Automobile filter air tightness detection compensation method and system

Abstract

The invention belongs to the technical field of air tightness detection, and discloses an automobile filter air tightness detection compensation method and system, and the method comprises the steps: placing a to-be-detected automobile filter in a detection cavity in a sealing manner, and applying a dynamically changing pressure to the detection cavity according to a preset time sequence; in the process of applying the dynamically changing pressure, a dynamic pressure signal and an acoustic signal are synchronously collected; performing time domain and / or frequency domain analysis on the acoustic signal to obtain a plurality of acoustic features; according to the dynamic pressure signal, analyzing a dynamic response mode of a plurality of acoustic characteristics, and extracting a leakage acoustic characteristic strongly associated with the dynamic change of the pressure; based on the leakage acoustic characteristics, whether the to-be-detected automobile filter leaks or not and the leakage degree are judged; therefore, the real leakage sound and various interference sounds can be effectively distinguished, and the detection accuracy is improved.

Description

Technical Field

[0001] The present application relates to the technical field of air tightness detection, and in particular to a method and system for compensating air tightness detection of an automobile filter. Background Art

[0002] Automotive filters accumulate impurities during use, changing their physical properties and presenting challenges for leak detection. Traditional acoustic testing methods rely on analyzing sound signals to identify leaks. However, due to the influence of impurities, the acoustic characteristics of used filters differ from those of new filters. These impurity-induced acoustic signatures may overlap with the leak sound. Furthermore, interference from testing equipment and ambient noise makes it difficult to accurately identify leaks by directly analyzing the raw acoustic signal.

[0003] Specifically, impurities alter the geometry and surface properties of the filter media channels, affecting airflow patterns and increasing airflow turbulence. This results in a used filter producing a different sound than a new one, even under normal conditions. More seriously, the acoustic signatures caused by impurities, such as airflow noise or the sound produced by impurity vibrations, can overlap with the leakage sound spectrum, making it difficult to distinguish using spectral analysis alone. Furthermore, the combined effects of various factors, such as noise from equipment like air pumps and valves, the resonant characteristics of the detection cavity, ambient noise, and sensor noise, make the overall signal a complex mixture of various acoustic components.

[0004] Therefore, when conducting acoustic testing on the air tightness of used automotive filters, the acoustic characteristics caused by impurity accumulation change greatly, and there may be spectral overlap with potential leakage sounds. At the same time, various other interfering noises are superimposed. Direct analysis of the collected original acoustic signals makes it difficult to accurately determine whether the filter has a leak and the extent of the leak, which seriously affects the accuracy and reliability of the test.

[0005] To address these issues, it is necessary to improve the existing acoustic detection methods to effectively distinguish real leakage sounds from various interference sounds and improve detection accuracy. Summary of the Invention

[0006] The purpose of this application is to provide a method and system for detecting and compensating air tightness of an automobile filter, which can effectively distinguish between real leakage sounds and various interference sounds and improve detection accuracy.

[0007] In a first aspect, the present application provides a method for compensating for air tightness detection of an automobile filter, which is used to perform air tightness detection on a used automobile filter. The method comprises the following steps:

[0008] S1. Place the automotive filter under test in a sealed state within the test chamber and apply a dynamically changing pressure to the chamber according to a preset time sequence.

[0009] S2. While applying dynamically changing pressure, synchronously collect dynamic pressure signals and acoustic signals;

[0010] S3. performing time domain and / or frequency domain analysis on the acoustic signal to obtain multiple acoustic features;

[0011] S4. Analyze the dynamic response patterns of multiple acoustic features based on the dynamic pressure signal and extract leakage acoustic features that are strongly correlated with dynamic pressure changes;

[0012] S5. Based on the leakage acoustic characteristics, determine whether the vehicle filter under test has a leak and the extent of the leak.

[0013] Preferably, step S1 includes:

[0014] S101. Obtaining a preset time series pressure change pattern; the pressure change pattern includes the amplitude, frequency and duration of the pressure change;

[0015] S102. Based on the pressure change pattern and the performance of the gas supply system, a pressure control signal is generated; the pressure control signal compensates for the response delay and nonlinear characteristics of the gas supply system to ensure that the actual pressure change in the detection cavity is consistent with the preset time series pressure change pattern;

[0016] S103. Control the gas supply system, and according to the pressure control signal, use a closed-loop control method to apply a dynamically changing pressure to the detection cavity according to a preset time sequence.

[0017] Preferably, step S102 includes:

[0018] generating an initial control signal based on the pressure variation pattern;

[0019] Based on the pressure variation pattern and the pre-built air supply system transfer function model, an inverse model control algorithm is used to calculate the pre-compensation control signal; the air supply system transfer function model characterizes the pressure response delay and nonlinear characteristics of the air supply system;

[0020] The pre-compensated control signal is added to the initial control signal to generate a pressure control signal.

[0021] Preferably, step S2 includes:

[0022] S201. Synchronously collecting dynamic pressure signals and acoustic signals;

[0023] S202. Preprocessing the synchronously collected dynamic pressure signal and acoustic signal respectively; the preprocessing includes bandpass filtering noise reduction and signal amplification;

[0024] S203. Based on the preprocessed dynamic pressure signal, calculate the pressure change rate to obtain the pressure change rate characteristic;

[0025] S204. Using a dynamic time warping algorithm, the preprocessed acoustic signal is time-aligned based on the pressure change rate characteristic to obtain a time-aligned acoustic signal.

[0026] Preferably, step S3 includes:

[0027] S301. Pre-emphasize the acoustic signal to increase the energy of the high-frequency signal to compensate for the attenuation of high-frequency leakage sound during propagation;

[0028] S302. Performing a short-time Fourier transform on the pre-emphasized acoustic signal to obtain a spectrum of the acoustic signal;

[0029] S303. Based on the spectrogram of the acoustic signal, extract the total sound pressure level, spectral centroid, spectral bandwidth and Mel-frequency cepstral coefficients;

[0030] S304. Perform time domain analysis on the pre-emphasized acoustic signal to calculate the peak factor and kurtosis;

[0031] S305. Normalize the total sound pressure level, spectrum centroid, spectrum bandwidth, Mel-frequency cepstral coefficient, peak factor, and kurtosis to obtain the final acoustic features.

[0032] Preferably, step S4 includes:

[0033] S401. For multiple acoustic features, calculate the mutual correlation coefficient between each acoustic feature and the dynamic pressure signal to obtain multiple mutual correlation coefficients;

[0034] S402. Setting a mutual correlation coefficient threshold, comparing multiple mutual correlation coefficients with the mutual correlation coefficient threshold, screening out acoustic features with mutual correlation coefficients greater than the mutual correlation coefficient threshold, and obtaining a candidate leakage acoustic feature set;

[0035] S403. For each candidate leakage acoustic feature in the candidate leakage acoustic feature set, calculate the energy change rate of the candidate leakage acoustic feature during the dynamic pressure change process to obtain multiple energy change rates;

[0036] S404. Set an energy change rate threshold, compare multiple energy change rates with the energy change rate threshold, screen out candidate leakage acoustic features with energy change rates greater than the energy change rate threshold, and obtain leakage acoustic features.

[0037] Preferably, step S401 includes:

[0038] The time window length and sliding step size for calculating the cross-correlation coefficient are set, and the dynamic pressure signal and each acoustic feature are divided into multiple time segments according to the time window length and sliding step size;

[0039] For each acoustic feature, calculate the mutual correlation coefficient between the acoustic feature in each time segment and the dynamic pressure signal in the corresponding time segment to obtain the mutual correlation value of the acoustic feature corresponding to each time segment;

[0040] For each acoustic feature, the mutual correlation values ​​of the acoustic feature corresponding to all time segments are weighted averaged to obtain the mutual correlation coefficient of the acoustic feature; wherein the weight of the time segment is proportional to the energy of the dynamic pressure signal in its corresponding time segment.

