Fabric air permeability detection method and system based on deep learning

Abstract

The invention relates to the field of intelligent detection, and provides a fabric air permeability detection method and system based on deep learning, and the method comprises the steps: firstly, flatly placing a fabric on a bottom plate, clamping the fabric with a negative pressure chamber, and then extracting air through an exhaust pipe to enable the air pressure in the chamber to be stable; the method comprises the following steps: firstly, measuring the air pressure of a fabric, then measuring the upheaval height of the fabric and the air flow value in an exhaust pipe in unit time when the air pressure is stable, then correcting the measured air flow value to obtain an air flow correction value, calculating the actually measured air permeability of the fabric according to the air flow correction value, and finally correcting the actually measured air permeability according to the upheaval height of the fabric to obtain the corrected air permeability. Therefore, the accuracy of the actually measured air permeability result can be ensured, and the reliability of the corrected air permeability result is further improved, so that the accuracy of fabric air permeability detection is effectively improved.

Description

Technical Field

[0001] The present application relates to the field of intelligent detection, and more specifically, to a method and system for detecting the air permeability of fabrics based on deep learning. Background Art

[0002] In today's textile industry, the air permeability of fabrics, as a key indicator for measuring their quality and functionality, plays a decisive role in the performance of various products and the user experience; from the sweat-venting and air-permeable requirements of sports clothing, to the comfortable and air-permeable standards of medical protective supplies, and then to the skin-friendly and air-permeable considerations of home textiles, accurately detecting the air permeability of fabrics is crucial.

[0003] The existing patent CN118483141B proposes a device and method for detecting the air permeability of fabrics. First, the fabric to be tested is placed flat and clamped, then the air is pumped to reach a set air pressure value, and a gas flow sensor is used to measure the gas flow rate, and the measured air permeability rate is calculated based on the gas flow rate; then, according to the prediction model, the surface area of the fabric after bulging is calculated based on the bulging height, and then the correction coefficient is obtained; finally, this correction coefficient is applied to adjust the measured air permeability rate to obtain a more accurate air permeability rate result.

[0004] In this patent, the air permeability rate is directly measured based on the actually measured gas flow rate value. However, during the air pumping process, the temperature gradient and pressure gradient inside and outside the negative pressure chamber may cause heat convection and gas density changes, thus interfering with the flow measurement; specifically, the temperature difference will cause the gas density to be uneven, affecting the accuracy of the flow sensor; at the same time, the pressure gradient will cause the non-ideal state of gas flow, further increasing the measurement error of the gas flow rate; these factors may cause the measured air permeability rate result to deviate from the true value, reducing the reliability and repeatability of the subsequent correction of the measured air permeability rate.

[0005] Therefore, a fabric air permeability detection solution based on deep learning is desired. Summary of the Invention

[0006] The present application aims at the deficiencies in the prior art and provides a method and system for detecting the air permeability of fabrics based on deep learning. According to one aspect of the present application, a method for detecting the air permeability of fabrics based on deep learning is provided, which includes:

[0007] Place the fabric flat on the bottom plate and clamp the fabric using a negative pressure chamber;

[0008] Pump air into the negative pressure chamber through an air extraction pipe so that the air pressure in the negative pressure chamber remains stable;

[0009] Detect the bulging height of the fabric when the air pressure is stable, and measure the gas flow rate value per unit time in the air extraction pipe;

[0010] Calibrate the gas flow value to obtain a corrected gas flow value, including: monitoring and collecting {internal temperature data, internal pressure data, external temperature data, and external pressure data} through a first sensor group deployed inside the negative pressure chamber and a second sensor group deployed outside the negative pressure chamber; performing an interactive response based on the core time series characteristics of temperature difference - pressure difference on {internal temperature data, internal pressure data, external temperature data, and external pressure data} to obtain the corrected gas flow value;

[0011] Calculate the measured air permeability of the fabric based on the corrected gas flow value;

[0012] Correct the measured air permeability of the fabric based on the bulging height of the fabric to obtain a corrected air permeability.

[0013] According to another aspect of the present application, there is provided a fabric air permeability detection system based on deep learning, which includes: a fabric clamping module for laying the fabric flat on the bottom plate and clamping the fabric using a negative pressure chamber;

[0014] An air pressure stabilization module for pumping air into the negative pressure chamber through an air extraction pipe to keep the air pressure inside the negative pressure chamber stable; a data detection module for detecting the bulging height of the fabric when the air pressure is stable and measuring the gas flow value per unit time in the air extraction pipe;

[0015] A flow rate calibration module for calibrating the gas flow value to obtain a corrected gas flow value, where the flow rate calibration module is used to: monitor and collect {internal temperature data, internal pressure data, external temperature data, and external pressure data} through a first sensor group deployed inside the negative pressure chamber and a second sensor group deployed outside the negative pressure chamber; perform an interactive response based on the core time series characteristics of temperature difference - pressure difference on {internal temperature data, internal pressure data, external temperature data, and external pressure data} to obtain the corrected gas flow value;

[0016] An air permeability calculation module for calculating the measured air permeability of the fabric based on the corrected gas flow value;

[0017] An air permeability correction module for correcting the measured air permeability of the fabric based on the bulging height of the fabric to obtain a corrected air permeability.

[0018] Due to the adoption of the above technical solutions, the present application has significant technical effects:

[0019] The fabric air permeability detection method and system based on deep learning provided by this application first place the fabric flat on the bottom plate and clamp it with a negative pressure chamber, then extract air through the air extraction pipe to make the air pressure in the chamber stable, then measure the fabric bulge height and the gas flow rate value per unit time in the air extraction pipe when the air pressure is stable, immediately correct the measured gas flow rate value to obtain a gas flow rate correction value, and calculate the measured air permeability of the fabric based on this, and finally correct the measured air permeability according to the height of the fabric bulge to obtain a corrected air permeability. In this way, the accuracy of the measured air permeability result can be ensured, and then the reliability of the corrected air permeability result can be improved, thereby effectively improving the accuracy of fabric air permeability detection. Description of the Drawings

[0020] By describing the embodiments of the present application in more detail in conjunction with the drawings, the above and other objects, features, and advantages of the present application will become more obvious. The drawings are used to provide a further understanding of the embodiments of the present application, and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation to the present application. In the drawings, the same reference numerals generally represent the same components or steps.

[0021] Figure 1 It is a flowchart of the fabric air permeability detection method based on deep learning according to an embodiment of the present application.

[0022] Figure 2 It is a flowchart of step S4 in the fabric air permeability detection method based on deep learning according to an embodiment of the present application.

[0023] Figure 3 It is a flowchart of step S42 in the fabric air permeability detection method based on deep learning according to an embodiment of the present application.

[0024] Figure 4 It is a flowchart of step S421 in the fabric air permeability detection method based on deep learning according to an embodiment of the present application.

[0025] Figure 5 It is a flowchart of step S423 in the fabric air permeability detection method based on deep learning according to an embodiment of the present application.

[0026] Figure 6 It is a flowchart of step S424 in the fabric air permeability detection method based on deep learning according to an embodiment of the present application.

[0027] Figure 7 It is a system block diagram of the fabric air permeability detection system based on deep learning according to an embodiment of the present application. Detailed Embodiments

[0028] Next, exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. It should be understood that the present application is not limited by the exemplary embodiments described herein.

