Vehicle-mounted audio shell waterproof performance detection method based on intelligent sensing technology

Abstract

The application discloses a vehicle-mounted audio shell waterproof performance detection method based on intelligent sensing technology and belongs to the technical field of vehicle-mounted audio shell detection. The method comprises the following steps: obtaining the initial vibration characteristic spectrum of the vehicle-mounted audio shell in a simulated rain environment; inputting the initial vibration characteristic spectrum into a pre-trained shell-sound field coupling model to generate an estimated water seepage probability distribution of each region of the shell; selecting the first few positions with the highest probability on the surface of the shell according to the estimated water seepage probability distribution and arranging a flexible piezoelectric sensor array; starting the flexible piezoelectric sensor array to collect the real-time resonant impedance spectrum of each sensor and synchronously collecting the acoustic emission signal of the inner wall of the shell; if the impedance drop slope exceeds the first dynamic threshold and the acoustic emission energy cumulative value exceeds the second dynamic threshold, it is determined that water seepage occurs in the corresponding sensor region; recording the positions of all water seepage regions and the time when water seepage occurs, and generating a shell water seepage space-time distribution map.

Description

Technical Field

[0001] This invention relates to the field of vehicle audio housing testing technology, specifically a method for testing the waterproof performance of vehicle audio housings based on intelligent sensing technology. Background Technology

[0002] The waterproof performance of a car audio enclosure directly affects the reliability and lifespan of the audio system, requiring testing under simulated rain conditions. Current technologies primarily rely on uniformly distributing sensors across the enclosure surface or using experience to place sensors at localized locations, collecting single physical quantities for leakage detection. The uniform distribution method fails to consider the varying leakage risks across different areas of the enclosure due to differences in structural stiffness and joint morphology. A large number of sensors are located in low-risk areas, wasting resources and potentially masking early signs of leakage in high-risk areas due to low-sensitivity signals. The experience-based placement method heavily relies on manual judgment, lacking a systematic analysis of the enclosure's acoustic-vibration coupling characteristics, easily overlooking actual high-risk leakage areas. In the leakage detection stage, judgment is typically based on a single characteristic quantity, such as impedance amplitude change or acoustic emission event count, using a fixed threshold. However, due to background interference from water flow impact and vibration table excitation in a rain environment, single characteristic quantities are easily affected by environmental factors and fluctuate. Fixed thresholds are difficult to adapt to changes in operating conditions, leading to frequent false alarms or missed detections. Therefore, it is necessary to solve the problem of how to pre-identify high-risk areas based on the characteristics of the shell structure and optimize the sensor layout accordingly, as well as the problem of how to integrate multi-source sensor features and adapt to environmental changes to improve the accuracy of water seepage detection. Summary of the Invention

[0003] This invention provides a method for testing the waterproof performance of a vehicle audio enclosure based on intelligent sensing technology. The purpose is to predict the probability distribution of water seepage by utilizing the acoustic response characteristics of the enclosure under vibration excitation, thereby guiding the efficient deployment of piezoelectric sensor arrays. Furthermore, through the dynamic joint determination of impedance spectrum and acoustic emission signal, the method achieves accurate capture and spatiotemporal location recording of water seepage events.

[0004] The objective of this invention can be achieved through the following technical solutions: This invention provides a method for testing the waterproof performance of a vehicle audio enclosure based on intelligent sensing technology, comprising the following steps: The initial vibration characteristic spectrum of the car audio enclosure in a simulated rain environment is obtained, and the initial vibration characteristic spectrum is input into the pre-trained enclosure. In the acoustic field coupling model, the estimated probability distribution of water seepage in each region of the shell is generated.

[0005] As a preferred technical solution, the process of obtaining the initial vibration characteristic spectrum involves mounting the car audio enclosure on a vibration table, attaching single-axis accelerometers to nine predetermined geometric center points on the enclosure surface, and controlling the vibration table to excite the enclosure in a sinusoidal sweep pattern from low to high frequencies, covering a range of 20 Hz to 20 kHz at a sweep rate of one octave per minute. The time-domain vibration signals from the nine accelerometers are simultaneously acquired, and nine raw spectra are obtained through Fast Fourier Transform. The average amplitude of the nine raw spectra at the same frequency points is calculated and arranged in frequency order to generate the initial vibration characteristic spectrum. This method can obtain the basic dynamic response characteristics of the enclosure under rain conditions, providing a global input for subsequent water seepage probability prediction.

[0006] When generating the estimated seepage probability distribution, the pre-trained shell-acoustic field coupling model adopts a graph neural network architecture. The training samples include vibration characteristic spectra under different shell damage levels and measured seepage probability distributions. Then, the initial vibration characteristic spectra are discretized into frequencies. Amplitude vector sequence, input to the shell The first graph convolutional layer of the acoustic field coupling model has graph structure nodes corresponding to each element of the shell finite element mesh. After outputting the element displacement field characteristics, it is input to the shell. The second graph convolutional layer of the acoustic field coupling model has graph structure edge weights determined by the sound pressure transfer function between adjacent shell elements, ultimately passing through the shell... The fully connected mapping network of the acoustic field coupling model outputs a probability vector with the same number of finite element mesh elements as the shell. Each element in the probability vector represents the estimated seepage probability of the corresponding element's region. The probability vector is then remapped to the shell surface according to the geometric coordinates of the shell's finite element mesh elements. This process achieves a nonlinear mapping from vibration response to seepage probability, enabling high spatial resolution seepage risk pre-assessment results without damaging the shell.

[0007] Based on the estimated water seepage probability distribution, flexible piezoelectric sensor arrays are deployed at the highest probability locations on the shell surface. A preferred deployment method involves extracting the probability values ​​of all shell surface units from the estimated water seepage probability distribution and sorting them from highest to lowest. The top M shell surface units are selected, where M is obtained by dividing the total shell area by the effective detection area of ​​a single flexible piezoelectric sensor and rounding up. The geometric center coordinates of each of the top M shell surface units are obtained as deployment points. A flexible piezoelectric sensor is attached to each deployment point. All flexible piezoelectric sensors are connected in series via a flexible circuit board to form a sensor array, and the output is connected to an impedance analyzer. This deployment strategy concentrates limited sensor resources in the areas with the highest water seepage risk, effectively improving the targeting and efficiency of detection.

[0008] A flexible piezoelectric sensor array is activated to acquire the real-time resonant impedance spectrum of each sensor, while simultaneously acquiring acoustic emission signals from the inner wall of the housing. The preferred acquisition method involves sending a start command to an impedance analyzer. The analyzer scans each sensor in the array at a rate of fifty times per second, covering a frequency range of 100 Hz to 100 kHz, recording the real and imaginary parts of the impedance to form the real-time resonant impedance spectrum. Simultaneously, a synchronous trigger pulse is sent to a dual-channel acoustic emission acquisition card mounted at the center of the inner wall of the housing. The two channels of the dual-channel acoustic emission acquisition card are connected to broadband acoustic emission sensors attached to the left and right inner sides of the inner wall of the housing, respectively. The card continuously acquires two signals at a sampling rate of two million points per second and transmits them to a buffer as acoustic emission signals. This synchronous acquisition method ensures the temporal correlation between impedance response and acoustic emission activity, providing a basis for multi-physics joint criteria.

