Acoustic array based industrial equipment abnormal sound field measurement method and system

Abstract

The present application relates to the technical field of sound wave measurement, and discloses an industrial equipment abnormal sound field measurement method and system based on a sound array, the method comprising: synchronously collecting multi-channel time-domain sound pressure based on a microphone array, obtaining a time-frequency complex sound pressure vector through short-time Fourier transform; constructing a cross-spectrum matrix and extracting a dominant sound propagation vector through characteristic decomposition; calculating a scanning grid point directional vector according to array geometry, and screening an abnormal time-frequency point set using the median of the spatial correlation coefficient between the dominant sound propagation vector and the directional vector; based on the abnormal complex sound pressure vector and the sound field transfer response, inverting equivalent source intensity, and then reconstructing sound pressure and particle velocity through the sound propagation coefficient from the equivalent source to the observation surface to obtain the abnormal sound field distribution; the present application can improve the accuracy of industrial equipment abnormal sound field measurement based on a sound array.

Description

Technical Field

[0001] This invention relates to the field of acoustic wave measurement technology, and in particular to a method and system for measuring abnormal sound fields of industrial equipment based on acoustic arrays. Background Technology

[0002] In existing technologies for measuring abnormal noise in industrial equipment, single-point sound pressure meters or handheld sound level meters can only obtain the sound pressure level value at a single point on the equipment, failing to reconstruct the spatial distribution of the entire sound field. This makes it difficult to determine the specific location of the abnormal noise source. Furthermore, single-point measurements are easily affected by ambient background noise and reflected sound, and the measurement results cannot reflect the true spatial propagation characteristics of the abnormal noise. While acoustic array imaging methods based on delay-summing beamforming can generate sound source distribution cloud maps, their spatial resolution is limited by the array aperture and signal frequency. In the low-frequency band or when the number of microphones is insufficient, they cannot distinguish multiple closely spaced abnormal noise sources, and sidelobe artifacts can mask the true location of the abnormal noise, leading to misjudgments.

[0003] Traditional cross-spectral matrix beamforming methods do not adaptively select time-frequency points when calculating the steering vector. All time-frequency points participate in sound source localization, resulting in steady-state background noise and transient non-abnormal pulses being mixed into the final result, reducing the signal-to-noise ratio for abnormal sound identification. Furthermore, these methods directly perform conventional beamforming on the cross-spectral matrix, failing to separate the dominant sound propagation direction from the scattered sound field. This makes it difficult to accurately extract the target abnormal sound components in industrial equipment with multiple sound sources emitting sound simultaneously. Near-field acoustic holography based on the equivalent source method requires pre-assuming the sound source distribution and manually setting regularization parameters. Inappropriate parameter selection can cause severe oscillations in the inversion results. The equivalent source intensity calculation process is extremely sensitive to array measurement noise; even a small voltage acquisition deviation can lead to orders of magnitude errors in the reconstructed sound pressure amplitude, making it impossible to stably obtain a reliable abnormal sound field distribution. Summary of the Invention

[0004] This invention provides a method and system for measuring the sound field of abnormal noises in industrial equipment based on an acoustic array. Its main purpose is to solve the problem of low accuracy in measuring the sound field of abnormal noises in industrial equipment based on an acoustic array.

[0005] To achieve the above objectives, the present invention provides a method for measuring the abnormal sound field of industrial equipment based on an acoustic array, comprising:

[0006] S1. Based on the target process, a microphone array synchronously collects the sound waves generated by industrial equipment to obtain the multi-channel time-domain sound pressure of the target process;

[0007] S2. Perform a short-time Fourier transform on the multi-channel time-domain sound pressure to obtain the time-frequency complex sound pressure vector of the target process;

[0008] S3. Construct the cross-spectral matrix of the time-frequency points in the target process based on the time-frequency complex sound pressure vector, and perform eigenvalue decomposition on the cross-spectral matrix to obtain the dominant sound propagation vector of the target process;

[0009] S4. Calculate the guiding vector of the scanning grid points in the target process according to the geometric position of the microphone array, and filter the abnormal sound time-frequency point set of the target process based on the median of the spatial correlation coefficient between the dominant sound propagation vector and the guiding vector.

[0010] S5. Based on the complex sound pressure vector of the abnormal noise at the time-frequency point and the sound field transmission response from the equivalent source to the array in the target process, the equivalent source intensity of the target process is inverted.

[0011] S6. Based on the equivalent source intensity and the sound propagation coefficient from the equivalent source to the observation surface corresponding to the array, reconstruct the sound pressure and particle velocity on the observation surface to obtain the abnormal sound field distribution of the target process.

[0012] In a preferred embodiment, the microphone array based on the target process synchronously acquires sound waves generated by industrial equipment to obtain the multi-channel time-domain sound pressure of the target process, including:

[0013] The microphone array is triggered by a synchronous clock signal to collect the equipment sound waves of the target process, thereby obtaining the multi-channel analog voltage signal of the target process;

[0014] The multi-channel analog voltage signal is converted from analog to digital to obtain the time-domain sound pressure sampling sequence of the target process;

[0015] The time-domain sound pressure sampling sequence is combined according to the microphone channel number to obtain the multi-channel time-domain sound pressure of the target process.

[0016] In a preferred embodiment, performing a short-time Fourier transform on the multi-channel time-domain sound pressure to obtain the time-frequency complex sound pressure vector of the target process includes:

[0017] Cosine square windows are applied to the channel time-domain sound pressures of the multi-channel time-domain sound pressures to perform frame division, thereby obtaining the windowed time-domain sound pressure sequence of the target process;

[0018] The windowed time-domain sound pressure sequence is subjected to a discrete Fourier transform to obtain the complex sound pressure spectrum of the target process;

[0019] Arrange the complex sound pressure spectrum in channel order to obtain the time-frequency complex sound pressure vector of the target process.

[0020] In a preferred embodiment, the step of constructing the cross-spectral matrix of the time-frequency points in the target process based on the time-frequency complex sound pressure vector, and performing eigenvalue decomposition on the cross-spectral matrix to obtain the dominant acoustic propagation vector of the target process, includes:

[0021] The cross-spectral matrix of the target process is obtained by performing an outer product operation between the time-frequency complex sound pressure vector and the corresponding conjugate transpose row vector;

[0022] The cross-spectral matrix is ​​decomposed into eigenvalues ​​to obtain the feature vector set of the target process, and the feature vector corresponding to the largest eigenvalue in the feature vector set is taken as the dominant acoustic propagation vector of the target process.

[0023] In a preferred embodiment, the step of calculating the steering vector of the scanned grid points in the target process based on the geometric position of the microphone array, and filtering the set of abnormal reverberation time-frequency points in the target process based on the median spatial correlation coefficient between the dominant acoustic propagation vector and the steering vector, includes:

[0024] Using the three-dimensional coordinates of the microphones in the microphone array in the spatial coordinate system as the reference origin, the three-dimensional spatial region around the industrial equipment in the target process is scanned to obtain the scanning grid points of the target process.