[0029] In the modern textile industry, fabric breathability, as a core indicator to measure its quality and functionality, has a decisive impact on the performance of various textile products and the user experience. Whether it is the sweat-venting and breathable requirements of sports clothing, the comfortable and breathable standards of medical protective supplies, or the skin-friendly and breathable requirements of home textiles, accurately detecting fabric breathability is extremely crucial. The existing patent CN118483141B provides a fabric breathability detection device and a detection method. First, the fabric to be tested is placed flat and clamped, and the air pressure is made to reach a set value by pumping air. Then, a gas flow sensor is used to measure the gas flow rate, and the measured air permeability is calculated based on this. Subsequently, according to the prediction model, the surface area after bulging is calculated based on the fabric bulging height, and then the correction coefficient is obtained. Finally, this correction coefficient is used to adjust the measured air permeability to obtain a more accurate detection result. However, it should be noted that during the air pumping process, the temperature gradient and pressure gradient generated inside and outside the negative pressure chamber may cause heat convection and changes in gas density, and these factors may affect the accuracy of the flow sensor, resulting in a deviation between the measured air permeability value and the true value, thereby affecting the reliability and repeatability of the final air permeability result.

[0030] Based on this, the present application proposes a fabric breathability detection method based on deep learning. Figure 1 It is a flowchart of the fabric breathability detection method based on deep learning according to an embodiment of the present application. As Figure 1As shown, the method for detecting the air permeability of a fabric based on deep learning according to an embodiment of the present application includes: S1, placing the fabric flat on the bottom plate and clamping the fabric using a negative pressure chamber; S2, pumping air into the negative pressure chamber through an air extraction pipe to keep the air pressure in the negative pressure chamber stable; S3, detecting the bulging height of the fabric when the air pressure is stable, and measuring the gas flow rate value per unit time in the air extraction pipe; S4, correcting the gas flow rate value to obtain a corrected gas flow rate value; S5, calculating the measured air permeability of the fabric based on the corrected gas flow rate value; S6, correcting the measured air permeability of the fabric based on the bulging height of the fabric to obtain a corrected air permeability. In step S1, the fabric is placed flat on the bottom plate and clamped using a negative pressure chamber. It should be understood that by placing the fabric flat and firmly clamping it, it can ensure that the fabric is in a standard state for testing, avoiding measurement errors caused by irregular placement of the fabric, so that more accurate and reproducible air permeability measurement results can be obtained. Moreover, the fabric will be subjected to various external forces in actual applications. When detecting, clamping the fabric by the negative pressure chamber can simulate the stress situation in its actual use scenario, thereby making the detection result more in line with the actual situation and improving the practicality and effectiveness of the detection. In step S2, pumping air into the negative pressure chamber through an air extraction pipe to keep the air pressure in the negative pressure chamber stable. It should be understood that many fabrics will face different air pressure environments in actual use scenarios. For example, when sports clothing is in use during human movement, the air pressure on both sides of the fabric will change. When detecting, pumping air into the negative pressure chamber through an air extraction pipe can simulate the air pressure difference situation that the fabric may encounter in actual use, and thus can provide an effective basis for evaluating the air permeability performance of the fabric in a real environment. It should be particularly noted that stable air pressure is the key to accurately measuring the relevant data of air permeability subsequently. Pumping air to keep the air pressure in the negative pressure chamber stable helps to reduce additional variables that may be caused by air pressure fluctuations, such as temperature gradients or non-ideal gas flow states, thereby improving the accuracy of air permeability measurement.

[0031] In step S3, detecting the bulging height of the fabric when the air pressure is stable, and measuring the gas flow rate value per unit time in the air extraction pipe. It should be understood that according to the principle of measuring the air permeability of a fabric by the pressure difference method, the air permeability is calculated by dividing the air flow rate passing through the fabric per unit time by the detection area of the fabric and then by the air pressure difference on both sides of the fabric. Measuring the gas flow rate value per unit time in the air extraction pipe directly provides one of the core data for this calculation. Only by obtaining the gas flow rate data can the measured air permeability of the fabric be calculated when the detection area and air pressure difference are known, so as to evaluate the air permeability performance of the fabric. At the same time, detecting the bulging height of the fabric is because the bulging of the fabric during the detection process will cause a change in the actual area, which will in turn affect the air permeability. Accurately measuring the bulging height helps to more accurately predict the surface area change after the fabric bulges subsequently, so as to reasonably correct the measured air permeability and make the finally obtained air permeability more in line with the actual air permeability of the fabric.

[0032] In step S4, the gas flow value is corrected to obtain a gas flow correction value. It should be understood that gas has the characteristics of thermal expansion and contraction, and temperature changes will change the density of the gas. At higher temperatures, the movement of gas molecules intensifies, and the distance between molecules increases, resulting in a decrease in gas density; conversely, when the temperature decreases, the gas density increases. And pressure changes will affect the flow rate of the gas. The greater the pressure difference, the faster the gas flow rate. In the fabric breathability test, the temperature and pressure inside and outside the negative pressure chamber are in dynamic changes, and these changes will directly change the physical state and flow characteristics of the gas, thereby significantly affecting the measurement result of the gas flow rate.

[0033] Based on this, the present application monitors and collects {internal temperature data, internal pressure data, external temperature data, and external pressure data} through a first sensor group deployed inside the negative pressure chamber and a second sensor group deployed outside the negative pressure chamber, and uses data analysis and coding techniques based on deep learning to regularize the collected data and calculate the internal and external time-sequence temperature difference and pressure difference. Then, time-sequence gradient feature extraction is performed on the calculated internal and external time-sequence temperature difference and pressure difference, and based on this, a gas flow correction coefficient is intelligently generated according to the core time-sequence interaction response representation between the internal and external temperature difference time-sequence gradient features and the internal and external pressure difference time-sequence gradient features obtained after time-sequence extraction, and the gas flow value is corrected based on this correction coefficient. By monitoring the temperature and pressure data inside and outside the negative pressure chamber and calculating the internal and external temperature difference and pressure difference, the present application can effectively compensate for the gas flow measurement error caused by the temperature gradient and pressure gradient. This makes the gas flow measurement closer to the true value, thereby improving the accuracy of the measured air permeability rate, and thus improving the credibility and accuracy of the final test.

[0034] Specifically, Figure 2 is a flowchart of step S4 in the fabric breathability detection method based on deep learning according to an embodiment of the present application. As Figure 2 shown, step S4 includes: S41, monitoring and collecting {internal temperature data, internal pressure data, external temperature data, and external pressure data} through a first sensor group deployed inside the negative pressure chamber and a second sensor group deployed outside the negative pressure chamber; S42, obtaining a gas flow correction value based on the core time-sequence feature interaction response of {internal temperature data, internal pressure data, external temperature data, and external pressure data}.