[0009] The real-time resonant impedance spectrum and acoustic emission signal are aligned with time windows, and the impedance drop slope and cumulative acoustic emission energy value within each time window are calculated. Specifically, the rising edge of the synchronous trigger pulse is taken as time zero. The real-time resonant impedance spectrum and acoustic emission signal are arranged along the same time axis. A time window with a length of 0.5 seconds and a sliding step of 0.1 seconds is set, and multiple continuous time windows are sequentially divided starting from time zero. For each time window, the real part of the impedance at 100 kHz for each sensor is extracted from the real-time resonant impedance spectrum within that window. These are then arranged in chronological order to form an impedance real part sequence and linearly fitted to obtain the impedance drop slope within that time window. For the same time window, the two acoustic emission signals are squared and integrated separately to obtain the energy values ​​of the two signals. The two energy values ​​are added together and the logarithm to the base 10 is taken to obtain the cumulative acoustic emission energy value. By continuously extracting features through sliding time windows, the subtle signal changes at the moment of seepage can be captured.

[0010] If the impedance drop slope exceeds the first dynamic threshold and the cumulative acoustic emission energy exceeds the second dynamic threshold, then water seepage is determined to have occurred in the corresponding sensor area.

[0011] As a preferred technical solution, the first dynamic threshold and the second dynamic threshold are recalculated once for each sliding time window during the detection process. The calculation method is to take the average of the impedance drop slope under the previous five consecutive time windows. and standard deviation and the average value of the cumulative acoustic emission energy and standard deviation Set the first dynamic threshold to the average value. Add three times the standard deviation Set the second dynamic threshold to the average value. Add three times the standard deviation The impedance drop slope of each sensor within the current time window is compared with a first dynamic threshold, and the cumulative acoustic emission energy is compared with a second dynamic threshold. When the impedance drop slope of any sensor exceeds the first dynamic threshold and the cumulative acoustic emission energy also exceeds the second dynamic threshold, the location of the sensor's deployment point is marked as a seepage area, and the end time of the current time window is recorded as the seepage occurrence time. The use of adaptive dynamic thresholds eliminates the risk of misjudgment caused by environmental fluctuations and individual differences, and the dual-criteria fusion logic ensures the reliability and accuracy of the seepage determination results.

[0012] Record the locations and times of all seepage areas to generate a spatiotemporal distribution map of shell seepage. When generating the map, create a spatiotemporal distribution data structure for shell seepage containing a list of regions and a time matrix. The region list stores the coordinates of all deployment points marked as seepage areas. The number of rows in the time matrix equals the length of the region list, and the number of columns equals the total number of time windows. For each marked deployment point, the value '1' is continuously entered into the column position corresponding to the current time window index in the row of the time matrix. If the same deployment point continues to meet the seepage condition in subsequent time windows, the value '1' is continuously entered into the subsequent column position of the corresponding row in the time matrix until the deployment point no longer meets the seepage condition. The coordinates of the deployment points in the region list are then linked row-wise with the time matrix to output the spatiotemporal distribution map of shell seepage. This map visually demonstrates the expansion of seepage locations over time.

[0013] As a technical solution of this invention, after generating the spatiotemporal distribution map of shell seepage, the method further includes tracing the seepage path of the map. The time window index of the first occurrence of a value of one in each seepage area is extracted from the time matrix and sorted. The area with the smallest time window index is selected as the seepage starting point area, and its geometric coordinates are extracted. The remaining seepage areas are traversed sequentially, and their Euclidean distances to the seepage starting point area are calculated and sorted by distance to generate a seepage diffusion sequence. The seepage occurrence times of two adjacent areas in the seepage diffusion sequence are subtracted to obtain the seepage diffusion time difference. The seepage diffusion time difference is divided by the corresponding Euclidean distance to obtain the seepage diffusion rate. This tracing process enables the location of the seepage source and a quantitative description of the diffusion rate.

[0014] Preferably, after obtaining the water diffusion rate, the spray parameters of the simulated rain environment are further corrected based on this rate. The preset spray water pressure, preset spray flow rate, and preset spray water temperature values ​​for the current simulated rain environment are obtained. The water diffusion rate is compared with a pre-stored standard diffusion rate threshold. If the water diffusion rate is greater than the standard threshold, the spray water pressure is reduced by the ratio of the diffusion rate to the standard threshold, and the spray flow rate is reduced by the ratio of the standard threshold to the water diffusion rate. If the water diffusion rate is less than the standard threshold, the spray water pressure is increased by the ratio of the standard threshold to the water diffusion rate, and the spray water temperature is increased by multiplying the ratio of the water diffusion rate to the standard threshold by 0.5. The corrected parameters are then sent to the programmable logic controller (PLC) of the simulated rain environment to update the spray parameters. Through closed-loop feedback of the water diffusion rate, dynamic matching between the rain test conditions and actual waterproof performance is achieved, improving the rationality and consistency of the testing process.

[0015] The beneficial effects of this invention are: The initial vibration characteristic spectrum of the car audio housing in a simulated rain environment was obtained and input into the pre-trained housing. A sound field coupling model is used to generate the estimated seepage probability distribution for each region of the shell. This coupling model uses a graph neural network architecture to learn the displacement field and sound pressure transmission relationship of the shell's finite element mesh elements under acoustic vibration, mapping the local stiffness attenuation and joint weakening features implicit in the vibration spectrum to seepage probabilities at the element scale. Based on this probability distribution, a flexible piezoelectric sensor array is deployed at several locations with the highest probability on the shell surface, concentrating the sensors in the high-probability seepage areas identified by the coupling model. Compared to uniform or empirically based placement methods, this data-driven sensor layout strategy significantly improves the probability of capturing impedance changes caused by early, minute seepage, avoids sensor redundancy in low-risk areas and detection blind spots in high-risk areas, thereby improving the detection sensitivity and spatial resolution of seepage signals without changing the number of sensors.

[0016] The real-time resonant impedance spectrum of each sensor acquired by the flexible piezoelectric sensor array is aligned with the acoustic emission signal synchronously acquired from the inner wall of the shell within a time window. The impedance drop slope and the cumulative acoustic emission energy value within each time window are calculated. The impedance drop slope reflects the continuous attenuation trend of the real part of the impedance at the resonant frequency of the piezoelectric sensor when water seeps into the joint or microcrack. The cumulative acoustic emission energy value corresponds to the transient elastic wave energy released by the interaction between the water flow and the shell material during the seepage process. Seepage is determined by the simultaneous exceeding of two parameters' respective dynamic thresholds. The first and second dynamic thresholds are recalculated every time a time window is slid during the detection process, and are automatically updated based on the statistical characteristics of the five consecutive time windows preceding the current time window plus three times the standard deviation. The adaptive adjustment of the dynamic thresholds with environmental noise and vibration conditions weakens the instantaneous fluctuations caused by rainwater impact and vibration table frequency sweep interference. The dual-parameter cross-verification mechanism suppresses false triggering that may occur if only a single impedance or acoustic emission amplitude is used, making the determination of the seepage time more reliable, effectively reducing the false alarm rate and accurately recording the time and area of ​​seepage. Attached Figure Description

[0017] The invention will now be further described with reference to the accompanying drawings.

[0018] Figure 1 This is a flowchart of the method for testing the waterproof performance of car audio enclosures; Figure 2 This is a flowchart of the method for deploying a flexible piezoelectric sensor array on the shell surface; Figure 3 This is a flowchart of dynamic threshold seepage determination and spatiotemporal distribution map generation; Figure 4 It is a flowchart for generating the spatiotemporal distribution map of water seepage in the shell, tracing the seepage path, and correcting the spray parameters. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] Example See Figure 1 This invention provides a method for testing the waterproof performance of a vehicle audio enclosure based on intelligent sensing technology. The overall implementation scheme is as follows: The initial vibration characteristic spectrum of the vehicle audio enclosure in a simulated rain environment is obtained. The initial vibration characteristic spectrum is input into a pre-trained enclosure-sound field coupling model to generate the estimated water seepage probability distribution of each region of the enclosure. Based on the estimated water seepage probability distribution, a flexible piezoelectric sensor array is deployed at the top several locations with the highest probability on the enclosure surface. The flexible piezoelectric sensor array is activated to collect the real-time resonant impedance spectrum of each sensor and simultaneously collect the acoustic emission signal of the inner wall of the enclosure. The real-time resonant impedance spectrum and the acoustic emission signal are aligned with time windows, and the impedance drop slope and the cumulative acoustic emission energy value within each time window are calculated. If the impedance drop slope exceeds the first dynamic threshold and the cumulative acoustic emission energy value exceeds the second dynamic threshold, water seepage is determined to have occurred in the corresponding sensor area. The location and time of all water seepage areas are recorded to generate a spatiotemporal distribution map of water seepage in the enclosure.

