One-way valve air tightness detection method and system

Abstract

The invention discloses a one-way valve air tightness detection method and system, and relates to the technical field of air tightness detection.According to the one-way valve air tightness detection method and system, through dynamic interaction of AI virtual sensor prediction and physical detection values and combination of time-frequency domain signal fusion, the extraction capacity of micro-leakage signals is remarkably enhanced, and misjudgment caused by background noise or residual gas in a traditional method is avoided; a multi-mode interference source correlation analysis and dynamic weight distribution mechanism is utilized to actively identify and compensate composite interference such as temperature drift and electromagnetic pulse, so that the reliability of a detection result in severe scenes such as workshop vibration and airflow disturbance is improved; and meanwhile, a model adaptation module based on transfer learning and non-contact vibration spectrum monitoring are adopted, manual parameter adjustment dependence is reduced, valve bodies of different specifications are rapidly adapted, early damage of the diaphragm is warned in advance, and the equipment maintenance period is prolonged.

Description

Technical Field

[0001] The present invention relates to the technical field of air tightness detection, and in particular to a one-way valve air tightness detection method and system. Background Art

[0002] In high-precision equipment such as hydrogen fuel cells and aerospace hydraulic systems, one-way valves are key fluid control components, and their airtightness directly affects system safety and reliability. The current industry requirement for leakage rate has been raised to 10^-3~10^-4Pa·m³ / s, equivalent to a leakage of less than 1 ml of gas per hour.

[0003] However, traditional detection methods such as differential pressure method, ammonia mass spectrometry or water detection method face serious bottlenecks when dealing with such ultra-low leakage scenarios: the differential pressure method is limited by sensor resolution and environmental thermodynamic disturbances, and small pressure changes are easily masked by noise; although ammonia mass spectrometry has high sensitivity, the detection cycle is long and there is residual gas interference; the water detection method cannot meet the needs of non-destructive and rapid online detection; more importantly, the above methods have insufficient ability to capture quasi-dynamic leaks (intermittent micro-leaks), and are difficult to adapt to the real-time leakage monitoring needs of precision equipment.

[0004] To cope with complex environmental interference, existing solutions generally adopt multi-sensor fusion technology, integrating high-precision differential pressure sensors, micro-mass flowmeters and environmental monitoring modules, and reducing the risk of misjudgment through synchronous data collection; some companies have introduced AI modeling modules (such as LSTM and Transformer), using historical data to train virtual sensors to predict leakage trends, and compare and correct them with physical sensor data; however, the time alignment accuracy of multimodal data aggregated by such solutions is limited by hardware synchronization errors, resulting in insufficient modeling of the spatiotemporal correlation between environmental disturbance characteristics and leakage signals; AI models rely on static training of historical data, and lack generalization capabilities when faced with new interference sources, such as sudden electromagnetic pulses and non-steady-state airflow, and are prone to over-correction or missed detection; therefore, a one-way valve air tightness detection solution is urgently needed to solve such problems. Summary of the Invention

[0005] In view of the above existing problems, the present invention is proposed.

[0006] The present invention provides a one-way valve air tightness detection method and system to solve the problems faced by existing one-way valve air tightness detection methods in ultra-low leakage scenarios, such as insufficient sensitivity, weak anti-interference ability, and reliance on experience for parameter adjustment. In particular, it is difficult to cope with the problem of temperature / electromagnetic / airflow composite interference and quasi-dynamic leakage identification requirements.

[0007] In order to solve the above technical problems, the present invention provides the following technical solutions: In a first aspect, an embodiment of the present invention provides a method for detecting air tightness of a one-way valve, which comprises: Step S1, real-time acquisition of multimodal sensing data by a physical detection unit, wherein the physical detection unit includes a pressure differential sensor, a micro mass flow meter, an ambient temperature sensor, an airflow disturbance monitoring module, and an electromagnetic interference monitoring module; Step S2, inputting the multimodal sensing data into a pre-trained AI prediction model to generate a virtual leakage prediction value; Step S3, comparing the virtual leakage prediction value with the real-time detection value of the physical detection unit, and dynamically correcting the environmental interference component based on the deviation analysis result; Step S4, determining the air tightness status of the one-way valve according to the corrected leakage rate and a preset threshold; Among them, the AI ​​prediction model is trained based on historical sample data, and the historical sample data includes multimodal sensor data sequences under normal no-leakage state, known leakage state and multiple types of environmental interference scenarios.