[0035] In step S41, the first sensor group deployed inside the negative pressure chamber and the second sensor group deployed outside the negative pressure chamber are used to monitor and collect {internal temperature data, internal pressure data, external temperature data, and external pressure data}. It should be understood that the internal temperature data reflects the thermal state of the gas inside the negative pressure chamber, and the external temperature data reflects the ambient temperature. Temperature changes can cause gases to expand and contract thermally, changing their density. For example, when the temperature rises, the molecular motion of the gas intensifies, the molecular spacing increases, the density decreases, and the mass of the gas decreases under the same volume. When measuring the flow rate, if the density change is not considered, the measurement result will deviate. The internal pressure data and the external pressure data reflect the pressure conditions between the negative pressure chamber and the external environment. The pressure difference is the driving force for gas flow. The greater the pressure difference, the faster the gas flow rate. During the detection process, the dynamic change of pressure will make the gas flow rate unstable, affecting the accuracy of flow rate measurement. Generally speaking, by integrating and analyzing the collected internal temperature data, internal pressure data, external temperature data, and external pressure data, the physical state changes and flow characteristics changes of the gas during the detection process can be comprehensively understood, which can provide rich and key information for generating the gas flow correction coefficient, and then achieve accurate correction of the gas flow value to improve the accuracy of fabric breathability detection.

[0036] In step S42, {internal temperature data, internal pressure data, external temperature data, and external pressure data} are used to obtain the gas flow correction value through the interaction response of the core time series features of temperature difference - pressure difference. Specifically, Figure 3 FIG. is a flowchart of step S42 in the fabric breathability detection method based on deep learning according to an embodiment of the present application. As Figure 3 shown, step S42 includes: S421, based on {internal temperature data, internal pressure data, external temperature data, and external pressure data}, determining the internal - external temperature difference time series vector and the internal - external pressure difference time series vector; S422, performing time series feature extraction on the internal - external temperature difference time series vector and the internal - external pressure difference time series vector to obtain the internal - external temperature difference time series gradient feature vector and the internal - external pressure difference time series gradient feature vector; S423, performing interaction response encoding driven by the core time series features of temperature difference - pressure difference on the internal - external temperature difference time series gradient feature vector and the internal - external pressure difference time series gradient feature vector to obtain the temperature difference - pressure difference time series gradient interaction response encoding vector; S424, based on the temperature difference - pressure difference time series gradient interaction response encoding vector, correcting the gas flow value to obtain the gas flow correction value. In step S421, based on {internal temperature data, internal pressure data, external temperature data, and external pressure data}, the internal - external temperature difference time series vector and the internal - external pressure difference time series vector are determined. Specifically, Figure 4 FIG. is a flowchart of step S421 in the fabric breathability detection method based on deep learning according to an embodiment of the present application. As Figure 4As shown, step S421 includes: S4211, regularizing {internal temperature data, internal pressure data, external temperature data, and external pressure data} to obtain an internal temperature time series vector, an internal pressure time series vector, an external temperature time series vector, and an external pressure time series vector; S4212, determining an internal-external temperature difference time series vector based on the internal temperature time series vector and the external temperature time series vector; S4213, determining an internal-external pressure difference time series vector based on the internal pressure time series vector and the external pressure time series vector.

[0037] In step S4211, {internal temperature data, internal pressure data, external temperature data, and external pressure data} are regularized to obtain an internal temperature time series vector, an internal pressure time series vector, an external temperature time series vector, and an external pressure time series vector. Correspondingly, considering that the collected internal temperature data, internal pressure data, external temperature data, and external pressure data may have different units, dimensions, and value ranges. For example, the temperature data may be in degrees Celsius and have a value range within a certain interval; the pressure data may be in kilopascals and have a completely different value range from the temperature data. Therefore, in order to uniformly convert these data into the form of time series vectors, the dimension differences can be eliminated, making different types of data comparable, and analyzing and processing the time series characteristics of each data more carefully and accurately. In the technical solution of this application, {internal temperature data, internal pressure data, external temperature data, and external pressure data} are regularized to obtain an internal temperature time series vector, an internal pressure time series vector, an external temperature time series vector, and an external pressure time series vector.

[0038] In step S4212, based on the internal temperature time series vector and the external temperature time series vector, an internal-external temperature difference time series vector is determined. Specifically, in the embodiments of the present application, step S4212 includes: calculating the position-wise difference between the internal temperature time series vector and the external temperature time series vector to obtain the internal-external temperature difference time series vector. It should be understood that temperature is one of the key factors affecting the physical properties and flow characteristics of gases. During the fabric breathability detection process, the temperature difference between the inside and outside of the negative pressure chamber will cause heat exchange of the gas, which in turn affects the density and flow rate of the gas, and ultimately affects the gas flow measurement result. In order to observe the dynamic change trend of the internal-external temperature difference, the present application determines the internal-external temperature difference time series vector based on the internal temperature time series vector and the external temperature time series vector. That is to say, by converting them into the internal-external temperature difference time series vector, the magnitude and change law of the internal-external temperature difference at different times can be clearly shown. For example, during the detection process, it may occur that the external temperature gradually increases while the internal temperature remains relatively stable due to operations such as air extraction. Through the internal-external temperature difference time series vector, this change process of the temperature difference can be intuitively reflected, providing a more intuitive data perspective for in-depth study of the influence of temperature on gas flow and providing key data support for calculating the flow correction coefficient. In particular, in a specific example of the present application, the internal-external temperature difference time series vector is obtained by calculating the position-wise difference between the internal temperature time series vector and the external temperature time series vector. That is, the internal temperature time series vector and the external temperature time series vector respectively record the changes in temperature inside and outside the negative pressure chamber over time. Through the position-wise difference operation, that is, subtracting the external temperature value from the internal temperature value at the corresponding moment, the temperature difference between the inside and outside of the chamber at each time point can be clearly shown, making the change trend of the temperature difference clearer. For example, during the detection process, the external temperature may gradually increase while the internal temperature first decreases and then increases. The internal-external temperature difference time series vector obtained by the position-wise difference can intuitively present the specific changes of this temperature difference at different times, providing an intuitive data basis for subsequent analysis.

[0039] In step S4213, based on the internal pressure time series vector and the external pressure time series vector, an internal-external pressure difference time series vector is determined. Correspondingly, considering that the pressure difference is the direct driving force for gas flow, in the fabric breathability detection, the pressure difference between the inside and outside of the negative pressure chamber directly affects the flow rate and flow of the gas through the fabric. The internal pressure time series vector and the external pressure time series vector respectively reflect the changes in their respective pressures over time. Based on this, in the technical solution of the present application, the internal-external pressure difference time series vector is determined based on the internal pressure time series vector and the external pressure time series vector. In this way, by determining the internal-external pressure difference time series vector, the magnitude and change trend of the pressure difference at different times can be intuitively presented, so as to better analyze the influence of pressure changes on fabric breathability, thereby improving the accuracy of gas flow value correction.

[0040] In step S422, temporal feature extraction is performed on the internal-external temperature difference time series vector and the internal-external pressure difference time series vector to obtain the internal-external temperature difference time series gradient feature vector and the internal-external pressure difference time series gradient feature vector. Specifically, in the embodiment of the present application, step S422 includes: passing the internal-external temperature difference time series vector and the internal-external pressure difference time series vector through a temporal feature extractor based on a multi-scale causal convolutional neural network to obtain the internal-external temperature difference time series gradient feature vector and the internal-external pressure difference time series gradient feature vector. It should be understood that the internal-external temperature difference and pressure difference time series vectors contain information at different time scales, which are difficult to comprehensively capture by traditional methods, that is, there are temporal variations in the short term and periodic changes in the long term, etc. Therefore, in the technical solution of the present application, the internal-external temperature difference time series vector and the internal-external pressure difference time series vector are passed through a temporal feature extractor based on a multi-scale causal convolutional neural network to simultaneously capture the features at these different time scales, and the internal-external temperature difference time series gradient feature vector and the internal-external pressure difference time series gradient feature vector are obtained. That is to say, the multi-scale causal convolutional neural network can process the time series vector at different convolutional kernel scales. Specifically, larger-scale convolutional kernels can capture long-period, macroscopic change trends, such as the influence caused by slow environmental changes during the detection process; smaller-scale convolutional kernels focus on short-period, local detail changes, such as the pressure fluctuations caused by the instant of air extraction. Through multi-scale operations, data features can be comprehensively mined from multiple angles without missing key information. And causal convolution ensures that when the network processes the data at the current moment, it only depends on the information at the past moment, which conforms to the causal logic of the actual physical process. In the fabric breathability detection, the internal-external temperature difference and pressure difference at the current moment are only affected by the past moment. In this way, the obtained internal-external temperature difference time series gradient feature vector and internal-external pressure difference time series gradient feature vector contain richer and deeper information. These features can more accurately reflect the influence law of temperature and pressure changes on gas flow.