[0021] In practical implementation, the method for obtaining the initial vibration characteristic spectrum of the vehicle audio enclosure in a simulated rain environment is as follows: The car audio system housing was mounted on a vibration table, and single-axis accelerometers were attached to nine pre-defined geometric center points on the housing surface. These nine geometric center points were determined by dividing the housing surface into nine equal-area rectangular regions, with the intersection of the diagonals of each region serving as the geometric center. The sensitivity direction of the single-axis accelerometer was perpendicular to the tangential plane of the housing surface at its respective geometric center point. When attaching the single-axis accelerometers, cyanoacrylate adhesive was used to bond the bottom surface of the single-axis accelerometer to the housing surface, with the adhesive layer thickness controlled to 0.05 mm.

[0022] The vibration table excites the shell using a sinusoidal sweep frequency pattern from low to high frequencies, with a sweep frequency range of 20 Hz to 20 kHz and a sweep rate of one octave per minute. The sinusoidal sweep excitation signal is generated by a signal generator and then amplified by a power amplifier to drive the vibration table. The acceleration amplitude of the vibration table surface remains constant throughout the sweep process, set at 1g. During the sweep, a data acquisition system simultaneously acquires time-domain vibration signals from nine single-axis accelerometers at a sampling rate of 50 kHz; the length of each time-domain vibration signal is equal to the sweep duration.

[0023] A Fast Fourier Transform (FFT) was performed on each time-domain vibration signal to obtain nine raw spectra. The FFT processing parameters were as follows: a Hanning window was applied to the time-domain vibration signal, the window function was of the same length as the signal, and the number of Fourier transform points was twice the number of signal sampling points.

[0024] The amplitudes of the nine original spectra at the same frequency point are arithmetically averaged to calculate the average amplitude. These average amplitudes are then arranged in ascending order of frequency to generate the initial vibration characteristic spectrum. The initial vibration characteristic spectrum is a one-dimensional array where each element corresponds to a discrete frequency point, and the value of each element is the arithmetic mean of the amplitudes of the nine original spectra at that frequency point.

[0025] In practical implementation, the initial vibration feature spectrum is input into the pre-trained shell-sound field coupling model to generate the estimated seepage probability distribution of each region of the shell. The shell-sound field coupling model adopts a graph neural network architecture, comprising three modules: a first graph convolutional layer, a second graph convolutional layer, and a fully connected mapping network. The graph structure of the first graph convolutional layer is defined by the shell finite element mesh. The nodes of the graph structure correspond to the individual elements of the shell finite element mesh. The adjacency matrix of the graph structure is determined by the adjacency relationship of the elements in the shell finite element mesh. If two finite element mesh elements share an edge, the corresponding adjacency matrix element value is 1; otherwise, it is 0. The input feature dimension of the first graph convolutional layer is the length of the initial vibration feature spectrum, and the output feature dimension of the first graph convolutional layer is set to 64 dimensions. The convolution operation formula for the first graph convolutional layer is:

[0026] in, Indicates the first The node feature matrix of the layer graph convolutional layer, the first Each row of the node feature matrix of the graph convolutional layer corresponds to the feature vector of a finite element mesh cell at the output of the graph convolutional layer. Indicates the first The learnable weight matrix of the layer graph convolutional layer is updated using the gradient descent algorithm during the neural network training process; This represents the adjacency matrix of the graph structure after adding self-loops. The adjacency matrix of the graph structure after adding self-loops is obtained by adding the original adjacency matrix and the identity matrix. express The degree matrix, The diagonal elements are The sum of all non-zero elements in the corresponding row, and all non-diagonal elements are zero; This represents the ReLU activation function. The first convolutional layer receives the frequency-amplitude vector sequence formed after discretizing the initial vibration feature spectrum as the input node feature matrix. The frequency-amplitude vector sequence is constructed by arranging the average amplitude corresponding to each frequency point of the initial vibration feature spectrum in frequency order to form a vector. The length of the vector is equal to the number of frequency points contained in the initial vibration feature spectrum.

[0027] After processing by the first convolutional layer, the unit displacement field features are output. These features are the node feature matrix output by the first convolutional layer, with a dimension equal to the number of finite element mesh elements multiplied by 64. The unit displacement field features are then input into the second convolutional layer. The graph structure nodes of the second convolutional layer are identical to those of the first convolutional layer. The edge weights of the adjacency matrix of the second convolutional layer are determined by the acoustic pressure transfer function between adjacent shell elements.

[0028] The sound pressure transfer function (SPD) is calculated as follows: For each pair of adjacent finite element mesh elements, a unit volume sound source excitation is applied to the geometric center of one element, and the sound pressure response is calculated at the geometric center of the other element. The ratio of the amplitude of the sound pressure response to the amplitude of the unit volume sound source excitation is used as the SPD value. The frequency range for calculating the SPD value is consistent with the frequency range of the initial vibration characteristic spectrum, and the SPD value is taken as the average of the amplitude-frequency response at all frequency points. The input feature of the second convolutional layer is the element displacement field feature output by the first convolutional layer, and the output feature dimension of the second convolutional layer is set to 32-dimensional. The convolution operation formula of the second convolutional layer is the same as that of the first convolutional layer. The adjacency matrix of the second convolutional layer is composed of the SPD values, which are directly used as the edge weights of the corresponding adjacent element positions in the adjacency matrix.

[0029] The output of the second convolutional layer is input to a fully connected mapping network, which consists of two fully connected layers. The first fully connected layer has an input dimension equal to the number of finite element mesh elements multiplied by 32, an output dimension of 128, and uses the ReLU activation function. The second fully connected layer has an input dimension of 128, an output dimension equal to the number of finite element mesh elements in the shell, and uses the Sigmoid activation function. The fully connected mapping network outputs a probability vector with a length equal to the number of finite element mesh elements in the shell. Each element in the probability vector ranges from 0 to 1, representing the estimated seepage probability of the corresponding element's region. The probability vector is remapped onto the shell surface according to the geometric coordinates of the shell finite element mesh elements, resulting in the estimated seepage probability distribution for each region. The remapping is achieved by assigning the probability value of each finite element mesh element to its corresponding geometric region on the shell surface, thus forming a seepage probability distribution map for each region on the shell surface.

[0030] The training process of the shell-sound field coupling model is as follows: Training samples are collected, including vibration characteristic spectra under different shell damage levels and measured seepage probability distributions. The method for obtaining the vibration characteristic spectra in the training samples is the same as that for obtaining the initial vibration characteristic spectra. The vibration characteristic spectra in the training samples are used as input to the shell-sound field coupling model, and the measured seepage probability distribution in the training samples is used as the supervision label for the shell-sound field coupling model. The mean squared error loss function is used to calculate the loss value between the model output probability vector and the measured seepage probability distribution. The Adam optimizer is used for gradient backpropagation to update the model parameters. The learning rate is set to 0.001, the batch size is set to 32, and the number of training iterations is set to 500 rounds. Training is terminated early when the validation set loss value no longer decreases in 20 consecutive iterations. The model parameters with the lowest validation set loss value are saved as the pre-trained shell-sound field coupling model.