[0008] As a preferred solution of the one-way valve air tightness detection method of the present invention, when the deviation between the virtual leakage prediction value and the real-time detection value exceeds a preset threshold, the following operations are performed: a) Extracting real-time data from the ambient temperature sensor, airflow disturbance monitoring module, and electromagnetic interference monitoring module; b) Calculate the correlation between the fluctuation amplitude of each interference source and the leakage signal based on the sliding time window; c) Generate compensation coefficients for interference sources with correlation greater than a set value and correct real-time detection values.

[0009] As a preferred solution of the one-way valve air tightness detection method of the present invention, step S3 includes: When the trend of the virtual leakage prediction value is consistent with the real-time detection value, the leakage signal weight is enhanced; When there is a significant deviation between the two and the environmental interference monitoring module detects abnormal disturbance, it generates an interference compensation coefficient and corrects the real-time detection value; The interference compensation coefficient is dynamically calculated by analyzing the time domain correlation of ambient temperature, airflow disturbance and electromagnetic interference.

[0010] As a preferred embodiment of the one-way valve air tightness detection method of the present invention, in step S3, during the process of dynamically correcting the environmental interference component based on the deviation analysis result, five sub-processes are continuously executed in each sampling period, and the logical order is as follows: 1) Construct instantaneous deviation: After obtaining the physical measured value and the virtual predicted value, calculate the instantaneous deviation. , in, represents the instantaneous deviation, Indicates time The measured value of the leakage rate, Indicates time The virtual leakage prediction value of is the discrete time index; 2) Normalized deviation: Use length The sliding window of is demeaned and normalized to obtain, , in, is the normalized deviation, is the mean deviation within the window, is the standard deviation of the deviation within the window, is the sliding window length; 3) Quantifying interference correlation: For the first Interference source signal Calculate the correlation coefficient, , in, Deviation and interference sources The correlation coefficient of is the lag index within the window, Interference source within the window The standard deviation of Interference source At the moment signal, , is the total number of interference sources; 4) Map the correlation coefficient to weight: The logistic mapping function is used to obtain continuous weights. , in, Interference source The weight of is the mapping slope coefficient, Enable thresholds for correlation; 5) Estimate the compensation coefficient and correct the measured value: Adopt belt The regular closed-form solution estimates the compensation, , , in, Interference source The compensation coefficient, is the regularization coefficient, is the corrected leak rate.

[0011] In a second aspect, the present invention provides a one-way valve air tightness detection system, comprising: include: Physical detection unit, used to collect pressure difference, mass flow, ambient temperature, air flow disturbance and electromagnetic interference data; The physical detection unit also includes: High-frequency vibration sensor, used to collect mechanical vibration signals in the detection environment; A data synchronization module, connected to the physical detection unit, to achieve time alignment of multi-channel signals; An AI modeling module, which includes a pre-trained LSTM or Transformer model, is used to receive multimodal sensor data and output a virtual leakage prediction value; Signal fusion unit, used to compare virtual leakage prediction values ​​with physical detection data, perform interference removal and signal enhancement; The determination module generates a qualification determination result according to the corrected leakage rate.

[0012] As a preferred solution of the one-way valve air tightness detection system of the present invention, the AI ​​modeling module further includes: The transfer learning submodule is used to adapt the historical training model to the detection scenario of the new specification one-way valve; The transfer learning submodule receives the geometric parameters and material properties of the new valve body and dynamically adjusts the model input layer dimensions and hidden layer weights; The dynamic adjustment of the model input layer dimension includes: According to the aperture size and material hardness of the new valve body, it is mapped into a standardized feature vector; The feature vector and the historical model input layer are dimensionally expanded or truncated, retaining the parameter weights.

[0013] As a preferred solution of the one-way valve air tightness detection system of the present invention, the signal fusion unit includes: Timing alignment submodule, used to match the timestamps of virtual leakage prediction values ​​and physical detection values; The weight allocation submodule dynamically adjusts the fusion weight of the virtual signal and the physical signal according to the output of the environmental interference monitoring module.