[0041] Preferably, step S403 includes:

[0042] For each candidate leakage acoustic feature in the candidate leakage acoustic feature set, dividing the frequency spectrum of each candidate leakage acoustic feature into a plurality of frequency bands according to a preset frequency range;

[0043] For each frequency band of each candidate leakage acoustic feature, the energy value within the frequency band is calculated to obtain multiple frequency band energy values;

[0044] For each frequency band energy value of each candidate leakage acoustic feature, the rate of change of the frequency band energy value during the dynamic pressure change process is calculated to obtain multiple frequency band energy change rates;

[0045] For each candidate leakage acoustic feature, the energy change rates of multiple frequency bands of the candidate leakage acoustic feature are weighted averaged according to a preset weight coefficient to obtain the energy change rate of the candidate leakage acoustic feature; wherein the preset weight coefficient is proportional to the center frequency of the frequency band.

[0046] Preferably, step S5 includes:

[0047] S501. Calculate a weight factor for each leakage acoustic feature for a plurality of leakage acoustic features; wherein the weight factor is proportional to the correlation coefficient between the leakage acoustic feature and the dynamic pressure signal;

[0048] S502. Perform weighted fusion on multiple leakage acoustic features according to the weight factor of each leakage acoustic feature to obtain a comprehensive leakage index;

[0049] S503. Compare the comprehensive leakage index with a preset leakage judgment threshold. When the comprehensive leakage index is greater than the leakage judgment threshold, determine that the filter has leakage, and determine the leakage degree based on the size of the comprehensive leakage index.

[0050] In a second aspect, the present application provides an automobile filter air tightness detection and compensation system for performing air tightness detection on used automobile filters, the system comprising a detection cavity, an air supply system, a pressure sensor, an acoustic sensor, and an industrial computer;

[0051] The detection cavity is used to seal and place the automobile filter to be tested;

[0052] The gas supply system is used to deliver gas to the detection cavity to adjust the pressure of the detection cavity;

[0053] The pressure sensor is used to detect the pressure signal of the detection cavity and upload it to the industrial computer;

[0054] The acoustic sensor is used to detect the acoustic signal in the detection cavity and upload it to the industrial computer;

[0055] The industrial computer is used to control the air supply system to apply dynamically changing pressure to the detection cavity according to a preset time sequence after the seal of the automobile filter to be tested is placed in the detection cavity; in the process of applying the dynamically changing pressure, the dynamic pressure signal and the acoustic signal are synchronously collected through the pressure sensor and the acoustic sensor; the acoustic signal is analyzed in the time domain and / or frequency domain to obtain multiple acoustic features; based on the dynamic pressure signal, the dynamic response patterns of the multiple acoustic features are analyzed to extract leakage acoustic features that are strongly correlated with the dynamic change of pressure; based on the leakage acoustic features, it is determined whether the automobile filter to be tested has a leak and the extent of the leak.

[0056] Beneficial effects: The present application provides a method and system for compensating the air tightness detection of an automobile filter. By analyzing the dynamic response patterns of multiple acoustic features, the system extracts leakage acoustic features that are strongly correlated with dynamic pressure changes, thereby solving the problem in the prior art of difficulty in distinguishing between real leakage sounds and various interference sounds. The system can effectively distinguish between real leakage sounds and various interference sounds, thereby improving detection accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] Figure 1 This is a flow chart of the automobile filter air tightness detection and compensation method provided in an embodiment of the present application.

[0058] Figure 2 This is a schematic structural diagram of the automobile filter air tightness detection and compensation system provided in an embodiment of the present application.

[0059] Explanation of the numbers: 1. Detection cavity; 2. Air supply system; 3. Pressure sensor; 4. Acoustic sensor; 5. Industrial computer. DETAILED DESCRIPTION

[0060] The technical solutions in this application will be clearly and completely described below in conjunction with the drawings in this application. Obviously, the described embodiments are only a part of the embodiments of this application, rather than all of the embodiments. The components of the present application generally described and shown in the drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the application for protection, but merely represents selected embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without making creative work fall within the scope of protection of this application.

[0061] It should be noted that similar reference numerals and letters represent similar items in the following drawings. Therefore, once an item is defined in one drawing, it does not need to be further defined or explained in subsequent drawings. At the same time, in the description of this application, the terms "first", "second", etc. are only used to distinguish the description and should not be understood as indicating or implying relative importance.

[0062] refer to Figure 1 The present application proposes a method for compensating the air tightness detection of automobile filters, which is used to detect the air tightness of used automobile filters. The method comprises the following steps:

[0063] S1. Place the automotive filter under test in a sealed state within the test chamber and apply a dynamically changing pressure to the chamber according to a preset time sequence.

[0064] S2. While applying dynamically changing pressure, synchronously collect dynamic pressure signals and acoustic signals;

[0065] S3. performing time domain and / or frequency domain analysis on the acoustic signal to obtain multiple acoustic features;

[0066] S4. Analyze the dynamic response patterns of multiple acoustic features based on the dynamic pressure signal and extract leakage acoustic features that are strongly correlated with dynamic pressure changes;

[0067] S5. Based on the leakage acoustic characteristics, determine whether the vehicle filter under test has a leak and the extent of the leak.

[0068] Among them, dynamically changing pressure refers to the pressure in the detection cavity changing over time according to a preset rule. It can be achieved by controlling the valve or pump speed in the air supply system, such as using a proportional control valve in conjunction with a pressure sensor for closed-loop control. Its main purpose is to stimulate potential leakage points, prompt the generation of leakage sound and show characteristics of changing with pressure.

[0069] Among them, synchronous acquisition of dynamic pressure signals and acoustic signals refers to the parallel acquisition of pressure data and acoustic data in the detection cavity under the same time reference. It can be achieved by connecting pressure sensors and acoustic sensors with a multi-channel data acquisition system. Its main purpose is to establish a time correspondence between pressure changes and acoustic responses, and provide an accurate data basis for subsequent analysis of the dynamic response of acoustic signals.

[0070] Among them, analyzing the dynamic response patterns of multiple acoustic features and extracting leakage acoustic features that are strongly correlated with dynamic pressure changes refers to studying the law of acoustic feature changes with pressure and identifying acoustic components that are closely related to the pressure change trend or rate. This can be achieved by calculating the cross-correlation between the acoustic feature sequence and the pressure signal or its rate of change, analyzing the rate of change of acoustic feature energy during the pressure rise or fall stage, etc. Its main purpose is to utilize the sensitivity of real leakage sound to pressure changes to distinguish it from interference sounds that are less affected by pressure changes or have different influence patterns.

[0071] The core innovation of this application lies in applying dynamically changing pressure and simultaneously collecting pressure and acoustic signals. This analysis then analyzes the dynamic response of acoustic features to pressure changes, thereby extracting leak acoustic signatures that are strongly correlated with dynamic pressure changes. This method exploits the sensitivity of real leak sounds to pressure changes and effectively distinguishes leak sounds from the sounds caused by impurities in used filters and other interfering noise, thereby improving detection accuracy.

[0072] Specifically, the technical solution of the present application operates in the following manner. First, the automobile filter to be tested is sealed and placed in the detection cavity. The air supply system accurately applies dynamically changing pressure through closed-loop control according to a preset time sequence. During the pressure change process, the pressure sensor and the acoustic sensor synchronously collect the pressure signal and acoustic signal in the cavity. The collected acoustic signal is analyzed in the time domain and / or frequency domain to extract multiple acoustic features reflecting the sound characteristics. The synchronously collected pressure signal is used to analyze the law or pattern of these acoustic features changing with pressure. For example, the cross-correlation between the acoustic feature and the pressure signal is calculated, or the change in the energy of the acoustic feature with the rate of change of pressure is analyzed. Acoustic features that show a strong correlation with the dynamic change of pressure are identified. These features are considered to be sounds originating from the leakage point, that is, leakage acoustic features. Based on the extracted leakage acoustic features, a comprehensive judgment is made as to whether the filter has a leak and the severity of the leak. The entire process achieves the purpose of effectively separating the real leakage sound from the complex noise environment by analyzing the dynamic correlation between the acoustic signal and the pressure signal.