[0025] The steering vector of the target process is calculated based on the spatial distance from the scanning grid points to the microphone and the corresponding angular frequency;

[0026] The time-frequency spatial correlation of the target process is calculated based on the dominant acoustic propagation vector and the steering vector;

[0027] Using the median of the maximum spatial correlation among the time-frequency points as the screening criterion, time-frequency points with a maximum spatial correlation greater than the median are retained to obtain the abnormal time-frequency point set of the target process.

[0028] In a preferred embodiment, calculating the steering vector of the target process based on the spatial distance from the scanning grid points to the microphone and the corresponding angular frequency includes:

[0029] The ratio of the spatial distance from the scanning grid point to the microphone to the speed of sound is used as the sound wave propagation time value of the target process;

[0030] The product of the sound wave propagation time value and the corresponding angular frequency is used as the phase delay value of the target process;

[0031] The cosine of the phase delay value is used as the real part and the sine of the phase delay value is used as the imaginary part to construct the complex value of the phase delay of the target process, and the complex value of the phase delay is arranged in channel order to form the guide vector of the target process.

[0032] In a preferred embodiment, the formula for calculating the spatial correlation of time and frequency points includes:

[0033]

[0034] in, The spatial correlation of the time-frequency points, This represents the total number of microphone channels. This is the sequence index of the microphone channel. The first of the dominant sound propagation vectors One element, The first guide vector One element, As a regularization factor, The dominant sound propagation vector, The guide vector is described above.

[0035] In a preferred embodiment, the step of retrieving the equivalent source intensity of the target process based on the complex sound pressure vector of the abnormal noise at the concentrated abnormal noise time-frequency points and the sound field transfer response from the equivalent source to the array in the target process includes:

[0036] The inverse acoustic field response matrix of the target process is obtained by performing a conjugate transpose on the equivalent source-to-array acoustic field transfer response matrix.

[0037] The sound field inverse response matrix is ​​regularized to obtain the stable transfer matrix of the target process, and the stable transfer matrix is ​​inverted to obtain the inversion matrix of the target process.

[0038] Multiplying the inversion matrix by the abnormal sound pressure vector yields the equivalent source intensity of the target process.

[0039] In a preferred embodiment, the step of reconstructing the sound pressure and particle velocity on the observation surface based on the equivalent source intensity and the sound propagation coefficient from the equivalent source to the observation surface corresponding to the array, to obtain the abnormal sound field distribution of the target process, includes:

[0040] An observation surface is defined within the spatial region surrounding the industrial equipment during the target process, and observation grid points for the target process are arranged on the observation surface.

[0041] The product of the equivalent source intensity and the sound propagation coefficient of the corresponding observation grid point is taken as the equivalent source sound pressure component. Then, the equivalent source sound pressure component is summed to obtain the complex sound pressure value of the target process.

[0042] The real part of the complex sound pressure value is taken as the sound pressure amplitude, and the sound pressure amplitude is combined to obtain the sound pressure amplitude distribution of the target process;

[0043] The ratio of the difference in sound pressure amplitude between adjacent observation grid points in the target process to the spatial distance between the adjacent observation grid points is taken as the sound pressure spatial gradient of the target process, and the ratio of the sound pressure spatial gradient to the air inertial acoustic impedance in the target process is taken as the particle velocity component of the target process.

[0044] The complex sound pressure value and the particle velocity component are cross-spectral processed to obtain the sound intensity vector of the target process, and the sound pressure amplitude distribution and the sound intensity vector are combined to form the abnormal sound field distribution of the target process.

[0045] To address the aforementioned problems, the present invention also provides an industrial equipment abnormal noise sound field measurement system based on an acoustic array, the system comprising:

[0046] A multi-channel time-domain sound pressure module synchronously acquires sound waves generated by industrial equipment based on a microphone array of the target process to obtain the multi-channel time-domain sound pressure of the target process;

[0047] The time-frequency complex sound pressure vector module performs a short-time Fourier transform on the multi-channel time-domain sound pressure to obtain the time-frequency complex sound pressure vector of the target process;

[0048] The dominant sound propagation vector module constructs a cross-spectral matrix of time-frequency points in the target process based on the time-frequency complex sound pressure vector, and performs eigenvalue decomposition on the cross-spectral matrix to obtain the dominant sound propagation vector of the target process;

[0049] The abnormal noise time-frequency point set module calculates the guide vector of the scanning grid points in the target process based on the geometric position of the microphone array, and filters the abnormal noise time-frequency point set of the target process based on the median of the spatial correlation coefficient between the dominant sound propagation vector and the guide vector.

[0050] The equivalent source strength module, based on the complex sound pressure vector of the abnormal noise at the time-frequency point and the sound field transmission response from the equivalent source to the array in the target process, inversely calculates the equivalent source strength of the target process.

[0051] The abnormal noise sound field distribution module reconstructs the sound pressure and particle velocity on the observation surface based on the equivalent source intensity and the sound propagation coefficient from the equivalent source to the observation surface corresponding to the array, thereby obtaining the abnormal noise sound field distribution of the target process.

[0052] Compared with the prior art, the present invention has the following beneficial effects:

[0053] 1. By triggering a microphone array with a synchronous clock to collect multi-channel time-domain sound pressure and performing short-time Fourier transform, a time-frequency complex sound pressure vector and cross-spectral matrix are constructed. Then, eigenvalue decomposition is performed to extract the dominant sound propagation vector corresponding to the largest eigenvalue. The spatial correlation of time-frequency points is calculated by combining the steering vector and the median is used as the threshold to filter the set of abnormal sound time-frequency points. This method accurately preserves the time-frequency points where the abnormal sound component is dominant, eliminates the interference of background noise and non-abnormal sound pulses, and significantly improves the signal-to-noise ratio and anti-environmental noise capability of abnormal sound time-frequency point identification.

[0054] 2. Based on the complex sound pressure vector of the time-frequency point set of the abnormal noise and the sound field transfer response from the equivalent source to the array, the equivalent source intensity is inverted. Then, the sound pressure amplitude distribution and particle velocity components on the observation surface are reconstructed by the equivalent source intensity and the sound propagation coefficient of the observation surface. The two are further cross-spectral processed to obtain the sound intensity vector. Finally, the sound pressure amplitude distribution and the sound intensity vector are combined to form the abnormal noise sound field distribution. This method provides the strength distribution of sound pressure energy and the direction information of sound energy propagation, and fully describes the spatial sound field characteristics of the abnormal noise. Moreover, the equivalent source inversion process adopts a stable transfer matrix with regularization, which ensures the numerical stability of the inversion results and avoids drastic fluctuations in sound pressure caused by small measurement errors. Attached Figure Description

[0055] Figure 1 This is a flowchart illustrating a method for measuring the abnormal noise sound field of industrial equipment based on an acoustic array, according to an embodiment of the present invention.

[0056] Figure 2 A functional block diagram of an industrial equipment abnormal noise sound field measurement system based on an acoustic array provided in an embodiment of the present invention;

[0057] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0058] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0059] This application provides a method for measuring the abnormal noise sound field of industrial equipment based on an acoustic array. The execution entity of this method includes, but is not limited to, at least one of the following electronic devices that can be configured to execute the method provided in this application: a server, a terminal, etc. In other words, the method for measuring the abnormal noise sound field of industrial equipment based on an acoustic array can be executed by software or hardware installed on a terminal device or a server device. The server includes, but is not limited to, a single server, a server cluster, a cloud server, or a cloud server cluster. The server can be an independent server or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks (CDNs), and big data and artificial intelligence platforms.