[0014] As a preferred solution of the one-way valve air tightness detection system of the present invention, the signal fusion unit further fuses the frequency domain characteristics of the vibration signal with the pressure difference data to separate the coupling interference of the vibration noise on the pressure signal; The separation of vibration noise includes: Perform wavelet packet decomposition on the vibration signal to extract the sub-band energy that overlaps with the pressure signal frequency band; When the sub-band energy exceeds a preset ratio, the corresponding frequency band component is deducted from the pressure signal; The signal fusion unit further includes: Dynamic parameter tuning module, used to adaptively adjust the sampling frequency of the differential pressure sensor and the integration time constant of the micro-mass flowmeter according to the interference intensity of the current detection environment; The adaptive adjustment includes: When the electromagnetic interference monitoring module detects that the intensity exceeds the preset limit, the sampling frequency of the differential pressure sensor is increased; When the airflow disturbance amplitude continues to increase, the integration time constant of the micro-mass flowmeter is shortened.

[0015] As a preferred solution of the one-way valve air tightness detection system of the present invention, the signal fusion unit performs vibration pressure difference fusion and coupling interference separation, including: Perform timing alignment on the data in the same window, and synchronize the data with a length of time window, constructing the vibration sequence With pressure difference series : , in, is the vibration signal window vector, is the pressure difference signal window vector, 、 Respectively indicate time Vibration and pressure difference sampling, is the lag index within the window, is the window length, the unit is the number of sampling points, is the current time index; right Perform discrete wavelet packet decomposition to obtain Subband coefficient vector : , in, Sub-band The coefficient vector of Indicates the The wavelet packet transform operator of the sub-bands, is the energy of this frequency band, , is the total number of frequency bands at the wavelet packet level, The coefficient vector Bit; First define the effective frequency band of pressure difference , then reconstruct the mask , like ,but ,otherwise , The coupling strength is then quantified and smoothed to weights: , , in, is the sub-band center frequency, is the lower limit of the pressure difference passband, is the upper limit of the differential pressure passband, For the mask, is the correlation coefficient, is the smoothing weight, is the mapping slope constant; The sub-bands selected by the mask are reconstructed into interference estimates after weight scaling : , in, Sub-band The inverse wavelet packet reconstruction operator, For the moment Vibration disturbance estimation, is the original pressure difference sampling, is the pressure difference value after vibration compensation.

[0016] As a preferred solution of the one-way valve air tightness detection system of the present invention, it further includes: A non-contact vibration monitoring unit, including a laser Doppler vibrometer, is used to collect the vibration spectrum of the one-way valve diaphragm; The determination module further coordinates and corrects the leakage state determination result according to the natural frequency offset of the vibration spectrum; The coordinated amendments include: When the natural frequency deviation of the vibration spectrum exceeds the preset threshold and is consistent with the leakage rate change trend, a leakage judgment is triggered; When the frequency offset is inconsistent with the leakage rate trend, the vibration spectrum judgment result is preferred.

[0017] The beneficial effects of the present invention are as follows: the present invention significantly enhances the ability to extract micro-leakage signals through the dynamic interaction of AI virtual sensor predictions and physical detection values, combined with time-frequency domain signal fusion, thereby avoiding misjudgments caused by background noise or residual gas in traditional methods; utilizing multi-modal interference source correlation analysis and dynamic weight distribution mechanism to actively identify and compensate for complex interference such as temperature drift and electromagnetic pulses, so as to ensure the reliability of detection results in harsh scenarios such as workshop vibration and airflow disturbance; at the same time, adopting a model adaptation module based on transfer learning and non-contact vibration spectrum monitoring to reduce dependence on manual parameter adjustment, quickly adapt to valve bodies of different specifications, and provide early warning of early diaphragm damage, thereby extending the equipment maintenance cycle; through multi-channel synchronous acquisition and adaptive parameter setting, while ensuring accuracy, the single detection cycle is shortened to meet the high throughput requirements of the production line. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0019] Figure 1 Schematic diagram of the flow of the one-way valve air tightness detection method in Example 1.

