechanical feature encoder, to obtain a shared feature representation; constructing a meta-learning model based on the shared feature representation, generating a support set and a query set by task sampling, performing feature enhancement training on the support set, to obtain a meta-learning feature model; The meta-learning feature model is quickly adapted on the query set to obtain an adaptive feature model, which includes a conductive feature adaptation branch and a mechanical feature adaptation branch; the conductive feature parameters and the mechanical feature parameters are respectively input into the corresponding branches of the adaptive feature model to perform feature reconstruction and optimization to obtain a reconstructed feature vector; the reconstructed feature vector is adaptively fused with the original feature parameters to obtain enhanced feature parameters; the conductive feature parameters are feature compensated according to the enhanced feature parameters to obtain compensated conductive feature parameters, and the mechanical feature parameters are feature compensated to obtain compensated mechanical feature parameters; the compensated conductive feature parameters and the compensated mechanical feature parameters are combined to obtain an enhanced feature vector.

[0198] In a specific embodiment, the execution step inputs the multi-scale feature vector into the second neural network in the double-layer stack quality detection model for quality detection analysis, the second neural network includes three parallel feature extraction branches and one feature fusion branch, and the process of outputting the quality detection result can specifically include the following steps:

[0199] Input the multi-scale feature vector into the first feature extraction branch of the second neural network for 3×3 convolution and maximum pooling processing to obtain a first feature map, input the multi-scale feature vector into the second feature extraction branch of the second neural network for 5×5 convolution and average pooling processing to obtain a second feature map, and input the multi-scale feature vector into the third feature extraction branch of the second neural network for 7×7 convolution and adaptive pooling processing to obtain a third feature map;

[0200] Perform channel attention calculation on the first feature map to obtain a first channel weight vector, perform channel attention calculation on the second feature map to obtain a second channel weight vector, and perform channel attention calculation on the third feature map to obtain a third channel weight vector;

[0201] The first channel weight vector is weighted with the first feature map to obtain a weighted first feature map, the second channel weight vector is weighted with the second feature map to obtain a weighted second feature map, and the third channel weight vector is weighted with the third feature map to obtain a weighted third feature map;

[0202] Input the weighted first feature map, the weighted second feature map and the weighted third feature map into the feature fusion branch, and obtain a fused feature vector through a fully connected layer.

[0203] The fused feature vector is processed by Softmax classification to obtain the quality category probability distribution, and the quality category probability distribution is compared with the preset quality classification threshold to output the quality detection result.

[0204] Specifically, let the multi-scale feature vector be F multi ∈R n×d×h×w , where n is the number of samples, d is the number of channels, h and w are the spatial dimensions of the feature map. multi Input the first feature extraction branch, use 3×3 convolution kernel for feature extraction, and use the maximum pooling operation to reduce the size of the feature map. The calculation formula is as follows:

[0205] F conv3 =MaxPool(ReLU(Conv3x3(F mu1ti )));

[0206] Among them, Conv3x3 represents a 3×3 convolution operation, ReLU is a nonlinear activation function, and MaxPool represents a maximum pooling operation to ensure that the strongest local features are retained. The multi-scale feature vector is input into the second feature extraction branch, and a 5×5 convolution kernel is used to extract features with a larger receptive field, and an average pooling operation is used to smooth the feature values. The calculation method is as follows:

[0207] F conv5 =AvgPool(ReLU(Conv5x5(F multi )));

[0208] Among them, Conv5x5 represents a 5×5 convolution operation, and AvgPool represents an average pooling operation, which ensures balanced calculation of feature information over a larger range. The multi-scale feature vector is input into the third feature extraction branch, and a 7×7 convolution kernel is used to extract features over a larger range, and an adaptive pooling operation is used to adapt to different input sizes. The calculation method is as follows:

[0209] F conv7 =AdaptivePool(ReLU(Conv7x7(F multi )));

[0210] Among them, Conv7x7 represents a 7×7 convolution operation, and AdaptivePool represents an adaptive pooling operation, which can automatically adjust the size of the output feature map to adapt to the fully connected layer input. conv3 , the second feature map Fconv5 And the third characteristic graph F conv7 Perform channel attention calculation to learn the importance weights of the channels and enhance key features. Calculate the channel attention weight matrix A of the first feature map conv3 :