[0073] Through the above scheme, the present application can effectively distinguish between real leakage sounds and sounds caused by impurities in used filters and interference noise caused by detection equipment and environment, thereby improving the accuracy and reliability of air tightness detection on used automotive filters, making it possible to perform performance evaluation and fault diagnosis on used filters.

[0074] In some embodiments, step S1 includes:

[0075] S101. Obtaining a preset time series pressure change pattern; the pressure change pattern includes the amplitude, frequency and duration of the pressure change;

[0076] S102. Based on the pressure change pattern and the performance of the gas supply system, a pressure control signal is generated; the pressure control signal compensates for the response delay and nonlinear characteristics of the gas supply system to ensure that the actual pressure change in the detection cavity is consistent with the preset time series pressure change pattern;

[0077] S103. Control the gas supply system, and according to the pressure control signal, use a closed-loop control method to apply a dynamically changing pressure to the detection cavity according to a preset time sequence.

[0078] The performance of the air supply system refers to the response characteristics of the air supply system to the control signal, which can be represented by the air supply system transfer function model, which can be obtained through system identification technology.

[0079] The pressure control signal refers to an electrical signal or pneumatic signal used to drive the air supply system to produce a target pressure change, which can be in the form of a voltage signal, a current signal or an air pressure signal.

[0080] Among them, compensating for the response delay and nonlinear characteristics of the air supply system means adjusting the waveform or timing of the pressure control signal to offset the inherent response lag and nonlinear distortion of the air supply system itself, so that the actual output pressure of the air supply system is closer to the desired preset pressure change. It can be achieved by feedforward control, inverse model control, adaptive control or lookup table compensation.

[0081] Among them, the closed-loop control method refers to using the actual pressure in the detection cavity as a feedback signal, comparing it with the preset pressure, and adjusting the control strategy of the pressure control signal in real time according to the comparison result. It can be implemented by proportional-integral-differential (PID) control, fuzzy control or model predictive control.

[0082] This solution aims to address the discrepancy between the actual pressure in the test chamber and the preset pressure due to the inherent characteristics of the gas supply system (response delay and nonlinearity), thereby improving the accuracy of airtightness testing. Its core is to compensate for deficiencies in the gas supply system through a pressure control signal, ensuring accurate pressure application.

[0083] Specifically, the system first acquires a pressure variation pattern over a preset time series, defining the desired pressure variation and laying the foundation for subsequent compensation control. The pressure variation pattern, encompassing amplitude, frequency, and duration, provides a complete description of the dynamic pressure variation, ensuring comprehensive control. Next, a pressure control signal is generated based on the pressure variation pattern and the performance of the gas supply system. This crucial step accounts for the response delay and nonlinear characteristics of the gas supply system, pre-compensating it through the control signal. This means the control signal does not simply replicate the preset pressure variation, but rather adjusts based on the actual performance of the gas supply system, thereby overcoming its inherent limitations. Generating a pressure control signal based on the pressure variation pattern and gas supply system performance enables precise control of the gas supply system, ensuring that its output pressure variation closely matches the preset pressure variation pattern. Finally, the gas supply system is controlled, and based on the pressure control signal, a closed-loop control method is used to dynamically apply pressure to the detection chamber according to a preset time series. The introduction of closed-loop control further improves the accuracy and stability of pressure control. By monitoring the actual pressure within the detection chamber in real time and comparing it with the preset pressure, the control signal is continuously adjusted to eliminate disturbances and errors, ensuring a high degree of consistency between the actual and preset pressures. According to the pressure control signal, a closed-loop control method is adopted to adjust the gas supply system in real time, thereby ensuring that the pressure in the detection cavity changes according to a preset time sequence.

[0084] Through the synergistic effect of the above steps, the solution of this application ensures that when dynamically changing pressure is applied to the automotive filter under test, the actual pressure within the test chamber accurately follows a preset time series variation pattern. This precise pressure control provides accurate and reliable input conditions for the subsequent simultaneous acquisition of dynamic pressure and acoustic signals, making it possible to analyze the acoustic signals and extract leak acoustic signatures that are strongly correlated with dynamic pressure changes, thereby improving the accuracy and reliability of the entire airtightness testing method.

[0085] Preferably, step S102 may include:

[0086] generating an initial control signal based on the pressure variation pattern;

[0087] Based on the pressure variation pattern and the pre-built air supply system transfer function model, an inverse model control algorithm is used to calculate the pre-compensation control signal; the air supply system transfer function model characterizes the pressure response delay and nonlinear characteristics of the air supply system;

[0088] The pre-compensated control signal is added to the initial control signal to generate a pressure control signal.

[0089] Among them, a pre-built gas supply system transfer function model is used to describe the dynamic response characteristics of the gas supply system to the control signal input, including the delay in pressure establishment, response speed, and possible nonlinear behavior. The model can be constructed using system identification technology. For example, by inputting a series of known test control signals into the gas supply system and measuring the actual pressure response data in the detection cavity, the model structure and parameters are determined using parameter estimation methods such as least squares or recursive least squares. The model can be a linear dynamic model, such as a transfer function or state space model, or a nonlinear model that can capture nonlinear characteristics.

[0090] The inverse model control algorithm calculates a pre-compensated control signal based on a constructed transfer function model of the air supply system. The basic idea is to utilize the inverse dynamic characteristics of the system to generate a control signal that, after passing through the air supply system, can offset or compensate for the system's own dynamic characteristics (such as delay and nonlinearity), allowing the system's output (actual pressure) to more accurately track the desired input (pressure change pattern). A variety of techniques can be used to implement the inverse model control algorithm. For example, if the system model is linear and reversible, the inverse of the model can be directly calculated. For nonlinear or irreversible systems, model-based predictive control, internal model control, or other advanced control strategies can be used to approximate the functionality of inverse model control.

[0091] Based on the above technical features, the specific operating principle of the pressure control signal generated by this application is as follows: This application is able to precisely control the pressure changes within the detection chamber, ensuring that they are highly consistent with a preset time-series pressure change pattern, due to the use of a model-based pre-compensation control strategy. First, based on the preset pressure change pattern, an initial control signal is generated. This signal represents the command that, under ideal conditions, drives the gas supply system to produce the desired pressure change. However, the gas supply system has inherent response delays and nonlinear characteristics, and directly using the initial control signal cannot ensure that the actual pressure accurately tracks the preset pattern. For this reason, this application introduces a pre-compensation step. By using a pre-constructed transfer function model of the gas supply system, the model accurately captures the dynamic response characteristics of the gas supply system, including its delays and nonlinearities. Based on this model, an inverse model control algorithm is employed to calculate a pre-compensated control signal. This pre-compensated signal is designed to offset or compensate for the dynamic characteristics of the gas supply system. Specifically, the inverse model control algorithm predicts the gas supply system's response to the initial control signal and calculates a correction signal so that, when this correction signal is superimposed on the initial signal, the gas supply system's output more closely approximates the desired pressure change. Finally, the calculated pre-compensated control signal is added to the initial control signal to form the final pressure control signal, which is then sent to the air supply system to drive its operation.

[0092] In this way, after receiving the pre-compensated control signal, the actual pressure change output by the gas supply system can significantly reduce the deviation caused by the system's own dynamic characteristics, thereby more accurately and in real time following the preset time series pressure change pattern. This model-based pre-compensation method, combined with closed-loop control, can further improve the accuracy and stability of pressure control, ensuring that the pressure environment during the air tightness testing process meets the requirements.