[0020] Figure 2 This is a schematic diagram of the framework of the one-way valve air tightness detection system in Example 1. DETAILED DESCRIPTION

[0021] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0022] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0023] Secondly, the term "one embodiment" or "embodiment" herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in various places throughout this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive of other embodiments.

[0024] Example 1, with reference to Figure 1 and Figure 2This embodiment provides a method for detecting air tightness of a one-way valve, comprising: Step S1, real-time acquisition of multimodal sensing data through a physical detection unit, the physical detection unit including a pressure differential sensor, a micro mass flow meter, an ambient temperature sensor, an airflow disturbance monitoring module, and an electromagnetic interference monitoring module; Step S2: input the multimodal sensing data into a pre-trained AI prediction model to generate a virtual leakage prediction value; Step S3, comparing the virtual leakage prediction value with the real-time detection value of the physical detection unit, and dynamically correcting the environmental interference component based on the deviation analysis result; When the deviation between the virtual leakage prediction value and the real-time detection value exceeds the preset threshold, the following operations are performed: a) Extracting real-time data from the ambient temperature sensor, airflow disturbance monitoring module, and electromagnetic interference monitoring module; b) Calculate the correlation between the fluctuation amplitude of each interference source and the leakage signal based on the sliding time window; c) Generate compensation coefficients for interference sources with correlation greater than a set value and correct real-time detection values; Step S3 includes: When the trend of the virtual leakage prediction value is consistent with the real-time detection value, the leakage signal weight is enhanced; When there is a significant deviation between the two and the environmental interference monitoring module detects abnormal disturbance, it generates an interference compensation coefficient and corrects the real-time detection value; The interference compensation coefficient is dynamically calculated by analyzing the time domain correlation of ambient temperature, airflow disturbance and electromagnetic interference; In step S3, during the process of dynamically correcting the environmental interference component based on the deviation analysis result, five sub-processes are continuously executed in each sampling period. The logical order is as follows: 1) Construct instantaneous deviation: After obtaining the physical measured value and the virtual predicted value, calculate the instantaneous deviation. , in, represents the instantaneous deviation, Indicates time The measured value of the leakage rate, Indicates time The virtual leakage prediction value of is the discrete time index; 2) Normalized deviation: Use length The sliding window of is demeaned and normalized to obtain, , in, is the normalized deviation, is the mean deviation within the window, is the standard deviation of the deviation within the window, is the sliding window length; 3) Quantifying interference correlation: For the first Interference source signal Calculate the correlation coefficient, , in, Deviation and interference sources The correlation coefficient of is the lag index within the window, Interference source within the window The standard deviation of Interference source At the moment signal, , is the total number of interference sources; 4) Map the correlation coefficient to weight: The logistic mapping function is used to obtain continuous weights. , in, Interference source The weight of is the mapping slope coefficient, Enable thresholds for correlation; 5) Estimate the compensation coefficient and correct the measured value: Adopt belt The regular closed-form solution estimates the compensation, , , in, Interference source The compensation coefficient, is the regularization coefficient, is the corrected leakage rate; Specifically, the normalized deviation The dimensions are unified to make the deviations under different measurement conditions comparable, and the correlation coefficient The coupling strength of slow-changing and sudden-changing interference is captured by window averaging, and the logic function Change the hard threshold into a smooth transition, alleviate the boundary jitter, and compensate the coefficient The regularization term Suppress the coefficient explosion caused by high collinear interference, while maintaining the analytical form to facilitate real-time calculation, and the final correction amount Maintains stable trends even in scenarios with dramatic temperature fluctuations, electromagnetic pulses, or significant airflow disturbances; Step S4, determining the air tightness status of the one-way valve according to the corrected leakage rate and a preset threshold; Among them, the AI ​​prediction model is trained based on historical sample data, which includes multimodal sensor data sequences under normal no-leakage state, known leakage state and multiple types of environmental interference scenarios.