[0211] A conv3 =σ(W attn ·GAP(F conv3 ));

[0212] Among them, W attn is a trainable weight matrix, GAP represents the global average pooling operation, and σ represents the sigmoid function, which ensures that the attention weight is normalized between [0,1]. Similarly, the channel attention weight matrix A of the second feature map is calculated conv5 :

[0213] A conv5 =σ(W attn ·GAP(F conv5 ));

[0214] And the channel attention weight matrix A of the third feature map conv7 :

[0215] A conv7 =σ(W attn ·GAP(F conv7 ));

[0216] The first channel weight vector is weighted with the first feature map to highlight the information of important channels. The calculation is as follows:

[0217]

[0218] Among them, ⊙ represents the channel-by-channel point multiplication operation. Similarly, the weighted second and third feature maps are calculated:

[0219]

[0220] All weighted feature maps are input into the feature fusion branch and fused through the fully connected layer:

[0221]

[0222] Among them, W fusion is the weight matrix of the fully connected layer, b fusion is the bias term, BN stands for batch normalization operation to accelerate training and improve model stability, and ReLU is used as the activation function to ensure nonlinear feature extraction capabilities. The fused feature vector is processed with Softmax to calculate the probability distribution of different quality categories:

[0223] y=Softmax(W out F fused +b out );

[0224] Among them, W out is the output layer weight matrix, b out It is a bias term, and Softmax ensures that the sum of the probabilities of all categories is 1. The output quality category probability distribution is compared with the preset classification threshold to determine whether the quality category of the sample meets the requirements, and output the final quality detection result.

[0225] See also Figure 2 , Figure 2 A schematic block diagram of the structure of a quality detection device 200 for a carbon fiber shadowless medical line provided in an embodiment of the present application, as shown in FIG. Figure 2 As shown, the quality inspection device 200 of carbon fiber shadowless medical wire includes:

[0226] An installation module 210 is used to install electrical connection terminals on the inner core conductor and shielding wire of the carbon fiber shadowless medical wire, and fix the carbon fiber shadowless medical wire on a test platform to obtain a sample to be tested;

[0227] The testing module 220 is used to perform a conductivity test on the sample to be tested to obtain a conductivity test data set, and to perform a mechanical property test on the sample to be tested to obtain a mechanical property test data set;

[0228] A feature extraction module 230 is used to perform feature extraction on the conductive performance test data set to obtain conductive feature parameters, and to perform feature extraction on the mechanical performance test data set to obtain mechanical feature parameters;

[0229] The quality detection module 240 is used to input the conductive characteristic parameters and the mechanical characteristic parameters into the double-layer stack quality detection model to perform quality detection and output the quality detection result.

[0230] Through the synergy of the above components, a dual-branch feature extraction structure is designed to process the conductive properties and mechanical properties respectively, and the feature interaction layer is used to achieve deep fusion of the two types of performance features, which effectively improves the comprehensiveness and accuracy of quality detection. The feature weighting method combining information entropy weight and coefficient of variation weight is adopted to reasonably allocate the importance of different feature parameters, overcoming the limitations of the traditional single weight method. The feature interaction layer design with attention mechanism and residual connection enhances the information interaction between conductive features and mechanical features and improves the feature expression ability. Three parallel feature extraction branches are designed, and multi-scale features are extracted through convolution kernels and pooling operations of different scales, which enhances the model's perception of features of different scales. Through the channel attention mechanism in the feature fusion branch, the importance of different feature channels is adaptively adjusted, which improves the accuracy and reliability of quality detection. The conductive performance test and mechanical performance test are standardized and automated, which significantly improves the detection efficiency and reduces the interference of human factors.

[0231] See also Figure 3 , Figure 3 A schematic block diagram of the structure of a carbon fiber shadowless medical line quality inspection device 300 provided in an embodiment of the present application, wherein the carbon fiber shadowless medical line quality inspection device 300 includes a processor 301 and a memory 302, wherein the processor 301 and the memory 302 are connected via a device bus 303, wherein the memory 302 may include a non-volatile storage medium and an internal memory.

[0232] The non-volatile storage medium can store a computer program. The computer program includes program instructions, and when the program instructions are executed by the processor 301, the processor 301 can execute any of the above-mentioned quality detection methods for carbon fiber shadowless medical wires.