[0041] In step S423, interaction response encoding driven by the core temporal features of the temperature difference-pressure difference is performed on the internal-external temperature difference time series gradient feature vector and the internal-external pressure difference time series gradient feature vector to obtain the temperature difference-pressure difference time series gradient interaction response encoding vector. Specifically, Figure 5 It is a flowchart of step S423 in the fabric breathability detection method based on deep learning according to the embodiment of the present application. As Figure 5As shown in the figure, step S423 includes: S4231, constructing the core time series anchor features of the internal and external temperature difference time series gradient feature vector and the internal and external pressure difference time series gradient feature vector to obtain the internal and external temperature difference time series gradient feature core time series anchor coding vector and the internal and external pressure difference time series gradient feature core time series anchor coding vector; S4232, performing temperature difference-pressure difference feature granularity response interaction coding on the internal and external temperature difference time series gradient feature core time series anchor coding vector and the internal and external pressure difference time series gradient feature core time series anchor coding vector to obtain the internal and external temperature difference-pressure difference time series gradient feature granularity response interaction coding vector;

[0042] S4233, performing temperature difference-pressure difference feature value granularity response interaction coding on the internal and external temperature difference time series gradient feature core time series anchor coding vector and the internal and external pressure difference time series gradient feature core time series anchor coding vector to obtain the internal and external temperature difference-pressure difference time series gradient feature value granularity response interaction coding vector; S4234, performing a concatenation process on the internal and external temperature difference-pressure difference time series gradient feature granularity response interaction coding vector and the internal and external temperature difference-pressure difference time series gradient feature value granularity response interaction coding vector to obtain the temperature difference-pressure difference time series gradient interaction response coding vector.

[0043] It should be understood that during the fabric breathability detection process, the temperature difference and pressure difference do not affect the gas flow in isolation, but interact with and influence each other. The internal and external temperature difference time series gradient feature vector and the internal and external pressure difference time series gradient feature vector respectively contain the change trends and change rate information of the temperature difference and pressure difference over time. However, analyzing these two vectors separately cannot reveal their internal non-linear interaction relationship. Based on this, in the technical solution of this application, temperature difference-pressure difference core time series feature-driven interaction response coding is performed on the internal and external temperature difference time series gradient feature vector and the internal and external pressure difference time series gradient feature vector to obtain the temperature difference-pressure difference time series gradient interaction response coding vector. In particular, this method uses the autocorrelation decoupling technology to extract the core anchor point information from the internal and external temperature difference and pressure difference time series gradient feature vectors, and through the two-level interaction modeling of feature granularity and feature value granularity, deeply explores the core interaction mode of the temperature difference and pressure difference at different times, and finds the key factor combination that affects the gas flow. Finally, through feature fusion, a highly compressed and structured feature interaction coding vector is generated to effectively integrate key information and lay a solid foundation for the subsequent accurate correction of gas flow.

[0044] Specifically, in the embodiment of this application, step S4231 includes: constructing the semantic autocorrelation association matrix of the internal and external temperature difference time series gradient feature vector and the internal and external pressure difference time series gradient feature vector to obtain the internal and external temperature difference time series gradient feature semantic autocorrelation association matrix and the internal and external pressure difference time series gradient feature semantic autocorrelation association matrix. This process can be expressed by the formula:

[0045]

[0046] Among them, v 1 is the time - series gradient feature vector of the internal - external temperature difference, v 2 is the time - series gradient feature vector of the internal - external pressure difference, T is the transpose operation, and φ(·) is the feature mapping function, such as a linear mapping or a non - linear kernel function, M 1 is the semantic self - correlation association matrix of the time - series gradient feature of the internal - external temperature difference, M 2 is the semantic self - correlation association matrix of the time - series gradient feature of the internal - external pressure difference;

[0047] Perform core time - series anchoring of self - correlation decoupling on the semantic self - correlation association matrix of the time - series gradient feature of the internal - external temperature difference and the semantic self - correlation association matrix of the time - series gradient feature of the internal - external pressure difference respectively to obtain the core time - series anchoring coding vector of the time - series gradient feature of the internal - external temperature difference and the core time - series anchoring coding vector of the time - series gradient feature of the internal - external pressure difference. This process can be expressed by the formula:

[0048]

[0049] Among them, M 1 is the semantic self - correlation association matrix of the time - series gradient feature of the internal - external temperature difference, M 2 is the semantic self - correlation association matrix of the time - series gradient feature of the internal - external pressure difference, f Anchor (·) is the core time - series anchoring operation, decouple(·) is the feature decoupling operation, x 11 , x 12 , x 1i and x 1n are respectively the row vectors of M 1 , W 1i and b 1i are respectively the weight matrix of the time - series gradient of the internal - external temperature difference and the bias vector of the time - series gradient of the internal - external temperature difference corresponding to x 1i , is matrix multiplication, is the scoring weight vector of the time - series gradient of the internal - external temperature difference, e 1i is the i - th key factor of the time - series gradient feature of the internal - external temperature difference in the set of key factors of the time - series gradient feature of the internal - external temperature difference, Sigmoid is the normalization function, a 1i is the i - th normalized key factor of the time - series gradient feature of the internal - external temperature difference in the set of normalized key factors of the time - series gradient feature of the internal - external temperature difference, n is the number of row vectors in M 1 , c 1 is the core time - series anchoring coding vector of the time - series gradient feature of the internal - external temperature difference, x 21 , x 22 , x 2i and x 2m are respectively the row vectors of M 2 , W 2i and b 2iare x respectively 2i the corresponding internal and external pressure difference time-sequence gradient weight matrix and the internal and external pressure difference time-sequence gradient bias vector is the internal and external pressure difference time-sequence gradient scoring weight vector, e 2i is the i-th internal and external pressure difference time-sequence gradient feature key factor in the set of internal and external pressure difference time-sequence gradient feature key factors, a 2i is the i-th normalized internal and external pressure difference time-sequence gradient feature key factor in the set of normalized internal and external pressure difference time-sequence gradient feature key factors, and m is M 2 the number of row vectors in the middle, and the quantities of m and n are equal, c 2 is the internal and external pressure difference time-sequence gradient feature core time-sequence anchoring coding vector