[0060] Reference Figure 1 The diagram shown is a flowchart illustrating a method for measuring the abnormal noise sound field of industrial equipment based on an acoustic array, according to an embodiment of the present invention. In this embodiment, the method for measuring the abnormal noise sound field of industrial equipment based on an acoustic array includes:

[0061] In this embodiment of the invention, when the microphone array based on the target process synchronously acquires the sound waves generated by industrial equipment to obtain the multi-channel time-domain sound pressure of the target process, it is specifically used for:

[0062] The microphone array is triggered by a synchronous clock signal to collect the equipment sound waves of the target process, thereby obtaining the multi-channel analog voltage signal of the target process;

[0063] The multi-channel analog voltage signal is converted from analog to digital to obtain the time-domain sound pressure sampling sequence of the target process;

[0064] The time-domain sound pressure sampling sequence is combined according to the microphone channel number to obtain the multi-channel time-domain sound pressure of the target process.

[0065] Specifically, when the microphone array is triggered by a synchronous clock signal to collect the sound waves of the target process equipment, the synchronous clock signal generator generates a uniform electrical pulse signal, which is simultaneously distributed to the preamplifier trigger port of each microphone in the microphone array through a coaxial cable.

[0066] Specifically, when performing analog-to-digital conversion on multi-channel analog voltage signals, the analog voltage signal of each channel is first input to the corresponding analog-to-digital converter input port, and the analog-to-digital converter internally uses a successive approximation method to quantize the analog voltage signal.

[0067] Specifically, when combining the time-domain sound pressure sampling sequences according to the microphone channel number, firstly, the time-domain sound pressure sampling sequence corresponding to each microphone channel is read. Each sequence contains binary numbers arranged in chronological order. Then, using the number of the first microphone channel as an index, the time-domain sound pressure sampling sequence of that channel is stored in the first column of a matrix.

[0068] Furthermore, each microphone begins to convert the diaphragm vibration caused by the sound wave into a continuously changing analog voltage the instant it receives the rising edge of the electrical pulse signal. All microphones simultaneously begin to collect data for a preset collection time, ultimately obtaining a multi-channel analog voltage signal for the target process.

[0069] Furthermore, the comparator inside the analog-to-digital converter compares the input voltage bit by bit with an internal reference voltage generated by the digital-to-analog converter, determining the binary value of each bit sequentially from the most significant bit to the least significant bit, thereby converting the continuously changing analog voltage amplitude into discrete digital code. At the same time, the internal clock of the analog-to-digital converter triggers the sample-and-hold circuit at a fixed sampling frequency to lock the voltage value at each sampling moment. After quantization and encoding, a string of binary digital sequences is output, which is the time-domain sound pressure sampling sequence. The above conversion process is performed in parallel on the analog voltage signals of all channels to obtain the time-domain sound pressure sampling sequence of the target process.

[0070] Furthermore, using the second microphone channel number as an index, the time-domain sound pressure sampling sequence of that channel is stored in the second column of the same matrix. This process is repeated until the time-domain sound pressure sampling sequences of all channels are sequentially filled into the corresponding columns of the matrix in ascending order of channel number. Each row of this matrix represents the sound pressure value of all microphone channels at the same sampling time, and each column represents the sound pressure value of the same microphone channel at all sampling times. This matrix is ​​the multi-channel time-domain sound pressure of the target process.

[0071] In summary, by triggering the microphone array with a synchronous clock signal to collect the device sound waves of the target process, all microphones start converting sound wave signals at the same time, eliminating the time delay error caused by inconsistent start times between channels, ensuring strict alignment of the multi-channel analog voltage signals on the time axis, and providing a synchronous reference for subsequent accurate calculation of the phase difference of sound waves arriving at different microphones.

[0072] In summary, the analog-to-digital conversion of the multi-channel analog voltage signal quantizes the continuous analog voltage into a discrete binary digital sequence, transforming the original acoustic signal into a digital form that can be directly stored and processed by a computer. At the same time, the sampling process preserves the complete information of the original signal on the time axis, providing accurate numerical input for subsequent short-time Fourier transform and cross-spectral matrix construction.

[0073] In summary, by combining the time-domain sound pressure sampling sequences according to the microphone channel number, the scattered channel data are integrated into a matrix structure according to a unified channel order. This ensures that the sound pressure values ​​of all microphones at the same sampling time are arranged in the same row, and the sound pressure values ​​of the same microphone at all sampling times are arranged in the same column. This regularized data organization method facilitates the subsequent algorithm to quickly index the data by channel or by time sequence, thereby improving the efficiency of multi-channel sound field processing.

[0074] In this embodiment of the invention, when performing a short-time Fourier transform on the multi-channel time-domain sound pressure to obtain the time-frequency complex sound pressure vector of the target process, the specific steps are as follows:

[0075] Cosine square windows are applied to the channel time-domain sound pressures of the multi-channel time-domain sound pressures to perform frame division, thereby obtaining the windowed time-domain sound pressure sequence of the target process;

[0076] The windowed time-domain sound pressure sequence is subjected to a discrete Fourier transform to obtain the complex sound pressure spectrum of the target process;

[0077] Arrange the complex sound pressure spectrum in channel order to obtain the time-frequency complex sound pressure vector of the target process.

[0078] Specifically, when applying a cosine square window to the channel time-domain sound pressure of the multi-channel time-domain sound pressure for framing, each column of the multi-channel time-domain sound pressure matrix is ​​taken as a channel time-domain sound pressure sequence. The sequence is segmented into non-overlapping segments according to a fixed frame length, and each segment is called a frame. For the sound pressure sampling points in each frame, the window function value of the cosine square window function at the current sampling point position is calculated sequentially.

[0079] Specifically, when performing a Discrete Fourier Transform on the windowed time-domain sound pressure sequence, the sequence is organized into a two-dimensional structure according to the frame index and the intra-frame sampling point index. The Discrete Fourier Transform is performed independently on the windowed time-domain sound pressure sequence of each frame. The specific execution method of the Discrete Fourier Transform is as follows: for a sequence with a fixed length in the current frame, the complex value corresponding to each discrete frequency point is calculated in sequence. During the calculation, the windowed sound pressure value of each sampling point in the frame is first summed with a rotation factor.

[0080] Specifically, when arranging the complex sound pressure spectrum by channel order, it is first determined that the complex sound pressure spectrum is a three-dimensional structure. The three dimensions are the microphone channel index, the frame index, and the frequency point index. When arranging, the frame index and the frequency point index are fixed, and the complex sound pressure values ​​of all microphone channels at the frequency point of the frame are extracted and arranged into a column vector in ascending order of microphone channel number.