[0233] The processor 301 is used to provide computing and control capabilities to support the operation of the entire carbon fiber shadowless medical line quality inspection device 300.

[0234] The internal memory provides an environment for the operation of the computer program in the non-volatile storage medium. When the computer program is executed by the processor 301, the processor 301 can execute any of the above-mentioned quality detection methods for carbon fiber shadowless medical wires.

[0235] Those skilled in the art will understand that Figure 3 The structure shown in the figure is merely a block diagram of a partial structure related to the present application scheme, and does not constitute a limitation on the quality inspection device 300 of the carbon fiber shadowless medical line involved in the present application scheme. The specific quality inspection device 300 of the carbon fiber shadowless medical line may include more or fewer components than shown in the figure, or combine certain components, or have a different arrangement of components.

[0236] It should be understood that the processor 301 may be a central processing unit (CPU), and the processor 301 may also be other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. Among them, the general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0237] It should be noted that technical personnel in the relevant field can clearly understand that, for the convenience and conciseness of description, the specific working process of the carbon fiber shadowless medical line quality inspection equipment 300 described above can refer to the corresponding process of the aforementioned carbon fiber shadowless medical line quality inspection method, and will not be repeated here.

[0238] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described systems, systems and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0239] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk and other media that can store program code.

[0240] As described above, the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A quality inspection method for carbon fiber shadowless medical wire, characterized in that: include: Install electrical connection terminals on the inner core conductor and shielding wire of the carbon fiber shadowless medical wire, and fix the carbon fiber shadowless medical wire on a test platform to obtain a sample to be tested; Performing a conductivity test on the sample to be tested to obtain a conductivity test data set, and performing a mechanical property test on the sample to be tested to obtain a mechanical property test data set; Performing feature extraction on the conductive performance test data set to obtain conductive feature parameters, and performing feature extraction on the mechanical performance test data set to obtain mechanical feature parameters; The conductive characteristic parameters and the mechanical characteristic parameters are input into a double-layer stack quality detection model for quality detection, and a quality detection result is output.

2. The quality inspection method of carbon fiber shadowless medical wire according to claim 1, characterized in that: The inner core conductor and the shielding wire of the carbon fiber shadowless medical wire are installed with electrical connection terminals, and the carbon fiber shadowless medical wire is fixed on a test platform to obtain a sample to be tested, including: Connecting the inner core conductor of the carbon fiber shadowless medical wire to the first connection part, and connecting the shielding wire of the carbon fiber shadowless medical wire to the second connection part to obtain an initial connection structure; The first connection part is subjected to gold plating so that the diameter of the first connection part matches the diameter of the inner core conductor to obtain an inner core conductor connection structure, and the second connection part is subjected to silver-plated copper mesh structure so that the second connection part matches the braided layer of the shielding wire to obtain a shielding wire connection structure; Fixing the inner core conductor connection structure and the shielding wire connection structure on a test platform, setting the temperature and relative humidity of the test platform to obtain an environmentally conditioned test platform, and installing a four-wire resistance measurement system on the environmentally conditioned test platform to obtain a resistance measurement structure; A high-frequency impedance analyzer and a network analyzer are installed on the environmentally conditioned test platform, the measurement ports of the high-frequency impedance analyzer and the network analyzer are connected to the resistance measurement structure via standard radio frequency cables, and the temperature of the aging test box is set to obtain a sample to be tested.

3. The quality inspection method of carbon fiber shadowless medical wire according to claim 1, characterized in that: The conducting of the conductivity test on the sample to be tested to obtain a conductivity test data set, and the conducting of the mechanical property test on the sample to be tested to obtain a mechanical property test data set, comprises: Performing four-wire resistance measurement on the sample to be tested to obtain temperature-resistance correspondence data; The transmission loss of the sample to be tested is measured, and the transmission loss parameter S21 and the crosstalk parameter S41 of the shielded wire are measured by a high-frequency impedance analyzer to obtain the transmission loss average value data; Performing an electromagnetic shielding test on the sample to be tested, performing a frequency sweep test and electromagnetic shielding effectiveness measurement using a network analyzer, and obtaining frequency-attenuation curve data; The sample to be tested is placed in an aging test box for aging treatment and conductivity parameter measurement to obtain aging conductivity data, and the temperature-resistance correspondence data, the transmission loss average data, the frequency-attenuation curve data and the aging conductivity data are combined into a conductivity test data set; Perform a tensile test on the sample to be tested according to a preset loading rate, perform a reciprocating bending test on the sample to be tested according to a preset bending radius, and perform a torsion test on the sample to be tested according to a preset torsion speed to obtain stress-strain data, bending number data, and torsion circle number data; Perform an alternating bending fatigue test on the sample to be tested, record load-displacement data, and combine the stress-strain data, the bending number data, the torsion circle number data, and the load-displacement data into a mechanical property test data set.