[0050] It should be understood that the time-sequence gradient feature vectors of the internal and external temperature difference and the internal and external pressure difference respectively contain the trend and rate information of the temperature difference and the pressure difference changing with time, but the mutual correlation between these information is not intuitive. Through the semantic self-correlation association matrix, the mutual correlation between the components in the feature vector can be made explicit. That is, the generated internal and external temperature difference time-sequence gradient feature semantic self-correlation association matrix and the internal and external pressure difference time-sequence gradient feature semantic self-correlation association matrix not only retain the global structure of the original corresponding respective feature vectors, that is, the overall change trend, but also capture the internal semantic patterns, such as the similarity and periodicity of the temperature difference or pressure difference changes in certain time periods. This can provide a more comprehensive and clearer data basis for further mining the core information in the original vector

[0051] Accordingly, considering that the generated internal and external temperature difference time-sequence gradient feature semantic self-correlation association matrix and the internal and external pressure difference time-sequence gradient feature semantic self-correlation association matrix may contain a large amount of redundant information, and this information is not helpful for accurately analyzing the influence of temperature and pressure on gas flow, and may even interfere with the analysis results. Based on this, it is necessary to perform the core time-sequence anchoring operation of self-correlation decoupling on these two semantic self-correlation association matrices respectively to refine and purify the internal information of the internal and external temperature difference time-sequence gradient semantic features and the internal and external pressure difference time-sequence gradient semantic features. That is, in the process of the model analyzing data, find the temperature and pressure change features that really have a key impact on gas flow, and filter out the irrelevant interference information. Generally speaking, the generated internal and external temperature difference time-sequence gradient feature core time-sequence anchoring coding vector and the internal and external pressure difference time-sequence gradient feature core time-sequence anchoring coding vector are more refined, only retaining the key information related to the core semantics of the internal and external temperature difference and the internal and external pressure difference, and removing the redundancy and noise, which can make the subsequent analysis focus more on the key factors, thereby improving the efficiency and accuracy of the analysis

[0052] Next, perform temperature-difference - pressure-difference feature granularity response interaction encoding on the core temporal anchor encoding vectors of the internal and external temperature difference temporal gradient features and the core temporal anchor encoding vectors of the internal and external pressure difference temporal gradient features to obtain the internal and external temperature difference - pressure difference temporal gradient feature granularity response interaction encoding vector. The above process can be expressed by the formula:

[0053]

[0054] Among them, c 1 is the core temporal anchor encoding vector of the internal and external temperature difference temporal gradient features, c 2 is the core temporal anchor encoding vector of the internal and external pressure difference temporal gradient features, is addition by position point, W VT and b VT are the response weight matrix and the response bias vector respectively, tanh is the hyperbolic tangent function, and E granular is the internal and external temperature difference - pressure difference temporal gradient feature granularity response interaction encoding vector. It should be understood that at the feature granularity level, there may be complex interactions between different semantic units of the temperature difference and the pressure difference (such as the change trends and change amplitudes in different time periods, etc.). To further analyze the interaction relationship between the temperature difference and the pressure difference, in this application, it is necessary to perform temperature-difference - pressure-difference feature granularity response interaction encoding on the core temporal anchor encoding vectors of the internal and external temperature difference temporal gradient features and the core temporal anchor encoding vectors of the internal and external pressure difference temporal gradient features to explore the response relationship of the two core temporal anchor encoding vectors at the semantic unit level, and further capture the semantic coupling at the microscopic level of the temperature difference and the pressure difference. That is, the obtained internal and external temperature difference - pressure difference temporal gradient feature granularity response interaction encoding vector records the interaction information of the temperature difference and the pressure difference at the feature granularity level, reflects their interaction patterns at different semantic units, can help the model deeply understand the complex interaction patterns between temperature and pressure, and can provide a reliable basis for subsequent gas flow correction.

[0055] Then, perform temperature-difference - pressure-difference feature value granularity response interaction encoding on the core temporal anchor encoding vectors of the internal and external temperature difference temporal gradient features and the core temporal anchor encoding vectors of the internal and external pressure difference temporal gradient features to obtain the internal and external temperature difference - pressure difference temporal gradient feature value granularity response interaction encoding vector. The above process can be expressed by the formula:

[0056]

[0057] Among them, c 1 is the core temporal anchor encoding vector of the internal and external temperature difference temporal gradient features, c 2 is the core temporal anchor encoding vector of the internal and external pressure difference temporal gradient features, and E valueIt is the interaction coding vector of the internal-external temperature difference-pressure difference time-sequence gradient eigenvalue granularity response. It should be understood that in addition to the interaction at the feature granularity level, the interdependence between the eigenvalues themselves also has an important impact on the gas flow rate. By performing interaction coding at the eigenvalue granularity level on the core time-sequence anchored coding vector of the internal-external temperature difference time-sequence gradient feature and the core time-sequence anchored coding vector of the internal-external pressure difference time-sequence gradient feature, the interdependence between the temperature difference and pressure difference eigenvalues can be characterized numerically. Specifically, through the element-wise division operation, it helps the model to deeply analyze the specific dependence between the temperature difference and pressure difference eigenvalues, supplementing the deficiencies in the analysis at the feature granularity level. This data processing method effectively enriches the representation dimension of the interaction modeling between the temperature difference and pressure difference features, enabling the model to analyze the interaction relationship between these two data more comprehensively and deeply.

[0058] Finally, the interaction coding vector of the internal-external temperature difference-pressure difference time-sequence gradient eigenvalue granularity response and the interaction coding vector of the internal-external temperature difference-pressure difference time-sequence gradient eigenvalue granularity response are cascaded to obtain the temperature difference-pressure difference time-sequence gradient interaction response coding vector. The above process can be expressed by the formula:

[0059] V f =concat[E granular ;E value

[0060] where E granular is the interaction coding vector of the internal-external temperature difference-pressure difference time-sequence gradient eigenvalue granularity response, E value is the interaction coding vector of the internal-external temperature difference-pressure difference time-sequence gradient eigenvalue granularity response, concat[·;·] is the cascading operation, and V f is the temperature difference-pressure difference time-sequence gradient interaction response coding vector.

[0061] It should be understood that the previous steps have analyzed the interaction relationship between the temperature difference and pressure difference from the feature granularity and eigenvalue granularity respectively, but this information is scattered. To comprehensively and synthetically reflect the influence of the temperature difference and pressure difference on the gas flow rate, it is necessary to effectively integrate the interaction coding vector of the internal-external temperature difference-pressure difference time-sequence gradient eigenvalue granularity response and the interaction coding vector of the internal-external temperature difference-pressure difference time-sequence gradient eigenvalue granularity response. Specifically, the cascading operation can combine these two vectors into a temperature difference-pressure difference time-sequence gradient interaction response coding vector, enabling the multi-granularity information to be expressed in this vector. This can avoid the dispersion and loss of information, and thus can better provide a complete and comprehensive data representation for the subsequent accurate correction of the gas flow rate.