[0031] The above-mentioned measured water seepage probability distribution was obtained by: conducting simulated rain tests on car audio housings with different degrees of damage; using dyeing penetrating liquid to indicate the water seepage location during the rain test; taking images of the dye distribution on the housing surface; segmenting the dyed area using an image segmentation algorithm; and calculating the measured water seepage probability of each unit based on the ratio of the dyed area to the total area of ​​the unit.

[0032] In practical implementation, based on the estimated water seepage probability distribution, the method for deploying a flexible piezoelectric sensor array at the top several locations with the highest probability on the shell surface is as follows: extract the estimated water seepage probability values ​​of all shell surface units from the estimated water seepage probability distribution, see [reference needed]. Figure 2 The estimated water seepage probability distribution is formed by remapping the probability vector output by the shell-acoustic field coupling model onto the shell surface according to the geometric coordinates of the shell finite element mesh element. Each shell surface element corresponds to an estimated water seepage probability value, and the numerical range of the estimated water seepage probability value is a real number between 0 and 1.

[0033] The estimated water seepage probability values ​​of all extracted shell surface elements are sorted from highest to lowest in descending order, with the shell surface element with the highest estimated water seepage probability value placed first in the sequence. If multiple shell surface elements have the same estimated water seepage probability value, their order is determined by their index number in the finite element mesh, from smallest to largest.

[0034] Select the first M shell surface elements from the sorted sequence. The value of M is determined by the following formula:

[0035] in, This indicates the number of shell surface units to be selected. This indicates the rounding up operation; The total area of ​​the shell is expressed in square millimeters. The total area of ​​the shell is obtained by summing the areas of all finite element mesh elements on the shell surface. The area of ​​each finite element mesh element is obtained from the element geometry information when the finite element mesh is generated. This represents the effective detection area of ​​a single flexible piezoelectric sensor, measured in square millimeters. The effective detection area is 80% of the area of ​​the piezoelectric ceramic sheet, which is circular with a diameter of 8 millimeters. The area of ​​the ceramic sheet is calculated using the formula for the area of ​​a circle, and the effective detection area of ​​a single flexible piezoelectric sensor is 40.21 square millimeters. The rounding up operation rounds the result of a division to the nearest integer towards positive infinity. When the decimal part of the result is zero, M equals the integer part of the result; when the decimal part is greater than zero, M equals the integer part of the result plus one.

[0036] After selecting the first M shell surface elements, the geometric center coordinates of each of the first M shell surface elements are obtained. The geometric center coordinates are obtained as follows: The coordinate values ​​of all nodes of each shell surface element are read from the element geometry information of the shell finite element mesh. The x-coordinates of all nodes of the shell surface element are summed and divided by the number of nodes to obtain the x-coordinate of the geometric center point. The y-coordinates of all nodes of the shell surface element are summed and divided by the number of nodes to obtain the y-coordinate of the geometric center point. The z-coordinates of all nodes of the shell surface element are summed and divided by the number of nodes to obtain the z-coordinate of the geometric center point. The x, y, and z coordinates constitute the geometric center coordinates of the shell surface element. Each geometric center coordinate is used as a placement point. The position of the placement point is determined by the geometric center coordinates of the corresponding shell surface element, and the position coordinates of the placement point are completely consistent with the geometric center coordinates of the corresponding shell surface element.

[0037] At each deployment point, a flexible piezoelectric sensor is attached. The flexible piezoelectric sensor consists of a piezoelectric ceramic sheet and a flexible polymer substrate, with the polarization direction of the piezoelectric ceramic sheet perpendicular to the attachment surface of the flexible piezoelectric sensor. The attachment method involves uniformly coating the bottom surface of the flexible piezoelectric sensor onto the deployment point using epoxy resin adhesive, with a coating thickness controlled to 0.1 mm. A uniform pressure of 0.5 Newtons is applied and maintained for 5 minutes to allow the epoxy resin adhesive to cure. All the attached flexible piezoelectric sensors are connected in series via a flexible circuit board to form a flexible piezoelectric sensor array. The flexible circuit board has pads corresponding to the number of flexible piezoelectric sensors printed on it. The positive and negative electrodes of each flexible piezoelectric sensor are soldered to the corresponding positive and negative conductive lines on the flexible circuit board, respectively. The conductive lines on the flexible circuit board connect the various flexible piezoelectric sensors in series, with the negative output terminal of the previous flexible piezoelectric sensor connected to the positive input terminal of the next flexible piezoelectric sensor. The output of the flexible circuit board is connected to the input port of the impedance analyzer via a shielded coaxial cable. The characteristic impedance of the shielded coaxial cable is 50 ohms, and the outer shield of the shielded coaxial cable is connected to the ground terminal of the impedance analyzer.

[0038] In practical implementation, the method for activating the flexible piezoelectric sensor array to acquire the real-time resonant impedance spectrum of each flexible piezoelectric sensor, and simultaneously acquiring the acoustic emission signal from the inner wall of the housing, is as follows: A start command is sent to the impedance analyzer. The start command is a standard command code sent via a universal interface bus. After receiving the start command, the impedance analyzer scans the flexible piezoelectric sensors in the array one by one according to preset parameters at a scan rate of fifty times per second. The impedance analyzer contains a multiplexer module, which selects the currently scanned flexible piezoelectric sensor sequentially according to the series connection order of the flexible piezoelectric sensors on the flexible circuit board during each scan. Each scan operates within a frequency range of 100 Hz to 100 kHz, employing a logarithmic frequency scan method with 200 frequency points. During each scan, the impedance analyzer measures the real and imaginary parts of the impedance of the currently scanned flexible piezoelectric sensor at each frequency point. The real part of the impedance is the real component of the complex impedance, measured in ohms, and the imaginary part is the imaginary component of the complex impedance, also measured in ohms. The real and imaginary parts of the impedance recorded for all frequency points in a single complete scan of a flexible piezoelectric sensor constitute a real-time resonant impedance spectrum frame for that sensor. The impedance analyzer repeats this scanning process 50 times per second. After completing a full scan of all flexible piezoelectric sensors, all real-time resonant impedance spectrum frames acquired in this round are packaged and sent to the host computer's buffer via a data transmission interface.

[0039] Simultaneously, a synchronization trigger pulse is sent to a dual-channel acoustic emission acquisition card installed at the center of the inner wall of the housing. This synchronization trigger pulse is generated by the same host computer via a digital input / output board, with a pulse width of 10 microseconds and an amplitude of 5 volts. The dual-channel acoustic emission acquisition card includes a first acquisition channel and a second acquisition channel. The first acquisition channel is connected to a broadband acoustic emission sensor attached to the left inner side of the inner wall of the housing, and the second acquisition channel is connected to a broadband acoustic emission sensor attached to the right inner side of the inner wall of the housing. The broadband acoustic emission sensor has a frequency response range of 100 Hz to 1 MHz. The broadband acoustic emission sensor is coupled to the surface of the inner wall of the housing via a coupling agent, which is vacuum silicone grease, with a coupling agent layer thickness of 0.2 mm. After receiving the rising edge of the synchronization trigger pulse, the dual-channel acoustic emission acquisition card performs continuous analog-to-digital conversion on the output signals of the two broadband acoustic emission sensors at a sampling rate of two million points per second. The analog-to-digital conversion resolution is 16 bits, and the input range is ±10 volts. The dual-channel acoustic emission acquisition card transmits the two digital signals after analog-to-digital conversion to the buffer area in real time. The two digital signals are labeled as the first acoustic emission signal and the second acoustic emission signal, respectively. The first acoustic emission signal corresponds to the broadband acoustic emission sensor on the left inner side of the inner wall of the housing, and the second acoustic emission signal corresponds to the broadband acoustic emission sensor on the right inner side of the inner wall of the housing. The two digital signals together constitute the acoustic emission signal.