[0093] In some embodiments, step S2 includes:

[0094] S201. Synchronously collecting dynamic pressure signals and acoustic signals;

[0095] S202. Preprocessing the synchronously collected dynamic pressure signal and acoustic signal respectively; the preprocessing includes bandpass filtering noise reduction and signal amplification;

[0096] S203. Based on the preprocessed dynamic pressure signal, calculate the pressure change rate to obtain the pressure change rate characteristic;

[0097] S204. Using a dynamic time warping algorithm, the preprocessed acoustic signal is time-aligned based on the pressure change rate characteristic to obtain a time-aligned acoustic signal.

[0098] In S201, during the application of dynamically changing pressure, pressure signals and acoustic signals are simultaneously acquired. This can be achieved using a pressure sensor and an acoustic sensor. The pressure sensor measures the pressure within the detection cavity, while the acoustic sensor measures the acoustic signal within the detection cavity. The data acquisition systems of the two sensors are configured to operate synchronously to ensure that the corresponding pressure values ​​and acoustic signal data are acquired at the same time.

[0099] Among them, in S202, preprocessing can be specifically implemented using a digital signal processor or computer software. Bandpass filtering noise reduction refers to retaining the components within a specific frequency range in the signal through a filter and removing noise outside the range. For example, a Butterworth filter or a Chebyshev filter can be used. Signal amplification refers to increasing the amplitude of the signal to facilitate subsequent processing and analysis. For example, an operational amplifier circuit or digital gain control can be used. Preprocessing the two signals separately is to improve the quality of their respective signals and provide a more reliable data basis for subsequent feature extraction and analysis.

[0100] In S203, the pressure change rate is calculated, which can be implemented using a numerical differentiation algorithm. For example, a numerical differentiation algorithm can be used to calculate the ratio of the difference between the pressure values ​​at adjacent time points to the time interval. Alternatively, a more complex differentiation or fitting method can be used to calculate the instantaneous pressure change rate. The resulting pressure change rate feature is a time series that reflects the dynamic process of pressure change over time.

[0101] Among them, in S204, the time axis of the acoustic signal is adjusted to align it with the time axis of the pressure change rate feature. Specifically, this can be achieved by using the dynamic time warping (DTW) algorithm. The DTW algorithm minimizes the distance between two time series by finding the best matching path between them, even if they are nonlinearly stretched or compressed on the time axis. Here, the pressure change rate feature is used as a reference sequence, and the acoustic signal is used as the sequence to be aligned. The DTW algorithm calculates the optimal time correspondence of the acoustic signal relative to the pressure change rate feature, and resamples or interpolates the acoustic signal according to this correspondence to obtain the acoustic signal after time axis alignment.

[0102] Specifically, during the pressure change process, the dynamic pressure signal and acoustic signal within the cavity are synchronously collected. The collected raw signals may contain interference such as equipment noise, environmental noise, and sensor noise. Due to factors such as system response, there may be time deviation between the pressure and acoustic signals. To improve signal quality and correct for time deviation, the collected pressure and acoustic signals are preprocessed separately. This preprocessing includes bandpass filtering to reduce noise, remove unwanted frequency components and random noise, and signal amplification to enhance weak leakage acoustic signals. The preprocessed signals have a higher signal-to-noise ratio and clearer features. Next, the pressure change rate is calculated based on the preprocessed dynamic pressure signal. The pressure change rate better reflects the transient characteristics of pressure changes, and leakage acoustics are often associated with rapid pressure changes. A dynamic time warping algorithm is then used to align the preprocessed acoustic signals with their time axis, using the calculated pressure change rate feature as a time reference. This alignment compensates for any nonlinear time delay between the pressure and acoustic signals, ensuring that, in subsequent analysis, specific changes in the acoustic signal are accurately associated with the pressure change phase that caused them. Through these steps, a high-quality acoustic signal that is accurately aligned with pressure changes is obtained, which facilitates the subsequent time domain / frequency domain analysis to extract acoustic features and analyze the dynamic response patterns between these features and the dynamic pressure signal. Ultimately, a solid foundation is laid for determining whether there is a leak in the filter and the extent of the leak based on the extracted leakage acoustic features, overcoming the adverse effects of original signal interference and timing misalignment on detection accuracy.

[0103] Through the above technical solution, this application improves the quality of the collected dynamic pressure and acoustic signals and corrects the time deviation between them. This provides a reliable data foundation for subsequent dynamic response pattern analysis to accurately identify the acoustic characteristics of the leak, and solves the problem of noise interference and timing misalignment affecting the analysis accuracy of the original signal.

[0104] In some embodiments, step S3 includes:

[0105] S301. Pre-emphasize the acoustic signal to increase the energy of the high-frequency signal to compensate for the attenuation of high-frequency leakage sound during propagation;

[0106] S302. Performing a short-time Fourier transform on the pre-emphasized acoustic signal to obtain a spectrum of the acoustic signal;

[0107] S303. Based on the spectrogram of the acoustic signal, extract the total sound pressure level, spectral centroid, spectral bandwidth and Mel-frequency cepstral coefficients;

[0108] S304. Perform time domain analysis on the pre-emphasized acoustic signal to calculate the peak factor and kurtosis;

[0109] S305. Normalize the total sound pressure level, spectrum centroid, spectrum bandwidth, Mel-frequency cepstral coefficient, peak factor, and kurtosis to obtain the final acoustic features.

[0110] Among them, pre-emphasis processing refers to a digital signal processing technology that uses a filter to increase the amplitude of the high-frequency part of the signal to compensate for the attenuation of the high-frequency signal during transmission or recording.

[0111] The short-time Fourier transform (STFT) is a time-frequency analysis method that divides a long signal into shorter time segments and performs a Fourier transform on each segment to analyze the signal's spectral characteristics over time. A spectrogram is a visual representation of the STFT results, with time on the horizontal axis and frequency on the vertical axis. Color or grayscale represents the energy or amplitude at the corresponding time and frequency points.

[0112] Among them, the total sound pressure level is a measure of the overall energy or intensity of the acoustic signal within a certain period of time. The spectral centroid describes the indicator of the center position of the spectral energy distribution. The spectral bandwidth describes the indicator of the range of spectral energy distribution. Mel-frequency cepstral coefficient (MFCC) is an acoustic feature based on the auditory characteristics of the human ear, which is obtained by Mel filtering, taking logarithms, discrete cosine transform and other steps on the spectrum. The peak factor is the ratio of the peak amplitude of the signal to the effective value, which reflects the sharpness of the signal waveform or the amount of impact components it contains. The kurtosis describes the degree of deviation of the signal probability density function from the normal distribution, reflecting the sharpness of the signal waveform or the amount of outliers it contains. The extraction or calculation methods of the above-mentioned acoustic features can adopt existing technologies and are not limited here.

[0113] Among them, the eigenvalues ​​of different dimensions or ranges are scaled to a uniform range or distribution through normalization processing to eliminate the dimension effect and facilitate subsequent processing.

[0114] The acoustic feature extraction method described in this application aims to overcome the challenges of high-frequency attenuation and noise interference in analyzing raw acoustic signals. First, pre-emphasis is used to boost the acoustic signal's energy in the high-frequency range. This is because leakage sound attenuates significantly in this high-frequency range, and pre-emphasis effectively compensates for this attenuation, preserving the characteristics of high-frequency leakage sound. Subsequently, a short-time Fourier transform is performed on the pre-emphasized signal to generate a spectrogram, converting the signal from the time domain to the time-frequency domain, revealing how the signal's frequency components change over time. Based on this spectrogram, frequency domain features such as total sound pressure level, spectral centroid, spectral bandwidth, and Mel-frequency cepstral coefficients are extracted. These features characterize the signal's spectral distribution and energy concentration from different dimensions, helping to distinguish leakage sound with specific spectral characteristics from background noise. Furthermore, time domain analysis is performed on the pre-emphasized signal to calculate the crest factor and kurtosis. These time domain features are crucial for capturing the potential impact or transient characteristics of leakage sound. Finally, all extracted frequency and time domain features are normalized to eliminate dimensional differences between the features and ensure comparability in subsequent analysis. By organically combining the above steps, this application provides a set of comprehensive acoustic features that include both spectral information and time domain characteristics. Pre-emphasis processing enhances sensitivity to high-frequency leakage sounds, and normalization improves the robustness of the features. These optimized acoustic features are used as input for subsequent analysis of the dynamic response pattern based on the dynamic pressure signal, extracting leakage acoustic features that are strongly correlated with pressure changes, and ultimately determining whether a leak exists and the extent of the leak. High-quality acoustic feature extraction provides a solid foundation for subsequent pattern analysis and leakage judgment, significantly improving the accuracy and reliability of the overall detection method.