4. The quality inspection method of carbon fiber shadowless medical wire according to claim 3 is characterized in that: The step of extracting features from the conductive performance test data set to obtain conductive feature parameters, and extracting features from the mechanical performance test data set to obtain mechanical feature parameters, includes: Performing linear regression on the temperature-resistance correspondence data in the conductive performance test data set, calculating the slope of the resistance-temperature curve to obtain the resistance temperature coefficient, and performing maximum-minimum value ratio calculation on the temperature-resistance correspondence data to obtain the resistance stability coefficient; Performing arithmetic averaging on the transmission loss average value data in the conductive performance test data set to obtain a transmission loss mean value, and performing numerical calculation on the crosstalk parameter S41 to obtain a crosstalk-to-noise ratio; Integrate and average the frequency-attenuation curve data in the conductive performance test data set to obtain the electromagnetic shielding attenuation coefficient, calculate the change rate of the aged conductive performance data to obtain the conductive performance aging rate, and combine the resistance temperature coefficient, the resistance stability coefficient, the transmission loss mean, the crosstalk noise ratio, the electromagnetic shielding attenuation coefficient and the conductive performance aging rate into a conductive characteristic parameter; Calculating the slope of the linear segment of the stress-strain data in the mechanical property test data set to obtain Young's modulus, and extracting the maximum value of the stress-strain data to obtain the breaking strength and breaking elongation; Standardizing the bending times data in the mechanical property test data set to obtain a bending life coefficient, and standardizing the torsion turns data to obtain a torsion strength coefficient; The load-displacement data is subjected to cumulative damage calculation to obtain fatigue life parameters, and the Young's modulus, the fracture strength, the fracture elongation, the bending life coefficient, the torsional strength coefficient and the fatigue life parameter are combined into mechanical characteristic parameters.

5. The quality inspection method of carbon fiber shadowless medical wire according to claim 1, characterized in that: The step of inputting the conductive characteristic parameter and the mechanical characteristic parameter into a double-layer stack quality detection model for quality detection and outputting a quality detection result comprises: Performing information entropy weight and variation coefficient weight calculation on the conductive characteristic parameters to obtain a conductive characteristic comprehensive weight coefficient, and performing information entropy weight and variation coefficient weight calculation on the mechanical characteristic parameters to obtain a mechanical characteristic comprehensive weight coefficient; The conductive feature parameters are weighted combined according to the conductive feature comprehensive weight coefficient to construct a conductive feature fusion vector, and the mechanical feature parameters are weighted combined according to the mechanical feature comprehensive weight coefficient to construct a mechanical feature fusion vector; Input the conductive feature fusion vector into the conductive feature branch of the first neural network in the double-layer stacking quality detection model, wherein the conductive feature branch includes a fully connected layer, a batch normalization layer, and a ReLU activation function layer, to obtain a conductive feature representation vector; input the mechanical feature fusion vector into the mechanical feature branch of the first neural network in the double-layer stacking quality detection model, wherein the mechanical feature branch includes a fully connected layer, a batch normalization layer, and a ReLU activation function layer, to obtain a mechanical feature representation vector; A feature interaction layer is set in the first neural network in the double-layer stacking quality detection model, and the conductive feature representation vector and the mechanical feature representation vector are weighted by an attention mechanism and subjected to residual connection processing to obtain an interaction feature vector; Performing feature recombination on the interaction feature vector, the conductive feature characterization vector, and the mechanical feature characterization vector to obtain a multi-scale feature vector; The multi-scale feature vector is input into the second neural network in the double-layer stacking quality detection model for quality detection analysis, and the second neural network includes three parallel feature extraction branches and one feature fusion branch, and outputs the quality detection result.