[0081] Furthermore, the cosine squared window function is constructed as follows: within the window length, from the starting point to the ending point, the window function value rises from zero to one and then falls back to zero, its shape being the square of the cosine function. In specific implementation, first determine the relative position ratio of each sampling point in the current frame within the window length, then calculate the angle value corresponding to this ratio, and then calculate the square of the cosine value of this angle as a weighting coefficient. Multiply the original sound pressure value of each sampling point in the current frame by the weighting coefficient of the corresponding position to obtain the windowed sound pressure value of the frame. Perform the above operation on all frames in sequence, and then splice the windowed sound pressure values ​​of all frames in time order to form a new sequence. This sequence is the windowed time-domain sound pressure sequence of the target process.

[0082] Furthermore, the rotation factor is determined by the position index of the sampling point within the frame and the current frequency index. The real part of the rotation factor is a cosine value, and the imaginary part is a negative sine value. The angle is equal to twice pi multiplied by the position index multiplied by the frequency index and then divided by the frame length. The product results of all sampling points are summed to obtain the complex value corresponding to the frequency point. The above process is repeated for all frequency points to obtain a set of complex results for the frame. The complex results are arranged in the order of the frequency index to obtain the complex sound pressure spectrum of the frame. The same operation is performed on all frames to obtain the complex sound pressure spectrum of the target process.

[0083] Furthermore, the first element of this column vector is the complex sound pressure value of the microphone with the smallest channel number at this frequency point in the frame, the second element is the complex sound pressure value of the microphone with the second smallest channel number, and so on until the last channel. This column vector is the time-frequency complex sound pressure vector at this frequency point in the frame. The above extraction and arrangement operations are repeated for all frame indices and all frequency point indices to obtain the set of time-frequency complex sound pressure vectors of the target process. Each time-frequency complex sound pressure vector corresponds to a unique frame time and a frequency point.

[0084] In summary, applying a cosine square window to each channel's time-domain sound pressure for framing can suppress the discontinuity of each time-domain sound pressure sequence at the frame boundary, avoid spectral leakage caused by signal abrupt changes during subsequent Fourier transform, and thus ensure that the frequency components of the windowed time-domain sound pressure sequence truly reflect the original acoustic wave characteristics of the industrial equipment.

[0085] In summary, performing a discrete Fourier transform on the windowed time-domain sound pressure sequence converts the time-domain sound pressure signal to the frequency domain, enabling the complex sound pressure spectrum at each frequency point to simultaneously represent the amplitude and phase information of that frequency component. This provides fundamental data for subsequent construction of time-frequency complex sound pressure vectors and analysis of the frequency domain characteristics of abnormal sounds.

[0086] In summary, by arranging the complex sound pressure spectrum in channel order and organizing the complex sound pressure values ​​of all microphone channels at the same time and frequency point into a column vector, the spatial information of each time and frequency point can be centrally expressed, which facilitates subsequent cross-spectral matrix construction and feature decomposition of the vector, thereby extracting the dominant sound propagation direction and identifying the time and frequency points of abnormal sounds.

[0087] In this embodiment of the invention, the step of constructing the cross-spectral matrix of the time-frequency points in the target process based on the time-frequency complex sound pressure vector, and performing eigenvalue decomposition on the cross-spectral matrix to obtain the dominant acoustic propagation vector of the target process, is specifically used for:

[0088] The cross-spectral matrix of the target process is obtained by performing an outer product operation between the time-frequency complex sound pressure vector and the corresponding conjugate transpose row vector;

[0089] The cross-spectral matrix is ​​decomposed into eigenvalues ​​to obtain the feature vector set of the target process, and the feature vector corresponding to the largest eigenvalue in the feature vector set is taken as the dominant acoustic propagation vector of the target process.

[0090] Specifically, when performing the outer product operation between the time-frequency complex sound pressure vector and the corresponding conjugate transpose row vector, the time-frequency complex sound pressure vector corresponding to a certain time-frequency point in the target process is first extracted. This vector is a column vector, and each element of it is a complex number. Then, the conjugate transpose operation is performed on this column vector.

[0091] Specifically, when performing eigenvalue decomposition on the cross-spectral matrix, a numerical iterative method is used to solve it step by step. First, an identity matrix with the same dimension as the cross-spectral matrix is ​​constructed as the initial eigenvector matrix, and a diagonal matrix is ​​set as the initial eigenvalue matrix. Then, the off-diagonal elements of the cross-spectral matrix are eliminated by repeatedly applying orthogonal transformations. Each orthogonal transformation selects the row and column containing the off-diagonal element with the largest absolute value in the cross-spectral matrix to construct a rotation matrix.

[0092] Furthermore, first, take the conjugate of each complex element in the column vector to obtain a new complex number. Then, change the orientation of the column vector from vertical to horizontal to form a row vector, which is the corresponding conjugate transpose row vector. Then, perform an outer product operation between the original column vector and the row vector. Specifically, multiply each complex element in the column vector by all elements in the row vector. The column vector has a first number of elements, and the row vector has the same number of elements. The result of the multiplication forms a square matrix. In the square matrix, the row number corresponds to the index of the element in the column vector, and the column number corresponds to the index of the element in the row vector. Each element in the square matrix is ​​the product of two complex numbers. This square matrix is ​​the cross-spectral matrix of the time-frequency point in the target process.

[0093] Furthermore, the rotation matrix has cosine and positive / negative sine values ​​only at the selected row and column positions, and the other positions are the same as the identity matrix. The rotation matrix is ​​multiplied left by the current cross-spectrum matrix and then right by the transpose of the rotation matrix, so that the selected off-diagonal elements become zero. At the same time, the eigenvector matrix is ​​updated. The above orthogonal transformation is repeated until the absolute value of all off-diagonal elements is less than the preset minimum threshold. At this time, the cross-spectrum matrix is ​​diagonalized, and the elements on the diagonal are the eigenvalues. The product of the accumulated rotation matrices is the eigenvector set. Each column vector in the eigenvector set corresponds to an eigenvalue. The eigenvalue with the largest value is found, and the column eigenvector corresponding to that eigenvalue is extracted. This column vector is the dominant sound propagation vector of the target process.

[0094] In summary, the cross-spectral matrix is ​​obtained by performing an outer product operation between the time-frequency complex sound pressure vector and the corresponding conjugate transpose row vector. This allows the spatial covariance information at each time-frequency point to be centrally expressed in a square matrix. This square matrix simultaneously preserves the amplitude correlation and phase correlation between each microphone channel, providing a complete second-order statistical basis for the subsequent extraction of the dominant propagation direction of the sound field.

[0095] In summary, by performing eigenvalue decomposition on the cross-spectral matrix to obtain an eigenvector set, and taking the eigenvector corresponding to the eigenvalue with the largest value as the dominant sound propagation vector, the strongest sound propagation mode can be separated from the mixed multi-channel sound pressure data. This dominant sound propagation vector carries the main arrival direction information of the abnormal noise source relative to the microphone array, thus providing a reliable spatial guidance basis for subsequent screening of the time-frequency point set of abnormal noise and inversion of the equivalent source intensity.