[0062] ​Preferably, in another example of the present application, the internal and external temperature difference - pressure difference time - sequence gradient feature - granularity response interaction coding vector and the internal and external temperature difference - pressure difference time - sequence gradient feature - value - granularity response interaction coding vector are cascaded to obtain the temperature difference - pressure difference time - sequence gradient interaction response coding vector, including: performing interaction - distribution growth perturbation correction on the internal and external temperature difference - pressure difference time - sequence gradient feature - granularity response interaction coding vector and the internal and external temperature difference - pressure difference time - sequence gradient feature - value - granularity response interaction coding vector to obtain the corrected internal and external temperature difference - pressure difference time - sequence gradient feature - granularity response interaction coding vector and the corrected internal and external temperature difference - pressure difference time - sequence gradient feature - value - granularity response interaction coding vector; performing cascading on the corrected internal and external temperature difference - pressure difference time - sequence gradient feature - granularity response interaction coding vector and the corrected internal and external temperature difference - pressure difference time - sequence gradient feature - value - granularity response interaction coding vector to obtain the temperature difference - pressure difference time - sequence gradient interaction response coding vector. This process is represented by the following formula:

[0063]

[0064] where, E gi is the eigenvalue at the i - th position in E granular ; E vi is the eigenvalue at the i - th position in E value ; cos is the cosine function; E′ gi is the corrected eigenvalue of E gi ; E′ vi is the corrected eigenvalue of E vi ; E′ granular is the corrected internal and external temperature difference - pressure difference time - sequence gradient feature - granularity response interaction coding vector; E′ value is the corrected internal and external temperature difference - pressure difference time - sequence gradient feature - value - granularity response interaction coding vector; V f is the temperature difference - pressure difference time - sequence gradient interaction response coding vector.

[0065] Here, considering the differences between the modeling representations of the internal and external temperature difference - pressure difference time - sequence gradient feature - granularity interaction and the internal and external temperature difference - pressure difference time - sequence gradient feature - value - granularity interaction may cause unstable perturbations in the feature manifold interface of the temperature difference - pressure difference time - sequence gradient interaction response coding vector with multi - granularity information expression after fusion. Taking the internal and external temperature difference - pressure difference time - sequence gradient feature - value - granularity interaction representation as an extensibility - constraint representation, for the overall feature interaction distribution growth under the per - eigenvalue diffusion process, based on the growth - index representation under extensibility constraints, the interface - shape perturbation deviation of the overall distribution of the internal and external temperature difference - pressure difference time - sequence gradient feature - granularity interaction is modeled. That is, taking the interaction - interface gradient under per - eigenvalue granularity as the perturbation - contribution factor, the growth - index stabilization contribution of gradient diffusion under eigenvalue interdependence is determined to achieve growth - mode gradient correction dominated by perturbation stabilization and improve the manifold - interface stability of the temperature difference - pressure difference time - sequence gradient interaction response coding vector.

[0066] In step S424, based on the temperature difference-pressure difference time series gradient interaction response coding vector, the gas flow value is corrected to obtain a gas flow correction value. Specifically, in Figure 6 FIG. 4 is a flowchart of step S424 in the fabric air permeability detection method based on deep learning according to an embodiment of the present application. As Figure 6 shown, step S424 includes: S4241, passing the temperature difference-pressure difference time series gradient interaction response coding vector through a gas flow corrector based on a decoder to obtain a gas flow correction coefficient; S4242, based on the gas flow correction coefficient, correcting the gas flow value to obtain a gas flow correction value.

[0067] In step S4241, the temperature difference-pressure difference time series gradient interaction response coding vector is passed through a gas flow corrector based on a decoder to obtain a gas flow correction coefficient. That is, the temperature difference-pressure difference time series gradient interaction response coding vector obtained by performing core time series interaction response using the internal and external temperature difference time series gradient feature vector and the internal and external pressure difference time series gradient feature vector is decoded, so as to intelligently generate a gas flow correction coefficient. It should be understood that the gas flow corrector based on a decoder has a powerful mapping ability, and can decode the complex information in the coding vector and convert it into a correction coefficient directly related to the gas flow. The decoder can establish a complex mapping model by learning the relationship between the coding vector and the actual gas flow correction requirement, and realize the conversion from the abstract coding to the practical correction coefficient. Specifically, in the embodiment of the present application, passing the temperature difference-pressure difference time series gradient interaction response coding vector through a gas flow corrector based on a decoder to obtain a gas flow correction coefficient includes: multiplying the decoding weight matrix of the decoder by the temperature difference-pressure difference time series gradient interaction response coding vector to obtain a temperature difference-pressure difference time series gradient interaction response decoding vector, and accumulating and summing all the eigenvalues of the temperature difference-pressure difference time series gradient interaction response decoding vector to obtain a gas flow correction coefficient.

[0068] In step S4242, the gas flow value is corrected based on the gas flow correction coefficient to obtain the corrected gas flow value. That is, in the actual detection process, due to the influence of factors such as the temperature difference, pressure difference between the inside and outside of the negative pressure chamber, and other environmental factors, the directly measured gas flow value often has errors. These errors will lead to inaccurate calculation of the fabric air permeability rate, and cannot truly reflect the air permeability performance of the fabric. The gas flow correction coefficient is obtained by comprehensively considering various influencing factors. By correcting the measured flow value with it, these errors can be compensated, making the measurement result closer to the true gas flow. In this way, the fabric air permeability rate calculated based on the corrected flow can more truly reflect the air permeability performance of the fabric. For example, when the fabric is detected under different environmental conditions, the corrected gas flow can eliminate the interference of environmental factors, making the calculated air permeability rate comparable, thereby improving the accuracy of the fabric air permeability rate calculation.

[0069] The following is a detailed elaboration of a specific implementation process of "correcting the gas flow value based on the gas flow correction coefficient to obtain the corrected gas flow value":

[0070] First, directly read the gas flow value passing through the fabric to be tested per unit time by a standard air permeability testing device. For example, in a specific example, assume that Q_raw = 500 L / min is obtained as the initially measured gas flow value. This value represents the volume of air passing through the fabric per unit time without considering any environmental impacts. However, as mentioned before, due to temperature and pressure differences that may cause changes in gas density and flow state, this value may not fully and accurately reflect the true air permeability performance of the fabric.

[0071] Next is the process of applying the gas flow correction coefficient to adjust the original gas flow value. The core idea here is to introduce the correction coefficient to compensate for the measurement deviation caused by changes in environmental conditions. Specifically, if a correction coefficient C = 0.95 has been determined, then according to the formula Q_corrected = Q_raw * C, the corrected gas flow value Q_corrected = 500 * 0.95 = 475 L / min can be calculated. This means that considering all known environmental factors, the true gas flow should be slightly lower than the initially measured value. Such correction not only helps to improve the accuracy of a single measurement, but also can significantly enhance the consistency and reliability of the overall air permeability assessment when applied to multiple samples or multiple repeated measurements.

[0072] To verify the effectiveness of the calibration effect, it is usually necessary to compare the gas flow values before and after calibration. In this process, it is important not only to focus on the numerical changes, but also to analyze whether these changes reasonably reflect the expected physical phenomena. For example, if the calibrated gas flow value is indeed closer to the results obtained by other independent methods (such as theoretical calculations or repeated experiments under different conditions), then the calibration strategy can be considered successful. In addition, this calibration method also allows fine-tuning of the calibration coefficient according to the actual situation when necessary to further optimize the measurement accuracy. For instance, if it is found that certain specific types of fabrics exhibit trends different from the normal in a specific environment, the model parameters can be adjusted or the algorithm can be retrained to adapt to the new situation.