6. The quality inspection method of carbon fiber shadowless medical wire according to claim 5, characterized in that: The feature interaction layer is set in the first neural network in the double-layer stacking quality detection model, and the conductive feature representation vector and the mechanical feature representation vector are weighted by the attention mechanism and the residual connection is processed to obtain the interaction feature vector, including: Performing self-attention calculation on the conductive feature representation vector to obtain a conductive feature attention weight matrix, and performing self-attention calculation on the mechanical feature representation vector to obtain a mechanical feature attention weight matrix; Performing cross-attention calculation on the conductive feature characterization vector and the mechanical feature characterization vector to obtain a cross-feature attention weight matrix, and performing a weighted summation operation based on the cross-feature attention weight matrix to obtain a weighted cross-feature vector; Performing a 1×1 convolution process on the conductive feature characterization vector to obtain a conductive feature residual vector, and performing a 1×1 convolution process on the mechanical feature characterization vector to obtain a mechanical feature residual vector; Performing a dot multiplication operation on the conductive feature attention weight matrix and the conductive feature residual vector to obtain a weighted conductive feature vector, and performing a dot multiplication operation on the mechanical feature attention weight matrix and the mechanical feature residual vector to obtain a weighted mechanical feature vector; The weighted conductive feature vector, the weighted mechanical feature vector and the weighted cross feature vector are feature spliced ​​to obtain a spliced ​​feature vector, and the spliced ​​feature vector is fully connected layer transformed and normalized to obtain an interactive feature vector.

7. The quality inspection method of carbon fiber shadowless medical wire according to claim 5, characterized in that: The multi-scale feature vector is input into the second neural network in the double-layer stacking quality detection model for quality detection analysis, wherein the second neural network includes three parallel feature extraction branches and one feature fusion branch, and outputs quality detection results, including: Input the multi-scale feature vector into the first feature extraction branch of the second neural network for 3×3 convolution and maximum pooling processing to obtain a first feature map, input the multi-scale feature vector into the second feature extraction branch of the second neural network for 5×5 convolution and average pooling processing to obtain a second feature map, and input the multi-scale feature vector into the third feature extraction branch of the second neural network for 7×7 convolution and adaptive pooling processing to obtain a third feature map; Performing channel attention calculation on the first feature map to obtain a first channel weight vector, performing channel attention calculation on the second feature map to obtain a second channel weight vector, and performing channel attention calculation on the third feature map to obtain a third channel weight vector; The first channel weight vector is weighted with the first feature map to obtain a weighted first feature map, the second channel weight vector is weighted with the second feature map to obtain a weighted second feature map, and the third channel weight vector is weighted with the third feature map to obtain a weighted third feature map; Inputting the weighted first feature map, the weighted second feature map and the weighted third feature map into a feature fusion branch, and obtaining a fused feature vector through a fully connected layer processing; The fused feature vector is subjected to Softmax classification processing to obtain a quality category probability distribution, and the quality category probability distribution is compared with a preset quality classification threshold to output a quality detection result.

8. A quality inspection device for carbon fiber shadowless medical wire, characterized in that: A method for detecting the quality of a carbon fiber shadowless medical wire according to any one of claims 1 to 7, comprising: An installation module is used to install electrical connection terminals on the inner core conductor and shielding wire of the carbon fiber shadowless medical wire, and fix the carbon fiber shadowless medical wire on a test platform to obtain a sample to be tested; A testing module, used to perform a conductivity test on the sample to be tested to obtain a conductivity test data set, and to perform a mechanical property test on the sample to be tested to obtain a mechanical property test data set; A feature extraction module, used to perform feature extraction on the conductive performance test data set to obtain conductive feature parameters, and to perform feature extraction on the mechanical performance test data set to obtain mechanical feature parameters; The quality detection module is used to input the conductive characteristic parameters and the mechanical characteristic parameters into a double-layer stacking quality detection model for quality detection and output a quality detection result.

9. A quality inspection device for carbon fiber shadowless medical wire, characterized in that: The quality inspection device of the carbon fiber shadowless medical line comprises: a memory and at least one processor, wherein the memory stores instructions; The at least one processor calls the instruction in the memory to enable the quality detection device of the carbon fiber shadowless medical wire to perform the quality detection method of the carbon fiber shadowless medical wire according to any one of claims 1 to 7.

Images