[0115] In a specific embodiment, when pre-emphasizing the collected acoustic signal, a digital first-order high-pass filter can be used, whose transfer function is designed to increase the gain of the high-frequency portion. Next, a short-time Fourier transform (SFT) is performed on the pre-emphasized signal. A Hanning window can be used as the window function, with an appropriate window length and overlap ratio set to obtain a time-frequency representation of the signal, i.e., a spectrogram. Based on this spectrogram, the total sound pressure level can be calculated, for example, by integrating the spectral energy and converting it to decibels; the spectral centroid and spectral bandwidth can be calculated, for example, by weighted averaging the frequency and calculating the energy distribution range; and the Mel-frequency cepstral coefficients can be calculated, for example, by mapping the spectrum to a Mel-scale filter bank, taking the logarithm, and then performing a discrete cosine transform. Furthermore, the crest factor (the ratio of the maximum absolute value of the signal to the effective value) of the pre-emphasized time-domain signal can be calculated; and the kurtosis can be calculated, for example, by using the ratio of the fourth-order central moment to the square of the variance. Finally, the calculated eigenvalues ​​of total sound pressure level, spectral centroid, spectral bandwidth, Mel-frequency cepstral coefficient, peak factor and kurtosis are normalized, for example, using the Z-score normalization method, so that the mean of all features is zero and the variance is one, thereby obtaining the final acoustic eigenvector for subsequent analysis.

[0116] By pre-emphasizing the acoustic signal, the energy of the high-frequency signal is increased, effectively compensating for the attenuation of high-frequency leakage sound during propagation, and enhancing the ability to capture weak high-frequency leakage sound. Combined with the spectrum obtained by short-time Fourier transform, frequency domain features such as total sound pressure level, spectrum centroid, spectrum bandwidth, and Mel-frequency cepstral coefficients are extracted, as well as time domain features such as peak factor and kurtosis calculated by time domain analysis. These features comprehensively describe the time domain and frequency domain characteristics of the acoustic signal, which helps to distinguish real leakage sound from background noise and sounds caused by impurities. These features are normalized to eliminate dimensional differences and improve the robustness of the features. Therefore, the present application provides a set of more accurate and discriminative acoustic features, which significantly improves the accuracy of air tightness detection of used automotive filters.

[0117] It should be noted that in step S3, only time domain analysis or only frequency domain analysis may be performed; when only time domain analysis is performed, the acoustic features may include peak factor and kurtosis; when only frequency domain analysis is performed, the acoustic features may include total sound pressure level, spectrum centroid, spectrum bandwidth and Mel-frequency cepstral coefficients.

[0118] In some embodiments, step S4 includes:

[0119] S401. For multiple acoustic features, calculate the mutual correlation coefficient between each acoustic feature and the dynamic pressure signal to obtain multiple mutual correlation coefficients;

[0120] S402. Setting a mutual correlation coefficient threshold, comparing multiple mutual correlation coefficients with the mutual correlation coefficient threshold, screening out acoustic features with mutual correlation coefficients greater than the mutual correlation coefficient threshold, and obtaining a candidate leakage acoustic feature set;

[0121] S403. For each candidate leakage acoustic feature in the candidate leakage acoustic feature set, calculate the energy change rate of the candidate leakage acoustic feature during the dynamic pressure change process to obtain multiple energy change rates;

[0122] S404. Set an energy change rate threshold, compare multiple energy change rates with the energy change rate threshold, screen out candidate leakage acoustic features with energy change rates greater than the energy change rate threshold, and obtain leakage acoustic features.

[0123] The cross-correlation coefficient is a measure of the linear correlation between two signals, reflecting their similarity at different time delays. In this application, the cross-correlation coefficient between acoustic features and dynamic pressure signals is calculated to quantify the synchronization or correlation strength of each acoustic feature with pressure changes. Standard signal processing algorithms can be used to calculate the cross-correlation coefficient.

[0124] The energy change rate refers to the degree or trend of the energy of an acoustic feature changing with dynamic pressure. In this application, calculating the energy change rate of an acoustic feature during a dynamic pressure change is intended to quantify the sensitivity of the acoustic feature energy to pressure changes. The energy change rate can be calculated based on how the energy value of the acoustic feature changes at different pressure levels or stages of pressure change.

[0125] This solution aims to more accurately extract acoustic features associated with leaks, improving leak detection accuracy through a two-step screening mechanism. First, for multiple acoustic features, the cross-correlation coefficient is calculated between each acoustic feature and the dynamic pressure signal. The cross-correlation coefficient reflects the degree of similarity between two signals and quantifies the degree of correlation between each acoustic feature and pressure fluctuations. This is because leak sounds typically vary with pressure, so acoustic features that strongly correlate with pressure fluctuations are more likely to be leaks. Next, a cross-correlation coefficient threshold is set and acoustic features with cross-correlation coefficients greater than the threshold are screened to obtain a set of candidate leak acoustic features. Setting the threshold allows for the initial elimination of acoustic features with weaker correlations with pressure fluctuations, reducing the computational effort required for subsequent analysis and improving the efficiency of leak acoustic feature extraction. Next, for each candidate leak acoustic feature in the set, the energy change rate during dynamic pressure fluctuations is calculated. The energy change rate reflects the magnitude of the change in acoustic feature energy with pressure. Leak sound energy typically increases with increasing pressure, so acoustic features with higher energy change rates are more likely to be leaks. Finally, an energy change rate threshold is set, and candidate leakage acoustic features with energy change rates greater than the energy change rate threshold are screened to obtain the final leakage acoustic features. By setting the energy change rate threshold, acoustic features with less obvious energy changes can be further eliminated, leading to more accurate extraction of leakage acoustic features.

[0126] This approach, through dual screening using the cross-correlation coefficient and energy change rate, effectively extracts leak acoustic features that are strongly correlated with dynamic pressure changes, providing a more reliable basis for subsequent leak detection. This two-stage screening method combines the time-domain correlation between acoustic features and pressure signals with the dynamic nature of acoustic feature energy changes with pressure, effectively distinguishing true leak acoustic features from complex background noise and non-leakage sounds. The leak acoustic features obtained in this way have an improved signal-to-noise ratio and are more closely correlated with the leak phenomenon, significantly improving the accuracy and reliability of leak detection based on these features.

[0127] Preferably, step S401 may include:

[0128] The time window length and sliding step size for calculating the cross-correlation coefficient are set, and the dynamic pressure signal and each acoustic feature are divided into multiple time segments according to the time window length and sliding step size;

[0129] For each acoustic feature, calculate the mutual correlation coefficient between the acoustic feature in each time segment and the dynamic pressure signal in the corresponding time segment to obtain the mutual correlation value of the acoustic feature corresponding to each time segment;

[0130] For each acoustic feature, the mutual correlation values ​​of the acoustic feature corresponding to all time segments are weighted averaged to obtain the mutual correlation coefficient of the acoustic feature; wherein the weight of the time segment is proportional to the energy of the dynamic pressure signal in its corresponding time segment.