[0073] In summary, step S4 is clearly described. It uses data analysis and coding techniques based on deep learning to regularize the {internal temperature data, internal pressure data, external temperature data, and external pressure data} collected inside and outside the negative pressure chamber and calculate the internal and external temporal temperature difference and pressure difference. Then, temporal gradient features are extracted from the calculated internal and external temporal temperature difference and pressure difference. Based on the core temporal interaction response representation between the internal and external temperature difference temporal gradient features and the internal and external pressure difference temporal gradient features obtained after temporal extraction, a gas flow calibration coefficient is intelligently generated, and the gas flow value is calibrated based on this calibration coefficient. In this way, by monitoring the temperature and pressure data inside and outside the negative pressure chamber and calculating the internal and external temperature difference and pressure difference, the measurement error of the gas flow caused by the temperature gradient and pressure gradient can be effectively compensated. This makes the gas flow measurement closer to the true value, thereby improving the accuracy of the measured air permeability and enhancing the credibility and precision of the final test.

[0074] In step S5, the measured air permeability of the fabric is calculated based on the gas flow calibration value. It should be understood that during the air extraction process, the temperature gradient and pressure gradient inside and outside the negative pressure chamber will interfere with the gas flow measurement, resulting in inaccurate directly measured flow values. By calibrating the gas flow value and removing the influence of these interference factors, calculating the measured air permeability based on the more accurate calibrated gas flow value can make the obtained result closer to the true air permeability of the fabric, providing a reliable basis for further correcting the air permeability subsequently and thus enhancing the accuracy and credibility of the entire fabric air permeability detection. Specifically, dividing the gas flow calibration value per unit time in the exhaust pipe by the area of the area to be detected and then by the air pressure difference on both sides of the fabric can obtain the measured air permeability of the fabric.

[0075] In step S6, based on the bulge height of the fabric, the measured air permeability of the fabric is corrected to obtain a corrected air permeability. It should be understood that in the fabric air permeability test, the fabric will bulge due to the pressure difference on both sides during the test, resulting in an increase in its actual area, and the gaps on the fabric will also become larger, thereby making the air permeability during the test greater than the air permeability under normal conditions. If the air permeability of the fabric is evaluated only based on the measured air permeability, the result will be too large and inaccurate. The measured air permeability is corrected based on the bulge height of the fabric in order to eliminate the influence of the area change caused by the bulge of the fabric on the air permeability measurement, and obtain a corrected air permeability that is closer to the actual air permeability of the fabric, thereby improving the accuracy and reliability of the test results, so that the test results can more truly reflect the air permeability of the fabric in actual use. Specifically, firstly, a model is established based on a neural network algorithm to predict the change of the surface area of ​​the fabric at different uplift heights. The input features of the model include at least parameters such as the elastic coefficient, thickness value and surface area of ​​the fabric in the non-uplifted state. Then, the uplift height data measured by multiple uplift height detection devices (such as laser rangefinders) arranged in the negative pressure chamber are input into the fabric surface area change prediction model to obtain the surface area A1 of the fabric after uplift under the current air pressure conditions. Then, the ratio of the surface area A2 of the fabric before uplift to the surface area A1 of the fabric after uplift is taken as the correction coefficient S. Finally, the correction coefficient S is used to correct the measured air permeability K according to the formula (K1=S×K) to obtain the corrected air permeability K1.

[0076] In summary, the fabric air permeability detection method based on deep learning according to the embodiment of the present application is explained, which first places the fabric flat on the bottom plate and clamps it with a negative pressure chamber, then extracts air through the exhaust pipe to stabilize the air pressure in the chamber, and then measures the fabric bulge height and the gas flow value per unit time in the exhaust pipe when the air pressure is stable, and then corrects the measured gas flow value to obtain the gas flow correction value, and calculates the measured air permeability of the fabric accordingly, and finally corrects the measured air permeability according to the height of the fabric bulge to obtain the corrected air permeability. In this way, the accuracy of the measured air permeability result can be ensured, thereby improving the reliability of the corrected air permeability result, thereby effectively improving the accuracy of fabric air permeability detection. Figure 7 : is a system block diagram of a fabric air permeability detection system based on deep learning according to an embodiment of the present application. Figure 7As shown, the deep learning-based fabric air permeability detection system 100 according to an embodiment of the present application includes: a fabric clamping module 110 for laying the fabric flat on a bottom plate and clamping the fabric using a negative pressure chamber; a gas pressure stabilization module 120 for pumping air into the negative pressure chamber through an air extraction pipe to keep the gas pressure in the negative pressure chamber stable; a data detection module 130 for detecting the bulging height of the fabric when the gas pressure is stable and measuring the gas flow rate value per unit time in the air extraction pipe; a flow rate correction module 140 for correcting the gas flow rate value to obtain a corrected gas flow rate value; an air permeability calculation module 150 for calculating the measured air permeability of the fabric based on the corrected gas flow rate value; and an air permeability correction module 160 for correcting the measured air permeability of the fabric based on the bulging height of the fabric to obtain a corrected air permeability.

[0077] Here, those skilled in the art can understand that the specific functions and operations of each unit and module in the above deep learning-based fabric air permeability detection system 100 have been introduced in detail in the description of the Figures 1 to 6 deep learning-based fabric air permeability detection method above, and thus, the repeated description thereof will be omitted.

[0078] In summary, the deep learning-based fabric air permeability detection system 100 according to an embodiment of the present application is elucidated. First, the fabric is laid flat on the bottom plate and clamped by a negative pressure chamber. Then, air is extracted through an air extraction pipe to make the gas pressure in the chamber stable. Next, when the gas pressure is stable, the bulging height of the fabric and the gas flow rate value per unit time in the air extraction pipe are measured. Immediately afterwards, the measured gas flow rate value is corrected to obtain a corrected gas flow rate value, and based on this, the measured air permeability of the fabric is calculated. Finally, the measured air permeability is corrected according to the bulging height of the fabric to obtain a corrected air permeability. In this way, the accuracy of the measured air permeability result can be ensured, and further, the reliability of the corrected air permeability result can be improved, thereby effectively improving the accuracy of fabric air permeability detection.

Claims

1. A fabric air permeability detection method based on deep learning, characterized in that: include: Laying the fabric flat on a bottom plate and clamping the fabric using a negative pressure chamber; Extracting air into the negative pressure chamber through an air extraction pipe so that the air pressure in the negative pressure chamber remains stable; Detecting the bulge height of the fabric when the air pressure is stable, and measuring the gas flow value per unit time in the air extraction pipe; Correcting the gas flow value to obtain a gas flow correction value, including: monitoring and collecting {internal temperature data, internal pressure data, external temperature data, and external pressure data} through a first sensor group deployed inside the negative pressure chamber and a second sensor group deployed outside the negative pressure chamber; performing interactive response based on the temperature difference-pressure difference core timing feature on the {internal temperature data, internal pressure data, external temperature data, and external pressure data} to obtain the gas flow correction value; Calculating the measured air permeability of the fabric based on the gas flow correction value; Based on the ridge height of the fabric, the measured air permeability of the fabric is corrected to obtain a corrected air permeability.

2. The fabric air permeability detection method based on deep learning according to claim 1 is characterized in that: The gas flow correction value is obtained by interactively responding the {internal temperature data, internal pressure data, external temperature data, and external pressure data} based on the temperature difference-pressure difference core timing characteristic, including: Based on the {internal temperature data, internal pressure data, external temperature data and external pressure data}, determine an internal and external temperature difference time series vector and an internal and external pressure difference time series vector; Performing time series feature extraction on the internal and external temperature difference time series vector and the internal and external pressure difference time series vector to obtain an internal and external temperature difference time series gradient feature vector and an internal and external pressure difference time series gradient feature vector; Performing temperature difference-pressure difference core timing feature driven interactive response coding on the internal and external temperature difference time series gradient feature vector and the internal and external pressure difference time series gradient feature vector to obtain a temperature difference-pressure difference time series gradient interactive response coding vector; Based on the temperature difference-pressure difference time series gradient interaction response encoding vector, the gas flow value is corrected to obtain the gas flow correction value.