[0096] In this embodiment of the invention, the step of calculating the steering vector of the scanning grid points in the target process based on the geometric position of the microphone array, and filtering the set of abnormal sound time-frequency points in the target process based on the median spatial correlation coefficient between the dominant sound propagation vector and the steering vector, is specifically used for:

[0097] Using the three-dimensional coordinates of the microphones in the microphone array in the spatial coordinate system as the reference origin, the three-dimensional spatial region around the industrial equipment in the target process is scanned to obtain the scanning grid points of the target process.

[0098] The steering vector of the target process is calculated based on the spatial distance from the scanning grid points to the microphone and the corresponding angular frequency;

[0099] The time-frequency spatial correlation of the target process is calculated based on the dominant acoustic propagation vector and the steering vector;

[0100] Using the median of the maximum spatial correlation among the time-frequency points as the screening criterion, time-frequency points with a maximum spatial correlation greater than the median are retained to obtain the abnormal time-frequency point set of the target process.

[0101] Specifically, when taking the three-dimensional coordinates of the microphones in the microphone array in the spatial coordinate system as the reference origin, a three-dimensional spatial coordinate system is first established, and the physical position of each microphone in the microphone array is represented by three coordinate values ​​under this coordinate system.

[0102] Specifically, when calculating the guide vector of the target process based on the spatial distance from the scanning grid point to the microphone and the corresponding angular frequency, for each scanning grid point, the spatial straight-line distance from that point to each microphone in the microphone array is calculated, and the time required for the sound wave to travel from the scanning grid point to the microphone is obtained by dividing the distance value by the speed of sound in the air.

[0103] Specifically, when calculating the spatial correlation of the time-frequency points of the target process based on the dominant sound propagation vector and the steering vector, the dominant sound propagation vector corresponding to a certain time-frequency point in the target process and the steering vector corresponding to the current frequency point of the current scanning grid point are first extracted. The lengths of these two vectors are equal to the total number of microphone channels. Then, the following calculation is performed channel by channel: For each microphone channel number, the element of the dominant sound propagation vector with that number is conjugate and multiplied by the element with the same number in the steering vector to obtain the product value of that channel. The product values ​​of all channels are summed to obtain a sum value.

[0104] Specifically, when using the median of the maximum spatial correlation among the time-frequency spatial correlations as the screening criterion, the maximum spatial correlation values ​​of each time-frequency point in the target process are first collected, and these values ​​are arranged into a sequence in ascending order.

[0105] Furthermore, using the coordinates of these microphones as the reference origin, a rectangular three-dimensional spatial region is delineated around the industrial equipment during the target process. The boundary of this region covers the entire spatial range where the industrial equipment may generate abnormal noises. Then, points are taken in three directions at fixed intervals within this region, with the distance between adjacent points in each direction being equal. The set of these points is the scanning grid point, and each scanning grid point corresponds to the three-dimensional coordinates of a spatial location.

[0106] Furthermore, for the current frequency point being processed, the angular frequency value of that frequency point is multiplied by the time value to obtain the phase delay value. Then, the cosine and sine values ​​of the phase delay value are calculated respectively. A complex number is constructed with the cosine value as the real part and the sine value as the imaginary part. This complex number represents the phase change of the sound wave as it propagates from the scanning grid point to the microphone. The above operation is performed on each microphone in the microphone array to obtain a complex number sequence. This sequence is arranged into a column vector according to the microphone channel number. This column vector is the steering vector of the scanning grid point at that frequency point.

[0107] Furthermore, the magnitudes of the dominant sound propagation vector and the steering vector are calculated separately. The two magnitudes are multiplied and then added to a small positive number to obtain the denominator. The absolute value of the sum obtained earlier is divided by the denominator to obtain the spatial correlation of the time frequency point at the scanning grid point. After traversing all scanning grid points, each time frequency point will obtain a spatial correlation value equal to the number of scanning grid points. The maximum value among them is taken as the maximum spatial correlation of the time frequency point. The above process is repeated for each time frequency point.

[0108] Furthermore, the value at the middle position of the sequence is then found. If the number of values ​​in the sequence is odd, the middle position is unique; if it is even, the average of the two middle values ​​is taken, which is the median. Then, the maximum spatial correlation of each time-frequency point is compared with the median, and those time-frequency points with a maximum spatial correlation greater than the median are retained. Each of the retained time-frequency points corresponds to an original time-frequency complex sound pressure vector. The set of all these retained time-frequency complex sound pressure vectors is the abnormal sound time-frequency point set of the target process.

[0109] In summary, by taking the three-dimensional coordinates of the microphones in the microphone array in the spatial coordinate system as the reference origin, the three-dimensional spatial region around the industrial equipment during the scanning process is used to obtain scanning grid points. The continuous spatial domain is discretized into a finite number of grid points, which transforms the problem of finding the location of abnormal noise sources into a problem of correlation judgment on known grid points, reducing computational complexity while ensuring the integrity of spatial coverage.

[0110] In summary, the steering vector is calculated based on the spatial distance from the scanning grid point to the microphone and the corresponding angular frequency, so that each scanning grid point obtains a complex vector of the theoretical sound propagation phase mode at each frequency point. This vector is used as a spatial matching filter to compare the similarity with the measured dominant sound propagation vector, thereby quantifying the possibility of abnormal noises occurring at different spatial locations.

[0111] In summary, the spatial correlation of time-frequency points is calculated based on the dominant sound propagation vector and the steering vector. The actual sound field propagation direction of each time-frequency point is numerically matched with the preset theoretical propagation direction of the scanning grid point to obtain a similarity index between zero and one. This index directly reflects whether the sound source at that time-frequency point is located near that scanning grid point.

[0112] In summary, by using the median of the maximum spatial correlation among time-frequency points as the screening criterion, time-frequency points with a maximum spatial correlation greater than the median are retained to obtain the set of abnormal noise time-frequency points. The median is used as an adaptive threshold to exclude time-frequency points with strong static background noise and random interference, and only abnormal noise components with sufficiently high spatial focus are retained, thereby improving the robustness and accuracy of abnormal noise time-frequency point screening.

[0113] In this embodiment of the invention, the step of calculating the steering vector of the target process based on the spatial distance from the scanning grid point to the microphone and the corresponding angular frequency is specifically used for:

[0114] The ratio of the spatial distance from the scanning grid point to the microphone to the speed of sound is used as the sound wave propagation time value of the target process;

[0115] The product of the sound wave propagation time value and the corresponding angular frequency is used as the phase delay value of the target process;

[0116] The cosine of the phase delay value is used as the real part and the sine of the phase delay value is used as the imaginary part to construct the complex value of the phase delay of the target process, and the complex value of the phase delay is arranged in channel order to form the guide vector of the target process.

[0117] Specifically, when the ratio of the spatial distance from the scanning grid point to the microphone to the speed of sound is used as the sound wave propagation time value of the target process, for each scanning grid point, the spatial straight-line distance from the scanning grid point to each microphone in the microphone array is first measured, that is, the Euclidean distance from the coordinates of the scanning grid point to the coordinates of the microphone.

[0118] Specifically, when using the product of the sound wave propagation time value and the corresponding angular frequency as the phase delay value of the target process, first take out the angular frequency value corresponding to the frequency point currently being processed. The angular frequency value is equal to twice pi multiplied by the frequency value of that frequency point. Then, multiply the angular frequency value by the sound wave propagation time value corresponding to each microphone.