[0131] The time window length refers to the duration of each time segment used to calculate the cross-correlation coefficient. It can be implemented with a fixed time length or a length that is adaptively adjusted based on signal characteristics. The sliding step size refers to the time interval between the starting points of adjacent time segments. It can be implemented with a fixed step size that is less than or equal to the time window length, or a step size that is dynamically adjusted based on signal characteristics. A time segment refers to a subsegment obtained by dividing the original signal according to the set time window length and sliding step size. It can be implemented using a signal interception method based on time index or sampling point index.

[0132] The cross-correlation coefficient refers to a statistic that measures the similarity between two signals at different time delays, and can be implemented using the Pearson correlation coefficient, the Spearman rank correlation coefficient, or other indicators that measure linear or nonlinear correlation of signals.

[0133] Weighted averaging refers to assigning a different weight to each value when calculating the average, then multiplying each value by its weight, summing the results, and finally dividing by the total weight. This can be achieved using a preset weighting function or dynamically calculating weights based on signal characteristics. Time segment weights refer to the relative importance assigned to the interrelationship values ​​of each time segment in the weighted averaging calculation. These can be determined proportionally to the energy of the dynamic pressure signal within the corresponding time segment, for example, by directly using the energy value or a function of the energy value as the weight.

[0134] The energy of the dynamic pressure signal refers to the intensity measurement of the dynamic pressure signal in a specific time segment, which can be achieved by calculating the sum of squares of the signal amplitudes, the root mean square value, or the signal power.

[0135] The improved method proposed in this application eliminates the need to simply calculate the cross-correlation coefficient for the entire time period when screening for leak acoustic signatures that are strongly correlated with dynamic pressure variations. Instead, a time window length and a sliding step size are first set. Based on these parameters, the dynamic pressure signal and each acoustic signature to be analyzed are segmented into a series of time segments. This segmentation allows for analysis of the local characteristics of the signal. Next, for each acoustic signature, the cross-correlation coefficient is calculated for each time segment between the signature and the dynamic pressure signal for the corresponding time segment. These calculated cross-correlation values ​​reflect the degree of correlation between the acoustic signature and the dynamic pressure signal within each local time segment, i.e., the instantaneous correlation. Subsequently, to obtain a comprehensive correlation assessment, these local cross-correlation values ​​are weighted averaged. The key to this approach lies in the weighting: the weight assigned to each time segment is proportional to the energy of the dynamic pressure signal within the corresponding time segment. This means that time segments with higher dynamic pressure signal energy have a greater weighting in the final weighted average. Given that leaks are often closely associated with pressure variations, especially when pressure variations are drastic and energy is high, leak acoustic signatures are more likely to exhibit a strong correlation with pressure. By assigning greater weights to these high-energy time periods, this scheme can more effectively highlight the correlation between acoustic features and pressure at critical moments of pressure change, thereby obtaining a cross-correlation coefficient that can more accurately reflect the degree of real leakage correlation.

[0136] This segmented calculation and pressure-energy-based weighted averaging method overcomes the problem of directly calculating the global cross-correlation coefficient, which can dilute the overall correlation due to signal uncorrelation in certain time periods. This allows the accurate identification of leak acoustic signatures that occur synchronously with dynamic pressure changes even in complex noise environments. This improved cross-correlation coefficient calculation method, combined with the overall airtightness testing method, can more accurately screen out true leak acoustic signatures, thereby improving the accuracy of determining the presence and extent of filter leaks.

[0137] Preferably, step S403 may include:

[0138] For each candidate leakage acoustic feature in the candidate leakage acoustic feature set, dividing the frequency spectrum of each candidate leakage acoustic feature into a plurality of frequency bands according to a preset frequency range;

[0139] For each frequency band of each candidate leakage acoustic feature, the energy value within the frequency band is calculated to obtain multiple frequency band energy values;

[0140] For each frequency band energy value of each candidate leakage acoustic feature, the rate of change of the frequency band energy value during the dynamic pressure change process is calculated to obtain multiple frequency band energy change rates;

[0141] For each candidate leakage acoustic feature, the energy change rates of multiple frequency bands of the candidate leakage acoustic feature are weighted averaged according to a preset weight coefficient to obtain the energy change rate of the candidate leakage acoustic feature; wherein the preset weight coefficient is proportional to the center frequency of the frequency band.

[0142] The term "preset frequency range" refers to one or more frequency intervals determined before spectrum analysis, based on the characteristics of the acoustic signal to be detected and the frequency distribution of potential leaks. For example, this can be the operating frequency range of the acoustic sensor or the frequency range where known leaks may occur. Frequency banding involves dividing the predefined frequency range into several consecutive sub-frequency intervals according to specific rules.

[0143] Calculating the energy value within a frequency band refers to integrating or summing the spectral components within each divided frequency interval to quantify the energy of the frequency band. This can be achieved by integrating the power spectral density of all frequency points in the frequency band, or by summing the squares of the amplitudes of all frequency points in the frequency band.

[0144] Calculating the rate of change of frequency band energy values ​​during dynamic pressure changes refers to analyzing the trend and amplitude of the energy value of each frequency band as it changes with time or pressure. This can be achieved by calculating the derivative of the energy value with respect to time, calculating the difference or ratio of the energy value under different pressure states, or fitting a curve of energy value changes with pressure and calculating its slope.

[0145] Among them, weighted averaging of the energy change rates of multiple frequency bands of the candidate leakage acoustic features refers to multiplying the energy change rates calculated in different frequency bands by the corresponding weight coefficients, summing them up, and then dividing them by the total weight (i.e., the sum of the weight coefficients).

[0146] Among them, the preset weight coefficient is proportional to the center frequency of the frequency band, which means that a weight coefficient is assigned to each frequency band, and the weight coefficient increases linearly or nonlinearly with the increase of the center frequency of the frequency band. Specifically, the weight coefficient can be set using the following formula: weight coefficient = k*center frequency + b, where k and b are constants, and k is greater than zero.

[0147] This solution refines the calculation of the energy change rate of candidate leak acoustic features in the method step. When analyzing the dynamic response pattern of acoustic features based on dynamic pressure signals and extracting leak acoustic features strongly correlated with dynamic pressure changes, it is necessary to calculate the energy change rate of the candidate leak acoustic features during dynamic pressure changes. By dividing the spectrum of the candidate leak acoustic features into multiple frequency bands, the energy change rate of each frequency band can be calculated separately. This frequency band division allows analysis of acoustic feature energy changes to move beyond the overall analysis and delve deeper into specific frequency regions. Calculating the energy value of each frequency band quantifies the energy distribution across different frequency ranges. Further calculating the energy change rate of each frequency band during dynamic pressure changes reveals the degree of response of different frequency components to pressure changes. Finally, a weighted average of these energy change rates is used to generate a comprehensive energy change rate index. The weighting coefficient is proportional to the center frequency of the frequency band, giving the energy change rate of high-frequency bands a greater weight in the final result.

[0148] This processing method is consistent with the physical characteristics of leakage sound. Leakage sound usually contains rich high-frequency components, and high-frequency signals attenuate greatly during propagation and are more sensitive to pressure changes. Therefore, energy changes in high-frequency bands can better reflect the actual leakage situation. Through this method based on frequency band division and weighted averaging, leakage acoustic features that are strongly associated with dynamic pressure changes can be more effectively captured, thereby improving the accuracy of subsequent leakage judgments. Through the above technical solution, the present application can more accurately calculate the energy change rate of candidate leakage acoustic features during dynamic pressure changes, effectively distinguish the response differences of different frequency bands to pressure changes, reduce the impact of noise in specific frequency bands, and thus improve the accuracy of leakage acoustic feature extraction.

[0149] In some embodiments, step S5 includes:

[0150] S501. Calculate a weight factor for each leakage acoustic feature for a plurality of leakage acoustic features; wherein the weight factor is proportional to the correlation coefficient between the leakage acoustic feature and the dynamic pressure signal;

[0151] S502. Perform weighted fusion on multiple leakage acoustic features according to the weight factor of each leakage acoustic feature to obtain a comprehensive leakage index;

[0152] S503. Compare the comprehensive leakage index with a preset leakage judgment threshold. When the comprehensive leakage index is greater than the leakage judgment threshold, determine that the filter has leakage, and determine the leakage degree based on the size of the comprehensive leakage index.