[0025] This embodiment also provides a one-way valve air tightness detection system, comprising: Physical detection unit, used to collect pressure difference, mass flow, ambient temperature, air flow disturbance and electromagnetic interference data; The physical detection unit also includes: High-frequency vibration sensor, used to collect mechanical vibration signals in the detection environment; The data synchronization module is connected to the physical detection unit to achieve time alignment of multi-channel signals; An AI modeling module, which includes a pre-trained LSTM or Transformer model, is used to receive multimodal sensor data and output a virtual leakage prediction value; The AI ​​modeling module also includes: The transfer learning submodule is used to adapt the historical training model to the detection scenario of the new specification one-way valve; The transfer learning submodule receives the geometric parameters and material properties of the new valve body and dynamically adjusts the model input layer dimensions and hidden layer weights; Dynamically adjust the model input layer dimensions including: According to the aperture size and material hardness of the new valve body, it is mapped into a standardized feature vector; Expand or truncate the feature vector and the historical model input layer, preserving the parameter weights; Signal fusion unit, used to compare virtual leakage prediction values ​​with physical detection data, perform interference removal and signal enhancement; The signal fusion unit includes: Timing alignment submodule, used to match the timestamps of virtual leakage prediction values ​​and physical detection values; The weight allocation submodule dynamically adjusts the fusion weight of virtual signals and physical signals according to the output of the environmental interference monitoring module; The signal fusion unit further fuses the frequency domain characteristics of the vibration signal with the pressure difference data to separate the coupling interference of vibration noise on the pressure signal; Separation of vibration noise includes: Perform wavelet packet decomposition on the vibration signal to extract the sub-band energy that overlaps with the pressure signal frequency band; When the sub-band energy exceeds a preset ratio, the corresponding frequency band component is deducted from the pressure signal; The signal fusion unit also includes: Dynamic parameter tuning module, used to adaptively adjust the sampling frequency of the differential pressure sensor and the integration time constant of the micro-mass flowmeter according to the interference intensity of the current detection environment; Adaptive adjustments include: When the electromagnetic interference monitoring module detects that the intensity exceeds the preset limit, the sampling frequency of the differential pressure sensor is increased; When the airflow disturbance amplitude continues to increase, the integration time constant of the micro-mass flowmeter is shortened; In the signal fusion unit, vibration pressure difference fusion and coupling interference separation are performed, including: Perform timing alignment on the data in the same window, and synchronize the data with a length of time window, constructing the vibration sequence With pressure difference series : , in, is the vibration signal window vector, is the pressure difference signal window vector, 、 Respectively indicate time Vibration and pressure difference sampling, is the lag index within the window, is the window length, the unit is the number of sampling points, is the current time index; right Perform discrete wavelet packet decomposition to obtain Subband coefficient vector : , in, Sub-band The coefficient vector of Indicates the The wavelet packet transform operator of the sub-bands, is the energy of this frequency band, , is the total number of frequency bands at the wavelet packet level, The coefficient vector Bit; First define the effective frequency band of pressure difference , then reconstruct the mask , like ,but ,otherwise , The coupling strength is then quantified and smoothed to weights: , , in, is the sub-band center frequency, is the lower limit of the pressure difference passband, is the upper limit of the differential pressure passband, For the mask, is the correlation coefficient, is the smoothing weight, is the mapping slope constant; The sub-bands selected by the mask are reconstructed into interference estimates after weight scaling : , in, Sub-band The inverse wavelet packet reconstruction operator, For the moment Vibration disturbance estimation, is the original pressure difference sampling, is the pressure difference value after vibration compensation; Specifically, the synchronization window ensures that the phase of the vibration and pressure difference signals is consistent, laying a time foundation for the subsequent frequency domain coupling analysis. The wavelet packet decomposition divides the frequency band in the logarithmic scale, which can simultaneously capture the transient impact and steady-state resonance of the vibration. Characterizes the contribution of each frequency band to the vibration intensity of the entire window, mask Filter out irrelevant frequency bands based on the effective bandwidth of the differential pressure sensor to reduce the amount of calculation and the correlation coefficient Accurately measure the same frequency coupling, and the logical mapping function makes the weight In the weak coupling range, it gradually becomes zero to avoid the sudden jump introduced by the hard threshold. The original time domain characteristics are retained, and the pressure difference sequence after compensation The high-frequency jitter amplitude is significantly reduced, and the low-frequency leakage trend remains intact, providing a purer pressure reference for leakage rate estimation; A determination module generates a qualification determination result according to the corrected leakage rate; The one-way valve air tightness detection system also includes: A non-contact vibration monitoring unit, including a laser Doppler vibrometer, is used to collect the vibration spectrum of the one-way valve diaphragm; The determination module further coordinates and corrects the leakage state determination result according to the natural frequency offset of the vibration spectrum; Synergy fixes include: When the natural frequency deviation of the vibration spectrum exceeds the preset threshold and is consistent with the leakage rate change trend, a leakage judgment is triggered; When the frequency offset is inconsistent with the leakage rate trend, the vibration spectrum judgment result is preferred.