[0119] Specifically, when constructing the complex value of phase delay for the target process by taking the cosine of the phase delay value as the real part and the sine of the phase delay value as the imaginary part, and arranging the complex value of phase delay in channel order as the guide vector of the target process, for each microphone, first calculate the cosine trigonometric function value of the phase delay value corresponding to the microphone, and take the cosine value as the real part of the complex number, then calculate the sine trigonometric function value of the same phase delay value, and take the sine value as the imaginary part of the complex number, and combine the real part and the imaginary part into a complex number, which is the complex value of phase delay corresponding to the microphone.

[0120] Furthermore, by dividing this distance value by the fixed speed at which sound waves travel in the air, the quotient is the time it takes for the sound wave to travel from the scanning grid point to the corresponding microphone. This time is called the sound wave propagation time value of the target process, and each microphone corresponds to an independent sound wave propagation time value.

[0121] Furthermore, for each microphone, the product of the angular frequency and the sound wave propagation time is the accumulated phase change when the sound wave propagates from the scanning grid point to the microphone. This change is called the phase delay value of the target process, and each microphone corresponds to an independent phase delay value.

[0122] Furthermore, the above operation is performed sequentially on all microphones in the microphone array to obtain a complex number equal to the number of microphone channels. These complex numbers are then arranged into a column vector in ascending order of microphone channel number. The first element of this column vector is the complex value of the phase delay of the microphone with the smallest channel number, the second element is the complex value of the phase delay of the microphone with the second smallest channel number, and so on. This column vector is the steering vector of the target process.

[0123] In summary, by using the ratio of the spatial distance from the scanning grid point to the microphone to the speed of sound as the sound wave propagation time value, the spatial geometric distance is converted into the time delay of sound wave propagation. This allows the arrival time difference between different microphones due to positional differences to be represented by a unified physical quantity, providing a time reference for subsequent phase delay calculations.

[0124] In summary, the product of the sound wave propagation time and the corresponding angular frequency is used as the phase delay value. Multiplying the time delay by the frequency yields a phase offset that varies linearly with the frequency. This allows the steering vector to accurately describe the phase differences of different frequency components during spatial propagation, thereby adapting to multiple frequency components that may be contained in the abnormal sound signal.

[0125] In summary, by constructing complex phase delay values ​​using the cosine of the phase delay as the real part and the sine as the imaginary part, and arranging them in channel order as steering vectors, and using complex exponential form to represent the phase delay of the sound wave, the steering vector contains both amplitude and phase information, and can be used for complex domain inner product operations with the measured complex sound pressure vector, thereby improving the numerical stability and physical consistency of spatial correlation calculation.

[0126] In this embodiment of the invention, the formula for calculating the spatial correlation of time-frequency points is specifically used for:

[0127]

[0128] in, The spatial correlation of the time-frequency points, This represents the total number of microphone channels. This is the sequence index of the microphone channel. The first of the dominant sound propagation vectors One element, The first guide vector One element, As a regularization factor, The dominant sound propagation vector, The guide vector is described above.

[0129] Specifically, the total number of microphone channels is directly determined by the number of physical channels in the microphone array, and the dominant sound propagation vector is the [missing information]. The element is a complex number extracted from the dominant acoustic propagation vector obtained after eigenvalue decomposition of the cross-spectral matrix, indexed by channel number. The nth element of the steering vector... The element is a complex number extracted from the guide vector calculated based on the spatial distance from the scanning grid point to the microphone and the corresponding angular frequency, indexed by the same channel number. The regularization factor is a small positive number calculated by regularizing the inverse response matrix of the sound field. The dominant sound propagation vector is the entire column vector, and the guide vector is the entire column vector. These parameters participate in the calculation in the formula in order.

[0130] Furthermore, this formula is used to calculate the spatial correlation of a certain time-frequency point at a certain scanning grid point during the target process. Its core meaning is to measure the similarity between the dominant sound propagation vector and the steering vector: the numerator of the formula first takes the conjugate of the elements of the dominant sound propagation vector in each channel and multiplies it with the corresponding elements of the steering vector and then sums the results of all channels. The denominator is multiplied by the magnitude of the dominant sound propagation vector and the magnitude of the steering vector and then a regularization factor is added to prevent the denominator from being zero. The larger the absolute value of the whole ratio, the more consistent the directions of the dominant sound propagation vector and the steering vector are, that is, the sound field propagation direction at that time-frequency point is highly consistent with the preset sound wave direction at that scanning grid point.

[0131] In general, when the dominant sound propagation vector and the steering vector are in the same direction, the absolute value of the numerator is equal to the product of the moduli in the denominator. At this time, the spatial correlation approaches its maximum value, indicating that the sound source at this time-frequency point is very likely located at this scanning grid point. When the dominant sound propagation vector and the steering vector are orthogonal or opposite in direction, the absolute value of the numerator is much smaller than the product of the moduli in the denominator, and the spatial correlation approaches zero, indicating that the sound field at this time-frequency point is unrelated to this scanning grid point. The existence of the regularization factor ensures that the denominator is always greater than the possible small errors in the absolute value of the numerator, ensuring that the value of the spatial correlation is stable and always varies between zero and one.

[0132] In this embodiment of the invention, the step of retrieving the equivalent source intensity of the target process based on the complex sound pressure vector of the abnormal noise at the concentrated abnormal noise time-frequency points and the sound field transfer response from the equivalent source to the array in the target process is specifically used for:

[0133] The inverse acoustic field response matrix of the target process is obtained by performing a conjugate transpose on the equivalent source-to-array acoustic field transfer response matrix.

[0134] The sound field inverse response matrix is ​​regularized to obtain the stable transfer matrix of the target process, and the stable transfer matrix is ​​inverted to obtain the inversion matrix of the target process.

[0135] Multiplying the inversion matrix by the abnormal sound pressure vector yields the equivalent source intensity of the target process.

[0136] Specifically, when performing the conjugate transpose of the equivalent source-to-array acoustic field transfer response matrix, the pre-calculated equivalent source-to-array acoustic field transfer response matrix is ​​first retrieved. The number of rows in this matrix is ​​equal to the total number of microphone channels, and the number of columns is equal to the total number of equivalent sources. Each element in the matrix is ​​a complex number, representing the acoustic field transfer characteristics from a certain equivalent source to a certain microphone. Then, the conjugate transpose operation is performed on this matrix.

[0137] Specifically, when performing regularization on the sound field inverse response matrix, the sound field inverse response matrix is ​​first multiplied by its own conjugate transpose to obtain a square matrix. Then, all elements on the diagonal of the square matrix are extracted. These diagonal elements are all positive real numbers. The average value of these diagonal elements is calculated, and then the average value is multiplied by a very small positive number to obtain the regularization factor. Finally, an identity matrix with the same number of columns as the sound field inverse response matrix is ​​constructed.

[0138] Specifically, when multiplying the inversion matrix with the abnormal sound pressure vector, the abnormal sound pressure vector corresponding to a certain time-frequency point is first extracted from the abnormal sound time-frequency point set. This vector is a column vector with the number of rows equal to the total number of microphone channels, and each element is a complex number. Then, the inversion matrix is ​​multiplied with this abnormal sound pressure vector.