[0040] In practical implementation, the real-time resonant impedance spectrum and acoustic emission signal are aligned with time windows. The method for calculating the impedance drop slope and cumulative acoustic emission energy within each time window is as follows: the rising edge of the synchronization trigger pulse sent by the host computer is taken as the zero time point. The accuracy of the zero time point is determined by the clock frequency of the digital input / output board, which is 100 MHz, and the time resolution of the zero time point is 10 nanoseconds. The timestamps of each frame of the real-time resonant impedance spectrum and the timestamps of each sampling point of the acoustic emission signal are arranged on the same time axis. The timestamps of the real-time resonant impedance spectrum are recorded by the impedance analyzer at the end of each scan, with a timestamp accuracy of 1 millisecond. The timestamps of the acoustic emission signal are recorded by the dual-channel acoustic emission acquisition card at each sampling, with a timestamp accuracy of 0.5 microseconds. The arrangement is as follows: all timestamps are mapped onto a continuous time axis with the rising edge of the synchronization trigger pulse as the zero time point, so that the real-time resonant impedance spectrum and the acoustic emission signal are synchronously aligned on the time axis.

[0041] Define a fixed-length time window of 0.5 seconds. The start time of the time window is denoted as the start time, and the end time as the end time. The difference between the start and end times is 0.5 seconds. Set the sliding step of the time window to 0.1 seconds. Starting from time zero, divide the time window into multiple consecutive windows. The first time window starts at time zero and ends at time zero plus 0.5 seconds; the second time window starts at time zero plus 0.1 seconds and ends at time zero plus 0.6 seconds; and so on. The start time of each subsequent time window is 0.1 seconds longer than the start time of the previous time window.

[0042] For each time window, the real part of the impedance of each flexible piezoelectric sensor at a frequency of 100 kHz is extracted from the real-time resonant impedance spectrum within that time window. The extraction method is as follows: for each frame of the real-time resonant impedance spectrum contained within the time window, the measurement point at a frequency of 100 kHz is located in the corresponding scan record, and the real part of the impedance at that measurement point is read. The real part of the impedance extracted from all real-time resonant impedance spectrum frames of a single flexible piezoelectric sensor within that time window is arranged in ascending order of timestamps, forming the real part sequence of the impedance of that flexible piezoelectric sensor within that time window. The length of the real part sequence is equal to the number of times the flexible piezoelectric sensor is scanned within that time window. Since the impedance analyzer's scan rate is 50 times per second, the time window length is 0.5 seconds, and the length of the real part sequence is 25. A linear fit is performed on the impedance real part sequence using the least squares method. The relative time corresponding to each value in the impedance real part sequence is used as the independent variable; the relative time is the difference between the timestamp corresponding to that value and the start time of the time window. The impedance real part value is used as the dependent variable. A straight line is obtained, and the slope of this line is taken as the impedance decrease slope for the flexible piezoelectric sensor within that time window. The impedance decrease slope is expressed as the change in the impedance real part within 0.1 seconds, representing the rate of decrease of the impedance real part value per unit time.

[0043] For the same time window, the energy values ​​of the two acoustic emission signals are obtained by performing square integration on the two signals respectively. The square integration is calculated as follows: All sampling points from the first acoustic emission signal between the start and end of the time window are extracted, the amplitude of each sampling point is squared, and the sum is accumulated. The accumulated result is multiplied by the sampling interval to obtain the first energy value. Similarly, all sampling points from the second acoustic emission signal between the start and end of the time window are extracted, the amplitude of each sampling point is squared, and the sum is accumulated. The accumulated result is multiplied by the sampling interval to obtain the second energy value. The sampling interval is the reciprocal of the sampling rate, i.e., the sampling interval is equal to 0.5 microseconds. The first and second energy values ​​are added together, and the logarithm to the base 10 is taken to obtain the cumulative acoustic emission energy value. The formula for calculating the cumulative acoustic emission energy value is as follows:

[0044] in, This represents the cumulative acoustic emission energy value, which is a dimensionless value. Represents the logarithmic function with base 10; This indicates the sampling interval, which is 0.5 microseconds and is determined by a sampling rate of two million points per second. This indicates the number of sampling points for the acoustic emission signal within the time window. The value is the time window length of 0.5 seconds multiplied by the sampling rate of 2 million points per second, i.e. It equals 1,000,000; Indicates the sampling point index. The value ranges from 1 to Positive integers; Indicates the first sound transmission signal at the 1st The amplitude at each sampling point, the first transmitted signal at the... The amplitude at each sampling point is acquired by the first acquisition channel of the dual-channel acoustic emission acquisition card, and the unit is volts; This indicates that the second acoustic transmission signal was at the... The amplitude at each sampling point, the second acoustic transmission signal at the... The amplitude at each sampling point is acquired by the second acquisition channel of the dual-channel acoustic emission acquisition card, and the unit is volts; This indicates the entirety of the first transmitted signal within the time window. Sum the squares of the amplitudes at each sampling point; This indicates the entirety of the second acoustic transmission signal within the time window. Sum the squares of the amplitudes at each sampling point.

[0045] In specific implementation, please refer to Figure 3The process of comparing the impedance drop slope of the current time window with the first dynamic threshold is as follows: For the current time window, the impedance drop slope of all flexible piezoelectric sensors within the five consecutive time windows preceding the current time window is obtained. This is done by reading the impedance drop slope values ​​of all flexible piezoelectric sensors within the first, second, third, fourth, and fifth time windows preceding the current time window from the buffer. These impedance drop slope values ​​were generated and stored during the time window calculation process in the above embodiment. The impedance drop slopes belonging to the same flexible piezoelectric sensor within the five obtained time windows are treated as a set of data, with each flexible piezoelectric sensor corresponding to a set of data containing five impedance drop slope values.

[0046] Calculate the average of the five impedance drop slope values ​​corresponding to each flexible piezoelectric sensor. ,average value The calculation method involves summing the five impedance drop slope values ​​and dividing by five to obtain the average impedance drop slope. The standard deviation is then calculated for the five impedance drop slope values ​​corresponding to each flexible piezoelectric sensor. Standard deviation The calculation method is as follows: calculate the difference between each of the five impedance drop slope values ​​and the average impedance drop slope, square each of the five differences, sum the results, divide the sum by four, and calculate the square root to obtain the standard deviation of the impedance drop slope. The first dynamic threshold for each flexible piezoelectric sensor within the current time window is set to the average impedance drop slope plus three times the corresponding standard deviation of the impedance drop slope. The three-times factor is determined based on the assumption that, under the normal distribution, the probability of a value falling outside the range of the average plus or minus three standard deviations is less than 0.3%. Setting this threshold to three standard deviations effectively distinguishes between abnormal impedance drops caused by water seepage and normal impedance fluctuations caused by environmental noise.

[0047] The second dynamic threshold process for calculating the cumulative acoustic emission energy value of the current time window is as follows: Obtain the cumulative acoustic emission energy values ​​of the five consecutive time windows preceding the current time window. These cumulative acoustic emission energy values ​​have been generated and stored during the calculation of each time window in the above embodiment. Combine the five cumulative acoustic emission energy values ​​into a dataset. Calculate the average value of the five cumulative acoustic emission energy values. The calculation method involves summing the five cumulative acoustic emission energy values ​​and dividing by five to obtain the average cumulative acoustic emission energy value. The standard deviation of the five cumulative acoustic emission energy values ​​is then calculated. The calculation method involves taking the difference between each of the five cumulative acoustic emission energy values ​​and the average cumulative acoustic emission energy value. Each of the five differences is squared, and the sum is calculated. The square root of the sum is then calculated to obtain the standard deviation of the cumulative acoustic emission energy value. The second dynamic threshold for the current time window is set to the average cumulative acoustic emission energy value plus three times the standard deviation of the cumulative acoustic emission energy value. The determination of this three-fold factor is based on the same criteria as the determination of the three-fold factor in the first dynamic threshold.