3. The fabric air permeability detection method based on deep learning according to claim 2 is characterized in that: Based on the {internal temperature data, internal pressure data, external temperature data and external pressure data}, determining the internal and external temperature difference time series vector and the internal and external pressure difference time series vector includes: Regularize the {internal temperature data, internal pressure data, external temperature data and external pressure data} to obtain an internal temperature time series vector, an internal pressure time series vector, an external temperature time series vector and an external pressure time series vector; Determining the internal and external temperature difference time series vector based on the internal temperature time series vector and the external temperature time series vector; The internal-external pressure difference time series vector is determined based on the internal pressure time series vector and the external pressure time series vector.

4. The fabric air permeability detection method based on deep learning according to claim 3 is characterized in that: Determining the internal-external temperature difference time series vector based on the internal temperature time series vector and the external temperature time series vector includes: calculating the position difference between the internal temperature time series vector and the external temperature time series vector to obtain the internal-external temperature difference time series vector.

5. The fabric air permeability detection method based on deep learning according to claim 4 is characterized in that: The internal and external temperature difference time series vector and the internal and external pressure difference time series vector are subjected to time series feature extraction to obtain the internal and external temperature difference time series gradient feature vector and the internal and external pressure difference time series gradient feature vector, including: passing the internal and external temperature difference time series vector and the internal and external pressure difference time series vector through a time series feature extractor based on a multi-scale causal convolutional neural network to obtain the internal and external temperature difference time series gradient feature vector and the internal and external pressure difference time series gradient feature vector.

6. The fabric air permeability detection method based on deep learning according to claim 2 is characterized in that: The internal and external temperature difference time series gradient feature vector and the internal and external pressure difference time series gradient feature vector are subjected to temperature difference-pressure difference core time series feature driven interactive response coding to obtain a temperature difference-pressure difference time series gradient interactive response coding vector, including: Constructing the core temporal anchoring features of the internal and external temperature difference temporal gradient feature vector and the internal and external pressure difference temporal gradient feature vector to obtain the core temporal anchoring coding vector of the internal and external temperature difference temporal gradient feature and the core temporal anchoring coding vector of the internal and external pressure difference temporal gradient feature; Performing temperature difference-pressure difference feature granularity response interactive coding on the internal and external temperature difference time series gradient feature core time series anchor coding vector and the internal and external pressure difference time series gradient feature core time series anchor coding vector to obtain an internal and external temperature difference-pressure difference time series gradient feature granularity response interactive coding vector; Performing temperature difference-pressure difference characteristic value granularity response interactive coding on the internal and external temperature difference time series gradient characteristic core time series anchor coding vector and the internal and external pressure difference time series gradient characteristic core time series anchor coding vector to obtain an internal and external temperature difference-pressure difference time series gradient characteristic value granularity response interactive coding vector; The internal and external temperature difference-pressure difference time series gradient characteristic granularity response interactive coding vector and the internal and external temperature difference-pressure difference time series gradient characteristic value granularity response interactive coding vector are cascaded to obtain the temperature difference-pressure difference time series gradient interactive response coding vector.

7. The fabric air permeability detection method based on deep learning according to claim 6 is characterized in that: Constructing the core temporal anchoring features of the internal and external temperature difference temporal gradient feature vector and the internal and external pressure difference temporal gradient feature vector to obtain the core temporal anchoring coding vector of the internal and external temperature difference temporal gradient feature and the core temporal anchoring coding vector of the internal and external pressure difference temporal gradient feature, including: Constructing the semantic autocorrelation association matrix of the internal and external temperature difference time series gradient feature vector and the internal and external pressure difference time series gradient feature vector to obtain the internal and external temperature difference time series gradient feature semantic autocorrelation association matrix and the internal and external pressure difference time series gradient feature semantic autocorrelation association matrix; The semantic autocorrelation association matrix of the internal and external temperature difference time series gradient feature and the semantic autocorrelation association matrix of the internal and external pressure difference time series gradient feature are respectively subjected to core time series anchoring by autocorrelation decoupling to obtain the core time series anchoring coding vector of the internal and external temperature difference time series gradient feature and the core time series anchoring coding vector of the internal and external pressure difference time series gradient feature.

8. The fabric air permeability detection method based on deep learning according to claim 7 is characterized in that: The internal and external temperature difference-pressure difference time series gradient characteristic granularity response interactive coding vector and the internal and external temperature difference-pressure difference time series gradient characteristic value granularity response interactive coding vector are cascaded to obtain the temperature difference-pressure difference time series gradient interactive response coding vector, including: Performing interactive distribution growth disturbance correction on the internal and external temperature difference-pressure difference time series gradient characteristic granularity response interactive coding vector and the internal and external temperature difference-pressure difference time series gradient characteristic value granularity response interactive coding vector to obtain a corrected internal and external temperature difference-pressure difference time series gradient characteristic granularity response interactive coding vector and a corrected internal and external temperature difference-pressure difference time series gradient characteristic value granularity response interactive coding vector; The corrected internal and external temperature difference-pressure difference time series gradient characteristic granularity response interactive coding vector and the corrected internal and external temperature difference-pressure difference time series gradient characteristic value granularity response interactive coding vector are cascaded to obtain the temperature difference-pressure difference time series gradient interactive response coding vector.

9. The fabric air permeability detection method based on deep learning according to claim 8, characterized in that: Based on the temperature difference-pressure difference time series gradient interaction response encoding vector, the gas flow value is corrected to obtain the gas flow correction value, including: Passing the temperature difference-pressure difference time series gradient interaction response encoding vector through a decoder-based gas flow corrector to obtain a gas flow correction coefficient; Based on the gas flow correction coefficient, the gas flow value is corrected to obtain the gas flow correction value.

10. A fabric air permeability detection system based on deep learning, characterized in that: include: A fabric clamping module, used for placing the fabric flat on a bottom plate and clamping the fabric using a negative pressure chamber; An air pressure stabilization module, used for pumping air into the negative pressure chamber through an air pumping pipe so that the air pressure in the negative pressure chamber remains stable; A data detection module, used to detect the bulge height of the fabric when the air pressure is stable, and to measure the gas flow value per unit time in the air extraction pipe; A flow correction module is used to correct the gas flow value to obtain a gas flow correction value, wherein the flow correction module is used to: monitor and collect {internal temperature data, internal pressure data, external temperature data and external pressure data} through a first sensor group deployed inside the negative pressure chamber and a second sensor group deployed outside the negative pressure chamber; and perform interactive response based on the temperature difference-pressure difference core timing feature on the {internal temperature data, internal pressure data, external temperature data and external pressure data} to obtain the gas flow correction value; An air permeability calculation module, used to calculate the measured air permeability of the fabric based on the gas flow correction value; The air permeability correction module is used to correct the measured air permeability of the fabric based on the ridge height of the fabric to obtain a corrected air permeability.

Images