[0139] Furthermore, the rows and columns of the matrix are first interchanged, so that the original rows become columns and the original columns become rows, resulting in a new matrix. Then, the conjugate of each complex element in the new matrix is ​​taken, that is, the imaginary part of each complex number is inverted while the real part remains unchanged. The matrix obtained after the row and column interchange and the conjugate of complex numbers is the acoustic field inverse response matrix of the target process.

[0140] Furthermore, multiply each diagonal element of the identity matrix by a regularization factor to obtain a regularized diagonal matrix. Then, add the product of the acoustic field inverse response matrix and its conjugate transpose to the regularized diagonal matrix to obtain a new square matrix. Finally, multiply the acoustic field inverse response matrix on the right by the inverse of this new square matrix. The resulting matrix is ​​the stable transfer matrix of the target process. Perform the inversion operation on this stable transfer matrix, i.e., solve for a matrix such that multiplying this matrix by the stable transfer matrix yields the identity matrix, to obtain the inversion matrix of the target process.

[0141] Furthermore, the number of rows in the inversion matrix equals the total number of equivalent sources, and the number of columns equals the total number of microphone channels. Each row of the inversion matrix is ​​multiplied by the abnormal sound pressure vector. That is, for a certain row of the inversion matrix, each element of the row is multiplied by the corresponding element in the abnormal sound pressure vector, and all products are added together to obtain a complex number. This complex number is the intensity value of the corresponding equivalent source. The above dot product operation is performed on each row of the inversion matrix to obtain a set of complex numbers equal to the total number of equivalent sources. These complex numbers are arranged into a column vector according to the equivalent source numbering order. This column vector is the equivalent source intensity of the target process.

[0142] In summary, the inverse sound field response matrix is ​​obtained by conjugate transpose of the sound field transfer response matrix from the equivalent source to the array. This transforms the forward sound propagation relationship into an inverse solution relationship, reversing the original mapping direction from the equivalent source to the microphone. This establishes the basic framework of algebraic equations for inferring the equivalent source intensity from the measured sound pressure of the array.

[0143] In summary, the sound field inverse response matrix is ​​regularized to obtain a stable transfer matrix, and the stable transfer matrix is ​​inverted to obtain an inversion matrix. By introducing a regularization factor into the inversion matrix, the phenomenon of small noise being drastically amplified when inverting the ill-conditioned matrix is ​​suppressed, thus ensuring the numerical stability of the inversion matrix. This allows for the acquisition of reliable equivalent source strength even in industrial environments with low signal-to-noise ratios.

[0144] In summary, by multiplying the inversion matrix with the abnormal sound pressure vector to obtain the equivalent source intensity, and directly applying the pre-calculated stable inversion matrix to the measured sound pressure data at the time-frequency points of the abnormal sound, an efficient linear mapping from the sound pressure received by the microphone array to the equivalent source intensity distribution is achieved. This avoids the computational burden of point-by-point iterative solution and provides accurate source intensity input for subsequent reconstruction of the abnormal sound field distribution on the observation surface.

[0145] In this embodiment of the invention, when reconstructing the sound pressure and particle velocity on the observation surface based on the equivalent source intensity and the sound propagation coefficient from the equivalent source to the observation surface corresponding to the array, and obtaining the abnormal sound field distribution of the target process, it is specifically used for:

[0146] An observation surface is defined within the spatial region surrounding the industrial equipment during the target process, and observation grid points for the target process are arranged on the observation surface.

[0147] The product of the equivalent source intensity and the sound propagation coefficient of the corresponding observation grid point is taken as the equivalent source sound pressure component. Then, the equivalent source sound pressure component is summed to obtain the complex sound pressure value of the target process.

[0148] The real part of the complex sound pressure value is taken as the sound pressure amplitude, and the sound pressure amplitude is combined to obtain the sound pressure amplitude distribution of the target process;

[0149] The ratio of the difference in sound pressure amplitude between adjacent observation grid points in the target process to the spatial distance between the adjacent observation grid points is taken as the sound pressure spatial gradient of the target process, and the ratio of the sound pressure spatial gradient to the air inertial acoustic impedance in the target process is taken as the particle velocity component of the target process.

[0150] The complex sound pressure value and the particle velocity component are cross-spectral processed to obtain the sound intensity vector of the target process, and the sound pressure amplitude distribution and the sound intensity vector are combined to form the abnormal sound field distribution of the target process.

[0151] Specifically, when defining the observation surface within the spatial area surrounding the industrial equipment during the target process, a cuboid spatial area is first delineated with the industrial equipment as the center. This area completely surrounds all locations where the equipment may produce abnormal noises. Then, a plane is selected within this area as the observation surface.

[0152] Specifically, when using the product of the equivalent source intensity and the sound propagation coefficient of the corresponding observation grid point as the equivalent source sound pressure component, first take out the equivalent source intensity vector that has been inverted. This vector contains the complex intensity value of each equivalent source. Then, for each observation grid point, calculate the sound propagation coefficient from each equivalent source to that observation grid point.

[0153] Specifically, when the real part of the complex sound pressure value is used as the sound pressure amplitude, for each observation grid point, the real part value of the complex sound pressure value at that point is taken out. The real part represents the magnitude of the instantaneous sound pressure at that point. The phase information represented by the imaginary part is ignored, and the real part value is used as the sound pressure amplitude of that observation grid point.

[0154] Specifically, when using the ratio of the difference in sound pressure amplitude between adjacent observation grid points in the target process to the spatial distance between the adjacent observation grid points as the sound pressure spatial gradient of the target process, firstly, each pair of adjacent observation grid points is found along the row and column directions on the observation surface, the spatial straight-line distance between these two observation grid points is calculated, and then the difference in sound pressure amplitude between these two observation grid points is calculated. The difference is divided by the spatial distance to obtain the rate of change in that direction. After performing the above operation on all adjacent point pairs, a set of rate of change values ​​is obtained, and this set of values ​​is the sound pressure spatial gradient.

[0155] Specifically, when performing cross-spectral processing on the complex sound pressure value and the particle velocity component, for each observation grid point, the complex sound pressure value of that point and the corresponding particle velocity component are taken out. The conjugate of the complex sound pressure value is multiplied by the particle velocity component to obtain a complex number. The real part of this complex number is the magnitude of the sound wave energy flux density at that point in the direction normal to the observation surface. This real part value is called the sound intensity value.

[0156] Furthermore, the observation surface is parallel to the plane where the microphone array is located and is situated between the device and the array. Then, a series of spatial points are uniformly arranged on the observation surface according to fixed row and column spacing. Each point is determined by its two-dimensional coordinates on the observation surface. The set of these points is the observation grid points of the target process.

[0157] Furthermore, this coefficient is equal to the complex attenuation factor when the sound wave propagates from the equivalent source location to the observation grid point location. Then, the intensity value of each equivalent source is multiplied by its corresponding sound propagation coefficient to obtain the equivalent source sound pressure component generated by the equivalent source at the observation grid point. Finally, all the equivalent source sound pressure components from all equivalent sources at the observation grid point are summed to obtain the complex result, which is the complex sound pressure value of the target process at the observation grid point.