[0048] The first and second dynamic thresholds are recalculated every time a time window slides during the detection process. The recalculation is triggered at the time window slide moment, which is controlled by the time window index flag signal generated by the host computer. When the time window index number increases, the host computer calls the dynamic threshold calculation program module. The dynamic threshold calculation program module reads the impedance drop slope and acoustic emission energy accumulation value of the latest five consecutive time windows, re-executes the average value calculation, standard deviation calculation and threshold setting operation, and updates the values ​​of the first and second dynamic thresholds.

[0049] The impedance drop slope of each flexible piezoelectric sensor within the current time window is compared one by one with the corresponding first dynamic threshold. The comparison is performed by comparing the impedance drop slope value of a single flexible piezoelectric sensor within the current time window with the corresponding first dynamic threshold value. The accumulated acoustic emission energy value of the current time window is then compared with the second dynamic threshold value. If, within the current time window, the impedance drop slope of any flexible piezoelectric sensor exceeds the corresponding first dynamic threshold, and simultaneously the accumulated acoustic emission energy value of the current time window exceeds the second dynamic threshold, then the location of the flexible piezoelectric sensor that meets the conditions is marked as a seepage area, and the end time of the current time window is recorded as the time of seepage occurrence in the seepage area. The end time of the time window is equal to the start time of the time window plus the window length of 0.5 seconds. The start time of the time window is determined by the time window index number and the sliding step size.

[0050] In practical implementation, the method for recording the location and time of all seepage areas to generate a spatiotemporal distribution map of shell seepage is as follows: A spatiotemporal distribution data structure for shell seepage is created, containing a list of regions and a time matrix. The region list is a one-dimensional array used to store the coordinates of all deployment points marked as seepage areas. Each deployment point coordinate consists of three floating-point numbers, representing the x-axis, y-axis, and z-axis coordinates of the deployment point in the geometric coordinate system of the shell surface. The time matrix is ​​a two-dimensional integer array. The number of rows in the time matrix equals the length of the region list, and the number of columns in the time matrix equals the total number of time windows divided from time zero to the end of the detection. The total number of time windows is calculated by subtracting the time window length of 0.5 seconds from the total detection duration, dividing by the sliding step size of 0.1 seconds, rounding the result down, and then adding one.

[0051] During the detection process of marking seepage areas, for each deployment point marked as a seepage area in the current time window, the system searches the area list to see if a record of that deployment point's coordinates already exists. The search method involves comparing the coordinates of the currently marked deployment point with the coordinates of all existing deployment points in the area list. If the coordinates of the deployment point do not exist in the area list, the coordinates of the currently marked deployment point are appended to the end of the area list. After appending, the length of the area list increases by one, and the number of rows in the time matrix increases accordingly. All elements in the newly added row are initialized to zero. If the coordinates of the deployment point already exist in the area list, the index position of that deployment point in the area list is obtained. For the row index position of the deployment point marked as a seepage area in the area list, the value 1 is entered into the column position of the time matrix corresponding to the current time window index. The time window index starts from zero, and each consecutive time window corresponds to an incrementally increasing time window index.

[0052] If the same deployment point continuously meets the seepage detection criteria in subsequent time windows, after each subsequent time window's detection period, if the deployment point is marked as a seepage area again, the value '1' will be added to the column position corresponding to the current time window index in the corresponding row of the time matrix, overwriting the original value of '0' in that column position. If the same deployment point no longer meets the seepage detection criteria in a certain time window, the addition of the value '1' will stop, and the value at the column position of the corresponding time window index in the corresponding row of the time matrix will remain zero. The deployment point coordinates in the region list are associated with the corresponding row's numerical sequence in the time matrix row by row. Each deployment point coordinate in the region list and the corresponding row's sequence of zeros and '1's in the time matrix are treated as a single entry. The collection of all such entries constitutes the spatiotemporal distribution map of shell seepage, which is output to the storage medium as a data file.

[0053] In practical implementation, the pre-trained shell-sound field coupling model adopts a graph neural network architecture. The core structure of the graph neural network architecture includes three modules: a first graph convolutional layer, a second graph convolutional layer, and a fully connected mapping network. The module composition and hierarchical connection relationship have been described in the above embodiment. See also... Figure 4 The training samples for the shell-sound field coupling model include vibration characteristic spectra under different shell damage levels and measured water seepage probability distributions. The different shell damage levels are defined as follows: Vehicle audio shell samples with pre-fabricated defects are prepared. These defects include through holes with diameters of 0.5 mm, 1.0 mm, 1.5 mm, and 2.0 mm drilled at different locations on the shell surface. Five different locations are set for each diameter of through hole, resulting in a total of twenty defect states. Additionally, five defect-free, intact shell samples are set, for a total of twenty-five shell damage levels. The method for acquiring vibration characteristic spectra under each shell damage level is consistent with the method for acquiring the initial vibration characteristic spectrum in the previous embodiment. During acquisition, the shell is mounted on a vibration table, and single-axis accelerometers are attached to nine pre-set geometric center points on the shell surface. The vibration table is controlled to perform sinusoidal frequency sweep excitation from 20 Hz to 20 kHz at a sweep rate of one octave per minute. The time-domain vibration signals from the nine single-axis accelerometers are simultaneously acquired and processed using fast Fourier transform and amplitude averaging to generate vibration characteristic spectra corresponding to each shell damage level. The method for obtaining the measured water seepage probability distribution under each degree of shell damage is as follows: Shell samples with corresponding damage degrees are placed in a simulated rain environment. The spray water pressure of the simulated rain environment is set to 0.2 MPa, the spray flow rate is set to 10 liters per minute, and the spray duration is set to 10 minutes. After the spraying is completed, a dyeing penetrating liquid is sprayed onto the shell surface. After the dyeing penetrating liquid penetrates the shell cracks and micropores, it forms a dyed area on the shell surface. The dyed distribution image on the shell surface is captured by an industrial camera. The image segmentation algorithm is used to segment the pixels of the dyed area and the pixels of the undyed area in the image. A mapping relationship is established between the geometric coordinates of the shell finite element mesh unit and the pixel coordinates of the dyed distribution image. The ratio of the number of dyed pixels in each finite element mesh unit to the total number of pixels is calculated as the measured water seepage probability of the finite element mesh unit. The measured water seepage probabilities of all finite element mesh units constitute the measured water seepage probability distribution. The training steps, loss function, optimizer parameters, learning rate, batch size, number of training iterations, and early termination conditions of the shell-sound field coupling model have been described in the above embodiment. During training, the vibration feature spectrum is used as the input feature of the first graph convolutional layer of the shell-sound field coupling model, and the measured water seepage probability distribution is used as the supervision label.