[0026] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A method for detecting the air tightness of a one-way valve, characterized in that: include, Step S1, real-time acquisition of multimodal sensing data by a physical detection unit, wherein the physical detection unit includes a pressure differential sensor, a micro mass flow meter, an ambient temperature sensor, an airflow disturbance monitoring module, and an electromagnetic interference monitoring module; Step S2, inputting the multimodal sensing data into a pre-trained AI prediction model to generate a virtual leakage prediction value; Step S3, comparing the virtual leakage prediction value with the real-time detection value of the physical detection unit, and dynamically correcting the environmental interference component based on the deviation analysis result; Step S4, determining the air tightness state of the one-way valve according to the corrected leakage rate and a preset threshold; Among them, the AI ​​prediction model is trained based on historical sample data, and the historical sample data includes multimodal sensor data sequences under normal no-leakage state, known leakage state and multiple types of environmental interference scenarios.

2. A one-way valve air tightness detection method according to claim 1, characterized in that: When the deviation between the virtual leakage prediction value and the real-time detection value exceeds a preset threshold, the following operations are performed: a) Extracting real-time data from the ambient temperature sensor, airflow disturbance monitoring module, and electromagnetic interference monitoring module; b) Calculate the correlation between the fluctuation amplitude of each interference source and the leakage signal based on the sliding time window; c) Generate compensation coefficients for interference sources with correlation greater than a set value and correct real-time detection values.

3. A one-way valve air tightness detection method according to claim 2, characterized in that: Step S3 includes: When the trend of the virtual leakage prediction value is consistent with the real-time detection value, the leakage signal weight is enhanced; When there is a significant deviation between the two and the environmental interference monitoring module detects abnormal disturbance, an interference compensation coefficient is generated and the real-time detection value is corrected; The interference compensation coefficient is dynamically calculated by analyzing the time domain correlation of ambient temperature, airflow disturbance and electromagnetic interference.

4. A one-way valve air tightness detection method according to claim 3, characterized in that: In step S3, in the process of dynamically correcting the environmental interference component based on the deviation analysis result, five sub-processes are continuously executed in each sampling period, and the logical order is as follows: 1) Construct instantaneous deviation: After obtaining the physical measured value and the virtual predicted value, calculate the instantaneous deviation. ， in, represents the instantaneous deviation, Indicates time The measured value of the leakage rate, Indicates time The virtual leakage prediction value of is the discrete time index; 2) Normalized deviation: Use length The sliding window of is demeaned and normalized to obtain, ， in, is the normalized deviation, is the mean deviation within the window, is the standard deviation of the deviation within the window, is the sliding window length; 3) Quantifying interference correlation: For the first Interference source signal Calculate the correlation coefficient, ， in, Deviation and interference sources The correlation coefficient of is the lag index within the window, Interference source within the window The standard deviation of Interference source At the moment signal, , is the total number of interference sources; 4) Map the correlation coefficient to weight: The logistic mapping function is used to obtain continuous weights. ， in, Interference source The weight of is the mapping slope coefficient, Enable thresholds for correlation; 5) Estimate the compensation coefficient and correct the measured value: Adopt belt The regular closed-form solution estimates the compensation, ， ， in, Interference source The compensation coefficient, is the regularization coefficient, is the corrected leak rate.

5. A one-way valve air tightness detection system, based on a one-way valve air tightness detection method according to any one of claims 1 to 4, characterized in that: include: Physical detection unit, used to collect pressure difference, mass flow, ambient temperature, air flow disturbance and electromagnetic interference data; The physical detection unit also includes: High-frequency vibration sensor, used to collect mechanical vibration signals in the detection environment; A data synchronization module, connected to the physical detection unit, to achieve time alignment of multi-channel signals; An AI modeling module, which includes a pre-trained LSTM or Transformer model, is used to receive multimodal sensor data and output a virtual leakage prediction value; Signal fusion unit, used to compare virtual leakage prediction values ​​with physical detection data, perform interference removal and signal enhancement; The determination module generates a qualification determination result according to the corrected leakage rate.