[0153] Step S501 calculates a weight factor for each leak acoustic feature. The weight factor reflects the importance or reliability of each leak acoustic feature in leak detection. The weight factor is proportional to the cross-correlation coefficient between the leak acoustic feature and the dynamic pressure signal. The cross-correlation coefficient measures the similarity or correlation between the two signals. Since leak sound is caused by pressure changes, the stronger the correlation between the acoustic feature and the dynamic pressure change, the more likely it is to be a true leak signal and the higher its reliability. Therefore, determining the weight based on the cross-correlation coefficient ensures that features with strong correlation with pressure changes play a greater role in subsequent detection. For example, the weight factor can be calculated using the following formula: wi = |ri| / (|r1| + |r2| + ... + |rn|), i∈[1,n], where wi is the weight factor for the i-th leak acoustic feature, ri is the cross-correlation coefficient between the i-th leak acoustic feature and the dynamic pressure signal, n is the number of leak acoustic features, and r1 through rn represent the cross-correlation coefficients between the 1st to nth leak acoustic features and the dynamic pressure signal.

[0154] In step S502, weighted fusion combines the information of multiple features according to their weights. This approach effectively combines the information of multiple leakage acoustic features with different reliability to form a more comprehensive and robust indicator. For example, if the leakage acoustic features are f1, f2, ..., fn, and the corresponding weight factors are w1, w2, ..., wn, respectively, then the comprehensive leakage index I can be calculated as I = w1*f1+w2*f2+...+wn*fn. This weighted summation method can highlight the role of important features and suppress the influence of secondary features or noise.

[0155] In step S503, the presence of a filter leak is determined by comparing the comprehensive leakage index with a preset threshold. When the comprehensive leakage index exceeds the preset leakage judgment threshold, the filter is deemed leaky. Furthermore, the severity of the leak can be further determined based on the magnitude of the comprehensive leakage index. For example, multiple thresholds can be set to divide the magnitude of the comprehensive leakage index into different intervals, each corresponding to a leakage severity level, such as slight leakage, moderate leakage, or severe leakage. This step makes the judgment based on the comprehensive information obtained through the weighted fusion process, making the judgment more objective and quantitative.

[0156] Specifically, this solution aims to address how to effectively utilize multiple identified leak acoustic features to improve the accuracy and reliability of leak detection. The core concept is to assign different weights to different leak acoustic features and then perform weighted fusion to form a comprehensive detection index. First, the cross-correlation coefficient between each leak acoustic feature and the dynamic pressure signal is calculated for each of the multiple leak acoustic features. A weighting factor is then determined based on the cross-correlation coefficient, with the larger the cross-correlation coefficient, the greater the weighting factor. Since leaks are caused by pressure fluctuations, acoustic features with stronger correlations to pressure fluctuations are more likely to be true leak signals and thus have higher reliability. Therefore, determining weights based on the cross-correlation coefficient ensures that features with strong correlations to pressure fluctuations play a greater role in subsequent detection, thereby enhancing detection accuracy. Next, based on the calculated weighting factors, multiple leak acoustic features are weightedly fused to produce a comprehensive leak index. This weighted fusion approach effectively combines information from multiple leak acoustic features with varying degrees of reliability to form a more comprehensive and robust indicator. Compared to using only a single feature or simple averaging, weighted fusion can highlight the role of important features and suppress the influence of secondary features or noise, thereby improving the stability of the final judgment. Finally, the resulting comprehensive leakage index is compared with the preset leakage judgment threshold. When the comprehensive leakage index exceeds the set threshold, the filter is determined to have a leak. The magnitude of the comprehensive leakage index can also be used to further determine the extent of the leak. This process, based on the comprehensive information obtained by the weighted fusion, makes the judgment more objective and quantitative, and can distinguish between different levels of leakage.

[0157] By calculating the correlation coefficient between each leak acoustic feature and the dynamic pressure signal as a weighting factor, the strength of the association between different features and the leak source can be quantified, thereby giving higher weight to features with strong correlations during weighted fusion. This weighted fusion method, based on correlation strength, enables the comprehensive leakage index to more accurately reflect the actual leakage situation, effectively reducing the negative impact of features with weak correlation or high interference on the judgment result. Comparing this more representative comprehensive index with the threshold can improve the accuracy of filter leak judgment. At the same time, the magnitude of the comprehensive index can more finely distinguish different leakage levels, providing more reliable evaluation results.

[0158] refer to Figure 2 , the present application provides an automobile filter air tightness detection and compensation system for performing air tightness detection on used automobile filters, the system comprising a detection cavity 1, an air supply system 2, a pressure sensor 3, an acoustic sensor 4 and an industrial computer 5;

[0159] The detection cavity 1 is used to seal and place the automobile filter to be tested;

[0160] The gas supply system 2 is used to deliver gas to the detection cavity 1 to adjust the pressure of the detection cavity 1;

[0161] The pressure sensor 3 is used to detect the pressure signal of the detection cavity 1 and upload it to the industrial computer 5;

[0162] The acoustic sensor 4 is used to detect the acoustic signal in the detection cavity 1 and upload it to the industrial computer 5;

[0163] The industrial computer 5 is used to control the air supply system 2 to apply dynamically changing pressure to the detection cavity 1 according to a preset time sequence after the seal of the automobile filter to be tested is placed in the detection cavity 1; in the process of applying the dynamically changing pressure, the dynamic pressure signal and the acoustic signal are synchronously collected through the pressure sensor 3 and the acoustic sensor 4; the acoustic signal is analyzed in the time domain and / or frequency domain to obtain multiple acoustic features; according to the dynamic pressure signal, the dynamic response patterns of the multiple acoustic features are analyzed to extract the leakage acoustic features that are strongly correlated with the dynamic change of pressure; based on the leakage acoustic features, it is judged whether there is a leak in the automobile filter to be tested and the degree of leakage (the specific process can be referred to steps S1 to S5 above).

[0164] The industrial computer 5 refers to a computer system used for industrial automation control and data processing, which can be implemented by an industrial personal computer, a programmable logic controller combined with a host computer, or an embedded system.

[0165] The detection cavity 1 can be made of metal material, and a rubber sealing ring and a quick clamp are used to seal the filter to be tested.

[0166] The gas supply system 2 may include a compressed air source, a proportional pressure regulating valve and an electromagnetic control valve. The industrial computer 5 controls the opening of the proportional pressure regulating valve through an analog output signal, thereby accurately adjusting the gas flow and pressure entering the detection cavity 1.

[0167] The pressure sensor 3 may be a piezoresistive pressure sensor, which is installed on the side wall of the detection cavity 1 and transmits the pressure signal to the industrial computer 5 through the analog input port.

[0168] The acoustic sensor 4 may be a high-sensitivity microphone installed inside the detection cavity 1 near the filter, and the acoustic signal is input to the industrial computer 5 through an audio acquisition card.

[0169] Through the above system, the following technical effects can be achieved.

[0170] The system provided in this application integrates a detection chamber 1, an air supply system 2, a pressure sensor 3, an acoustic sensor 4, and an industrial computer 5, providing a complete hardware and software platform capable of accurately implementing a method for airtightness testing of used automotive filters. The system precisely controls dynamic pressure changes within the detection chamber, simultaneously acquiring pressure and acoustic signals with high precision, and leveraging the powerful data processing capabilities of the industrial computer 5 to effectively distinguish true leak sounds from various interference sounds, thereby improving the accuracy and reliability of the test. This systematic design ensures the repeatability and automation of the test process, making it suitable for rapid and accurate airtightness assessment of used filters in industrial production lines or after-sales maintenance.