[0054] In the specific implementation, after generating the spatiotemporal distribution map of shell seepage, the process also includes tracing the seepage path. The seepage path tracing is implemented as follows: Extract the time window index of the first occurrence of a value of one for each seepage area from the time matrix of the shell seepage spatiotemporal distribution map. The rows of the time matrix correspond to the coordinates of deployment points stored in the region list. For each row of the time matrix, scan the values ​​of each column from left to right. When a column value of one is encountered, record the column index as the time window index of the first occurrence of a value of one for that seepage area. Stop scanning that row and continue the same scanning operation for the next row until all rows of the time matrix have been scanned. Sort all extracted seepage areas in ascending order of their time window indices, with the seepage area with the smallest time window index at the beginning of the sequence and the seepage area with the largest time window index at the end. The seepage starting region is selected from the sorted sequence with the smallest time window index. The geometric coordinates of this region on the shell surface are extracted from the region list; these coordinates consist of x-axis, y-axis, and z-axis values. The remaining seepage regions in the sorted sequence, excluding the starting region, are traversed sequentially. For each traversed region, the Euclidean distance between the remaining region and the starting region on the shell surface is calculated. The Euclidean distance is calculated by obtaining the geometric coordinates of the starting region from the region list and denoting them as coordinate points. Obtain the geometric coordinates of the remaining seepage areas that are currently being traversed, and denote them as coordinate points. Euclidean distance on the shell surface The calculation formula is:

[0055] in, This represents the Euclidean distance between the seepage initiation area and the remaining seepage areas in the geometric coordinate system of the shell surface, in millimeters; The x-axis coordinate value represents the geometric coordinates of the region where the seepage begins, in millimeters; The y-axis coordinate value represents the geometric coordinates of the region where the seepage begins, in millimeters; The z-axis coordinate value represents the geometric coordinates of the region where the seepage begins, in millimeters; The x-axis coordinate value represents the geometric coordinates of the remaining seepage areas that have been traversed, in millimeters; The y-axis coordinate value represents the geometric coordinates of the remaining seepage areas that have been traversed, in millimeters; The z-axis coordinate value represents the geometric coordinates of the remaining seepage areas that have been traversed, in millimeters; This represents the square of the difference between the x-axis coordinate value of the seepage initiation area and the x-axis coordinate values ​​of the remaining seepage areas; This represents the square of the difference between the y-axis coordinate value of the seepage initiation area and the y-axis coordinate values ​​of the remaining seepage areas; This represents the square of the difference between the z-axis coordinate value of the seepage initiation region and the z-axis coordinate values ​​of the remaining seepage regions; This represents the square root operation.

[0056] After calculating the Euclidean distances between all remaining seepage areas and the seepage initiation area on the shell surface, the remaining seepage areas are sorted in ascending order of Euclidean distance to generate a seepage diffusion sequence. The first element of the seepage diffusion sequence is the seepage initiation area, and subsequent elements are arranged in ascending order of Euclidean distance. The seepage occurrence time difference is obtained by subtracting the seepage occurrence times of two adjacent seepage areas in the seepage diffusion sequence. The seepage occurrence time is obtained from the end time of the first seepage determination time window corresponding to each seepage area in the region list of the shell seepage spatiotemporal distribution map, in seconds. The seepage diffusion time difference is calculated as follows: for the i-th and (i+1)-th seepage areas in the seepage diffusion sequence, the difference between the end time of the time window when the (i+1)-th seepage area is marked as a seepage area and the end time of the time window when the i-th seepage area is marked as a seepage area is used as the seepage diffusion time difference. Divide the time difference of water diffusion by the Euclidean distance on the shell surface between the i-th and (i+1)-th water diffusion regions in the water diffusion sequence. The Euclidean distance on the shell surface has been calculated and stored during the generation of the water diffusion sequence to obtain the water diffusion rate. The unit of the water diffusion rate is millimeters per second, which represents how quickly water diffuses from one water diffusion region to an adjacent water diffusion region on the shell surface.

[0057] In practical implementation, after obtaining the seepage diffusion rate, the spray parameters of the simulated rain environment are also corrected based on the seepage diffusion rate. The correction of the spray parameters is achieved by obtaining the preset spray water pressure, preset spray flow rate, and preset spray water temperature values ​​for the current simulated rain environment. The preset spray water pressure is the setpoint water pressure at the output of the spray pump in the simulated rain environment, in megapascals; the preset spray flow rate is the setpoint flow rate of the flow regulating valve in the simulated rain environment, in liters per minute; and the preset spray water temperature is the setpoint temperature of the water temperature controller in the simulated rain environment, in degrees Celsius. The seepage diffusion rate is compared with a pre-stored standard diffusion rate threshold, which is pre-stored in the register of the programmable logic controller. The specific value of the standard diffusion rate threshold is 0.5 millimeters per second. The standard diffusion rate threshold is determined based on statistical data of water diffusion rates of various sizes of car audio housings under standard rain test conditions. The standard rain test conditions refer to a rain environment with a spray water pressure of 0.15 MPa, a spray flow rate of 8 liters per minute, and a spray water temperature of 25 degrees Celsius.

[0058] If the seepage diffusion rate is greater than the standard diffusion rate threshold, the preset spray water pressure value is reduced by the first correction factor. The corrected spray water pressure value is equal to the preset spray water pressure value multiplied by the first correction factor, which is the ratio of the seepage diffusion rate to the standard diffusion rate threshold. At the same time, the preset spray flow rate value is reduced by the second correction factor. The corrected spray flow rate value is equal to the preset spray flow rate value multiplied by the second correction factor, which is the standard diffusion rate threshold divided by the seepage diffusion rate. If the seepage diffusion rate is less than the standard diffusion rate threshold, the preset spray water pressure value is increased according to the third correction coefficient. The corrected spray water pressure value is equal to the preset spray water pressure value multiplied by the third correction coefficient, which is the ratio of the standard diffusion rate threshold to the seepage diffusion rate. At the same time, the preset spray water temperature value is increased according to the fourth correction coefficient. The corrected spray water temperature value is equal to the preset spray water temperature value multiplied by the fourth correction coefficient, which is the ratio of the seepage diffusion rate to the standard diffusion rate threshold multiplied by a coefficient of 0.5. The coefficient of 0.5 is determined based on the fact that the sensitivity of the spray water temperature to the seepage diffusion rate is about half that of the sensitivity of the spray water pressure to the seepage diffusion rate. Therefore, the water temperature correction range is taken as half of the water pressure correction range.

[0059] The corrected spray water pressure, spray flow rate, and spray water temperature values ​​are sent to the data register of the programmable logic controller (PLC) simulating the rain environment via a general industrial communication protocol. After receiving the corrected parameter values, the PLC writes the corrected spray water pressure value to the pressure setpoint register of the spray pump frequency converter, the corrected spray flow rate value to the opening setpoint register of the flow regulating valve, and the corrected spray water temperature value to the temperature setpoint register of the water temperature controller. In the next control cycle, the PLC adjusts the spray pump speed, the opening of the flow regulating valve, and the heating power of the water temperature controller according to the updated setpoints, thus completing the update of the spray parameters.

[0060] The foregoing has provided a detailed description of one embodiment of the present invention, but this description is merely a preferred embodiment and should not be construed as limiting the scope of the invention. All equivalent variations and modifications made within the scope of the claims of this invention should still fall within the patent coverage of this invention.