6. A one-way valve air tightness detection system according to claim 5, characterized in that: The AI ​​modeling module also includes: The transfer learning submodule is used to adapt the historical training model to the detection scenario of the new specification one-way valve; The transfer learning submodule receives the geometric parameters and material properties of the new valve body and dynamically adjusts the model input layer dimensions and hidden layer weights; The dynamic adjustment of the model input layer dimension includes: According to the aperture size and material hardness of the new valve body, it is mapped into a standardized feature vector; The feature vector and the historical model input layer are dimensionally expanded or truncated, retaining the parameter weights.

7. A one-way valve air tightness detection system according to claim 5, characterized in that: The signal fusion unit includes: Timing alignment submodule, used to match the timestamps of virtual leakage prediction values ​​and physical detection values; The weight allocation submodule dynamically adjusts the fusion weight of the virtual signal and the physical signal according to the output of the environmental interference monitoring module.

8. A one-way valve air tightness detection system according to claim 7, characterized in that: The signal fusion unit further fuses the frequency domain characteristics of the vibration signal with the pressure difference data to separate the coupling interference of the vibration noise on the pressure signal; The separation of vibration noise includes: Perform wavelet packet decomposition on the vibration signal to extract the sub-band energy that overlaps with the pressure signal frequency band; When the sub-band energy exceeds a preset ratio, the corresponding frequency band component is deducted from the pressure signal; The signal fusion unit further includes: Dynamic parameter tuning module, used to adaptively adjust the sampling frequency of the differential pressure sensor and the integration time constant of the micro-mass flowmeter according to the interference intensity of the current detection environment; The adaptive adjustment includes: When the electromagnetic interference monitoring module detects that the intensity exceeds the preset limit, the sampling frequency of the differential pressure sensor is increased; When the airflow disturbance amplitude continues to increase, the integration time constant of the micro-mass flowmeter is shortened.

9. A one-way valve air tightness detection system according to claim 8, characterized in that: In the signal fusion unit, vibration pressure difference fusion and coupling interference separation are performed, including: Perform timing alignment on the data in the same window, and perform synchronization on the data with a length of time window, constructing the vibration sequence With pressure difference series : ， in, is the vibration signal window vector, is the pressure difference signal window vector, 、 Respectively indicate time Vibration and pressure difference sampling, is the lag index within the window, is the window length, the unit is the number of sampling points, is the current time index; right Perform discrete wavelet packet decomposition to obtain Subband coefficient vector : ， in, Sub-band The coefficient vector of Indicates the The wavelet packet transform operator of the sub-bands, is the energy of this frequency band, , is the total number of frequency bands at the wavelet packet level, The coefficient vector Bit; First define the effective frequency band of pressure difference , then reconstruct the mask , like ,but ,otherwise , The coupling strength is then quantified and smoothed to weights: ， ， in, is the sub-band center frequency, is the lower limit of the pressure difference passband, is the upper limit of the differential pressure passband, For the mask, is the correlation coefficient, is the smoothing weight, is the mapping slope constant; The sub-bands selected by the mask are reconstructed into interference estimates after weight scaling : ， in, Sub-band The inverse wavelet packet reconstruction operator, For the moment Vibration disturbance estimation, is the original pressure difference sampling, is the pressure difference value after vibration compensation.

10. A one-way valve air tightness detection system according to claim 5, characterized in that: Also includes: A non-contact vibration monitoring unit, including a laser Doppler vibrometer, is used to collect the vibration spectrum of the one-way valve diaphragm; The determination module further coordinates and corrects the leakage state determination result according to the natural frequency offset of the vibration spectrum; The coordinated amendments include: When the natural frequency offset of the vibration spectrum exceeds the preset threshold and is consistent with the leakage rate change trend, a leakage judgment is triggered; When the frequency offset is inconsistent with the leakage rate trend, the vibration spectrum judgment result is preferred.

Images