[0171] The above description is merely an embodiment of the present application and is not intended to limit the scope of protection of the present application. For those skilled in the art, various modifications and variations of the present application are possible. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application shall be included in the scope of protection of the present application.

Claims

1. A method for detecting and compensating air tightness of automobile filters, which is used to detect air tightness of used automobile filters, characterized in that: The method comprises the following steps: S1. Place the automotive filter under test in a sealed state within the test chamber and apply a dynamically changing pressure to the chamber according to a preset time sequence. S2. While applying dynamically changing pressure, synchronously collect dynamic pressure signals and acoustic signals; S3. performing time domain and / or frequency domain analysis on the acoustic signal to obtain multiple acoustic features; S4. Analyze the dynamic response patterns of multiple acoustic features based on the dynamic pressure signal and extract leakage acoustic features that are strongly correlated with dynamic pressure changes; S5. Based on the leakage acoustic characteristics, determine whether the vehicle filter under test has a leak and the extent of the leak.

2. The automobile filter air tightness detection and compensation method according to claim 1, characterized in that: Step S1 includes: S101. Obtaining a preset time series pressure change pattern; the pressure change pattern includes the amplitude, frequency and duration of the pressure change; S102. Based on the pressure change pattern and the performance of the gas supply system, a pressure control signal is generated; the pressure control signal compensates for the response delay and nonlinear characteristics of the gas supply system to ensure that the actual pressure change in the detection cavity is consistent with the preset time series pressure change pattern; S103. Control the gas supply system, and according to the pressure control signal, use a closed-loop control method to apply a dynamically changing pressure to the detection cavity according to a preset time sequence.

3. The automobile filter air tightness detection and compensation method according to claim 2, characterized in that: Step S102 includes: generating an initial control signal based on the pressure variation pattern; Based on the pressure variation pattern and the pre-built air supply system transfer function model, an inverse model control algorithm is used to calculate the pre-compensation control signal; the air supply system transfer function model characterizes the pressure response delay and nonlinear characteristics of the air supply system; The pre-compensated control signal is added to the initial control signal to generate a pressure control signal.

4. The automobile filter air tightness detection and compensation method according to claim 1, characterized in that: Step S2 includes: S201. Synchronously collecting dynamic pressure signals and acoustic signals; S202. Preprocessing the synchronously collected dynamic pressure signal and acoustic signal respectively; the preprocessing includes bandpass filtering noise reduction and signal amplification; S203. Based on the preprocessed dynamic pressure signal, calculate the pressure change rate to obtain the pressure change rate characteristic; S204. Using a dynamic time warping algorithm, the preprocessed acoustic signal is time-aligned based on the pressure change rate characteristic to obtain a time-aligned acoustic signal.

5. The automobile filter air tightness detection and compensation method according to claim 1, characterized in that: Step S3 includes: S301. Pre-emphasize the acoustic signal to increase the energy of the high-frequency signal to compensate for the attenuation of high-frequency leakage sound during propagation; S302. Performing a short-time Fourier transform on the pre-emphasized acoustic signal to obtain a spectrum of the acoustic signal; S303. Based on the spectrogram of the acoustic signal, extract the total sound pressure level, spectral centroid, spectral bandwidth and Mel-frequency cepstral coefficients; S304. Perform time domain analysis on the pre-emphasized acoustic signal to calculate the peak factor and kurtosis; S305. Normalize the total sound pressure level, spectrum centroid, spectrum bandwidth, Mel-frequency cepstral coefficient, peak factor, and kurtosis to obtain the final acoustic features.

6. The automobile filter air tightness detection and compensation method according to claim 1, characterized in that: Step S4 includes: S401. For multiple acoustic features, calculate the mutual correlation coefficient between each acoustic feature and the dynamic pressure signal to obtain multiple mutual correlation coefficients; S402. Setting a mutual correlation coefficient threshold, comparing multiple mutual correlation coefficients with the mutual correlation coefficient threshold, screening out acoustic features with mutual correlation coefficients greater than the mutual correlation coefficient threshold, and obtaining a candidate leakage acoustic feature set; S403. For each candidate leakage acoustic feature in the candidate leakage acoustic feature set, calculate the energy change rate of the candidate leakage acoustic feature during the dynamic pressure change process to obtain multiple energy change rates; S404. Set an energy change rate threshold, compare multiple energy change rates with the energy change rate threshold, screen out candidate leakage acoustic features with energy change rates greater than the energy change rate threshold, and obtain leakage acoustic features.

7. The automobile filter air tightness detection and compensation method according to claim 6, characterized in that: Step S401 includes: The time window length and sliding step size for calculating the cross-correlation coefficient are set, and the dynamic pressure signal and each acoustic feature are divided into multiple time segments according to the time window length and sliding step size; For each acoustic feature, calculate the mutual correlation coefficient between the acoustic feature in each time segment and the dynamic pressure signal in the corresponding time segment to obtain the mutual correlation value of the acoustic feature corresponding to each time segment; For each acoustic feature, the mutual correlation values ​​of the acoustic feature corresponding to all time segments are weighted averaged to obtain the mutual correlation coefficient of the acoustic feature; wherein the weight of the time segment is proportional to the energy of the dynamic pressure signal in its corresponding time segment.

8. The automobile filter air tightness detection and compensation method according to claim 6, characterized in that: Step S403 includes: For each candidate leakage acoustic feature in the candidate leakage acoustic feature set, dividing the frequency spectrum of each candidate leakage acoustic feature into a plurality of frequency bands according to a preset frequency range; For each frequency band of each candidate leakage acoustic feature, the energy value in the frequency band is calculated to obtain multiple frequency band energy values; For each frequency band energy value of each candidate leakage acoustic feature, the rate of change of the frequency band energy value during the dynamic pressure change process is calculated to obtain multiple frequency band energy change rates; For each candidate leakage acoustic feature, the energy change rates of multiple frequency bands of the candidate leakage acoustic feature are weighted averaged according to a preset weight coefficient to obtain the energy change rate of the candidate leakage acoustic feature; wherein the preset weight coefficient is proportional to the center frequency of the frequency band.

9. The automobile filter air tightness detection and compensation method according to claim 1, characterized in that: Step S5 includes: S501. Calculate a weight factor for each leakage acoustic feature for a plurality of leakage acoustic features; wherein the weight factor is proportional to the correlation coefficient between the leakage acoustic feature and the dynamic pressure signal; S502. Perform weighted fusion on multiple leakage acoustic features according to the weight factor of each leakage acoustic feature to obtain a comprehensive leakage index; S503. Compare the comprehensive leakage index with a preset leakage judgment threshold. When the comprehensive leakage index is greater than the leakage judgment threshold, determine that the filter has leakage, and determine the leakage degree based on the size of the comprehensive leakage index.

10. An automobile filter air tightness detection and compensation system, used for performing air tightness detection on used automobile filters, characterized in that: The system includes a detection chamber, an air supply system, a pressure sensor, an acoustic sensor and an industrial computer; The detection cavity is used to seal and place the automobile filter to be tested; The gas supply system is used to deliver gas to the detection cavity to adjust the pressure of the detection cavity; The pressure sensor is used to detect the pressure signal of the detection cavity and upload it to the industrial computer; The acoustic sensor is used to detect the acoustic signal in the detection cavity and upload it to the industrial computer; The industrial computer is used to control the air supply system to apply a dynamically changing pressure to the test cavity according to a preset time sequence after the automobile filter seal to be tested is placed in the test cavity; during the process of applying the dynamically changing pressure, the dynamic pressure signal and the acoustic signal are synchronously collected through the pressure sensor and the acoustic sensor; Performing time domain and / or frequency domain analysis on acoustic signals to obtain multiple acoustic features; Based on the dynamic pressure signal, the dynamic response patterns of multiple acoustic features are analyzed to extract leakage acoustic features that are strongly correlated with dynamic pressure changes. Based on the leakage acoustic features, it is determined whether the vehicle filter under test is leaking and the extent of the leakage.

Images