Claims

1. A method for testing the waterproof performance of a vehicle audio enclosure based on intelligent sensing technology, characterized in that, Includes the following steps: Obtain the initial vibration characteristic spectrum of the car audio housing in a simulated rain environment; The initial vibration characteristic spectrum is input into a pre-trained shell-acoustic field coupling model to generate the estimated seepage probability distribution of each region of the shell, specifically including: The pre-trained shell-sound field coupling model adopts a graph neural network architecture, and the training samples include vibration characteristic spectra and measured seepage probability distribution under different shell damage levels. The initial vibration characteristic spectrum is discretized into a frequency-amplitude vector sequence and input into the first graph convolutional layer of the shell-acoustic field coupling model. The graph structure nodes of the first graph convolutional layer correspond to each element of the shell finite element mesh. After processing by the first graph convolutional layer, the unit displacement field features are output. The unit displacement field features are then input into the second graph convolutional layer of the shell-sound field coupling model. The graph structure edge weights of the second graph convolutional layer are determined by the sound pressure transfer function between adjacent shell units. The output of the second convolutional layer is input to the fully connected mapping network of the shell-acoustic field coupling model. The fully connected mapping network outputs a probability vector with the same number of finite element mesh elements as the shell. Each element in the probability vector represents the estimated water seepage probability of the corresponding element's region. Based on the geometric coordinates of the shell finite element mesh, the probability vector is remapped onto the shell surface to obtain the estimated seepage probability distribution of each region of the shell; Based on the estimated water seepage probability distribution, flexible piezoelectric sensor arrays are deployed at the top several locations with the highest probability on the shell surface. The flexible piezoelectric sensor array is activated to acquire the real-time resonant impedance spectrum of each sensor, and the acoustic emission signal of the inner wall of the housing is acquired simultaneously. The real-time resonant impedance spectrum is aligned with the acoustic emission signal by time windows, and the impedance drop slope and the cumulative acoustic emission energy value are calculated in each time window. If the impedance drop slope exceeds the first dynamic threshold and the acoustic emission energy accumulation value exceeds the second dynamic threshold, it is determined that water seepage has occurred in the corresponding sensor area. Record the location and time of all seepage areas to generate a spatiotemporal distribution map of seepage in the shell.

2. The method for testing the waterproof performance of a vehicle audio enclosure based on intelligent sensing technology according to claim 1, characterized in that, The acquisition of the initial vibration characteristic spectrum of the vehicle audio enclosure in a simulated rain environment specifically includes: Install the car audio housing on a vibration table, and attach single-axis accelerometers to nine preset geometric center points on the surface of the housing. The vibration table is controlled to excite the shell in a sinusoidal sweep frequency pattern from low frequency to high frequency, with a sweep frequency range of 20 Hz to 20 kHz and a sweep frequency rate of one octave per minute. The time-domain vibration signals of nine accelerometers were simultaneously acquired during the frequency sweep process. A fast Fourier transform was performed on each time-domain vibration signal to obtain nine raw spectra. Calculate the average amplitude of the nine original spectra at the same frequency point, and arrange the average amplitude in frequency order to generate the initial vibration characteristic spectrum.

3. The method for testing the waterproof performance of a vehicle audio enclosure based on intelligent sensing technology according to claim 2, characterized in that, Based on the estimated water seepage probability distribution, a flexible piezoelectric sensor array is deployed at the highest probability locations on the shell surface, specifically including: Extract the probability values ​​of all shell surface units from the estimated water seepage probability distribution and sort them from high to low probability values; Select the top M shell surface units in the sort, where M is obtained by dividing the total shell area by the effective detection area of ​​a single flexible piezoelectric sensor and then rounding up. Obtain the geometric center coordinates of the first M shell surface units, and use each geometric center coordinate as a layout point; A flexible piezoelectric sensor is attached to each deployment point. All flexible piezoelectric sensors are connected in series via a flexible circuit board to form the flexible piezoelectric sensor array. The output of the flexible circuit board is then connected to an impedance analyzer.

4. The method for testing the waterproof performance of a vehicle audio enclosure based on intelligent sensing technology according to claim 3, characterized in that, The flexible piezoelectric sensor array is activated to acquire the real-time resonant impedance spectrum of each sensor, and the acoustic emission signal of the inner wall of the housing is acquired simultaneously, specifically including: A start command is sent to the impedance analyzer. The impedance analyzer scans each sensor in the flexible piezoelectric sensor array at a rate of fifty times per second. The frequency range of each scan is from 100 Hz to 100 kHz. The real part and imaginary part of the impedance of each sensor in the scanning frequency range are recorded to form the real resonant impedance spectrum. Simultaneously, a synchronous trigger pulse is sent to a dual-channel acoustic emission acquisition card installed at the center of the inner wall of the housing. The two channels of the dual-channel acoustic emission acquisition card are respectively connected to broadband acoustic emission sensors pasted on the left and right inner sides of the inner wall of the housing. The dual-channel acoustic emission acquisition card continuously acquires the output signals of two broadband acoustic emission sensors at a sampling rate of two million points per second, and transmits the two signals to the buffer area in real time as the acoustic emission signals.

5. The method for testing the waterproof performance of a vehicle audio enclosure based on intelligent sensing technology according to claim 4, characterized in that, Aligning the real-time resonant impedance spectrum with the acoustic emission signal within a time window, and calculating the impedance drop slope and cumulative acoustic emission energy within each time window, specifically including: With the rising edge of the synchronous trigger pulse as the zero moment, the real-time resonant impedance spectrum and the acoustic emission signal are arranged along the same time axis; Set a fixed-length time window of 0.5 seconds and a sliding step of 0.1 seconds, and divide the time window into multiple consecutive time windows starting from time zero. For each time window, the real part of the impedance of each sensor at 100 kHz is extracted from the real-time resonant impedance spectrum within that time window, and the real part of the impedance is arranged in time sequence. The real part of the impedance is then linearly fitted to the real part of the impedance sequence to obtain the impedance drop slope within that time window. For the same time window, the two signals of the acoustic emission signal are integrated squared to obtain the energy values ​​of the two signals. The energy values ​​of the two signals are added together and the logarithm to the base 10 is taken to obtain the cumulative acoustic emission energy value.

6. The method for testing the waterproof performance of a vehicle audio enclosure based on intelligent sensing technology according to claim 5, characterized in that, If the impedance drop slope exceeds a first dynamic threshold and the accumulated acoustic emission energy exceeds a second dynamic threshold, then it is determined that water seepage has occurred in the corresponding sensor area, specifically including: Calculate the average impedance drop slope of the five consecutive time windows preceding the current time window. and standard deviation The first dynamic threshold is set to the average value. Add three times the standard deviation ; Calculate the average of the cumulative acoustic emission energy values ​​of the five consecutive time windows preceding the current time window. and standard deviation The second dynamic threshold is set to the average value. Add three times the standard deviation ; The impedance drop slope of each sensor within the current time window is compared with the first dynamic threshold, and the cumulative acoustic emission energy value of the current time window is compared with the second dynamic threshold. If the impedance drop slope of any sensor exceeds the first dynamic threshold and the cumulative acoustic emission energy exceeds the second dynamic threshold, then the location of the deployment point corresponding to the sensor is marked as a seepage area, and the end time of the current time window is recorded as the time when seepage occurs in that area.

7. The method for testing the waterproof performance of a vehicle audio enclosure based on intelligent sensing technology according to claim 6, characterized in that, The process of recording the location and time of all seepage areas to generate a spatiotemporal distribution map of shell seepage specifically includes: Create a spatiotemporal distribution data structure for shell seepage. The spatiotemporal distribution data structure for shell seepage includes a region list and a time matrix. The region list is used to store the coordinates of all deployment points marked as seepage regions. The number of rows in the time matrix is ​​equal to the length of the region list, and the number of columns is equal to the total number of time windows. For each deployment point marked as a seepage area, the value 1 is continued to be filled into the row corresponding to the deployment point and the column position corresponding to the current time window index in the time matrix; If the same deployment point continues to meet the seepage judgment condition in subsequent time windows, the value 1 will be continuously filled in the subsequent column position of the corresponding row of the time matrix until the deployment point no longer meets the seepage judgment condition. The coordinates of the deployment points in the region list are associated with the time matrix row by row, and the output is the spatiotemporal distribution map of the shell seepage.

8. The method for testing the waterproof performance of a vehicle audio enclosure based on intelligent sensing technology according to claim 6, characterized in that, The first dynamic threshold and the second dynamic threshold are recalculated once for each time window during the detection process.

Images