[0158] Furthermore, the above operation is repeated for all observation grid points to obtain a set of sound pressure amplitude values ​​that correspond one-to-one with the observation grid points. These values ​​are then arranged into a two-dimensional array according to the row and column order of the observation grid points on the observation surface. This array is the sound pressure amplitude distribution of the target process.

[0159] Furthermore, each rate of change in the sound pressure spatial gradient is divided by the inertial acoustic impedance of the air in the target process. The air inertial acoustic impedance is equal to the air density multiplied by the speed of sound. The quotient obtained is the velocity component of the air particle vibration at each location. These velocity components are arranged according to the location of the observation grid points to obtain the particle vibration velocity components of the target process.

[0160] Furthermore, the above calculation is performed on all observation grid points to obtain a set of sound intensity values. Arranging these sound intensity values ​​into a two-dimensional array according to the position of the observation grid points gives the sound intensity vector of the target process. Finally, the previously obtained sound pressure amplitude distribution is combined with this sound intensity vector. The sound pressure amplitude distribution describes the spatial distribution of the intensity of the abnormal noise, and the sound intensity vector describes the propagation direction of the abnormal noise energy. The combination of the two constitutes the abnormal noise sound field distribution of the target process.

[0161] In summary, by defining an observation surface within the spatial region surrounding the industrial equipment during the target process and arranging observation grid points on the observation surface, the continuous abnormal sound field reconstruction area is discretized into a finite number of grid positions, allowing sound field reconstruction to be performed on regular discrete points, reducing computational complexity while ensuring the controllability of spatial resolution.

[0162] In summary, the product of the equivalent source intensity and the sound propagation coefficient at the corresponding observation grid point is used as the equivalent source sound pressure component, and then the complex sound pressure value is obtained by summing them. By superimposing the contributions of all equivalent sources at the observation grid point, the forward propagation calculation from source intensity to sound field is realized, avoiding the physical limitation of directly measuring the sound pressure on the observation surface, so that the sound field on any virtual observation surface can be reconstructed.

[0163] In summary, by taking the real part of the complex sound pressure value as the sound pressure amplitude and combining them to obtain the sound pressure amplitude distribution, the amplitude information in the sound pressure signal is extracted while the phase information is discarded. The spatial distribution of the strength of abnormal sound pressure is intuitively displayed in the form of a pure scalar field, which makes it easier for engineers to quickly locate the area with the maximum sound pressure.

[0164] In summary, the ratio of the difference in sound pressure amplitude between adjacent observation grid points to the spatial distance is used as the sound pressure spatial gradient, and the ratio of the sound pressure spatial gradient to the air inertial acoustic impedance is used as the particle velocity component. The vibration velocity field of air particles can be derived from the sound pressure amplitude distribution, and vector information in the sound field can be obtained without the need for additional particle velocity sensors.

[0165] In summary, the sound intensity vector is obtained by cross-spectral processing of the complex sound pressure value and the particle velocity component. The sound pressure amplitude distribution and the sound intensity vector together constitute the abnormal sound field distribution. The sound intensity vector indicates the direction and magnitude of sound energy propagation. Combined with the sound pressure amplitude distribution, it provides two complementary pieces of information: the energy intensity and the energy flow direction of the abnormal sound, thus fully describing the sound field characteristics of the abnormal sound.

[0166] Compared with the prior art, the present invention has the following beneficial effects:

[0167] 1. By triggering a microphone array with a synchronous clock to collect multi-channel time-domain sound pressure and performing short-time Fourier transform, a time-frequency complex sound pressure vector and cross-spectral matrix are constructed. Then, eigenvalue decomposition is performed to extract the dominant sound propagation vector corresponding to the largest eigenvalue. The spatial correlation of time-frequency points is calculated by combining the steering vector and the median is used as the threshold to filter the set of abnormal sound time-frequency points. This method accurately preserves the time-frequency points where the abnormal sound component is dominant, eliminates the interference of background noise and non-abnormal sound pulses, and significantly improves the signal-to-noise ratio and anti-environmental noise capability of abnormal sound time-frequency point identification.

[0168] 2. Based on the complex sound pressure vector of the time-frequency point set of the abnormal noise and the sound field transfer response from the equivalent source to the array, the equivalent source intensity is inverted. Then, the sound pressure amplitude distribution and particle velocity components on the observation surface are reconstructed by the equivalent source intensity and the sound propagation coefficient of the observation surface. The two are further cross-spectral processed to obtain the sound intensity vector. Finally, the sound pressure amplitude distribution and the sound intensity vector are combined to form the abnormal noise sound field distribution. This method provides the strength distribution of sound pressure energy and the direction information of sound energy propagation, and fully describes the spatial sound field characteristics of the abnormal noise. Moreover, the equivalent source inversion process adopts a stable transfer matrix with regularization, which ensures the numerical stability of the inversion results and avoids drastic fluctuations in sound pressure caused by small measurement errors.

[0169] like Figure 2 The diagram shown is a functional block diagram of an industrial equipment abnormal noise sound field measurement system based on an acoustic array, provided in an embodiment of the present invention.

[0170] The industrial equipment abnormal noise sound field measurement system 100 based on acoustic array described in this invention can be installed in an electronic device. Depending on the functions implemented, the industrial equipment abnormal noise sound field measurement system 100 based on acoustic array may include a multi-channel time-domain sound pressure module 101, a time-frequency complex sound pressure vector module 102, a dominant sound propagation vector module 103, an abnormal noise time-frequency point set module 104, an equivalent source intensity module 105, and an abnormal noise sound field distribution module 106. The module described in this invention can also be called a unit, which refers to a series of computer program segments that can be executed by the processor of an electronic device and can perform a fixed function, and is stored in the memory of the electronic device.

[0171] In this embodiment, the functions of each module / unit are as follows:

[0172] A multi-channel time-domain sound pressure module synchronously acquires sound waves generated by industrial equipment based on a microphone array of the target process to obtain the multi-channel time-domain sound pressure of the target process;

[0173] The time-frequency complex sound pressure vector module performs a short-time Fourier transform on the multi-channel time-domain sound pressure to obtain the time-frequency complex sound pressure vector of the target process;

[0174] The dominant sound propagation vector module constructs a cross-spectral matrix of time-frequency points in the target process based on the time-frequency complex sound pressure vector, and performs eigenvalue decomposition on the cross-spectral matrix to obtain the dominant sound propagation vector of the target process;

[0175] The abnormal noise time-frequency point set module calculates the guide vector of the scanning grid points in the target process based on the geometric position of the microphone array, and filters the abnormal noise time-frequency point set of the target process based on the median of the spatial correlation coefficient between the dominant sound propagation vector and the guide vector.

[0176] The equivalent source strength module, based on the complex sound pressure vector of the abnormal noise at the time-frequency point and the sound field transmission response from the equivalent source to the array in the target process, inversely calculates the equivalent source strength of the target process.

[0177] The abnormal noise sound field distribution module reconstructs the sound pressure and particle velocity on the observation surface based on the equivalent source intensity and the sound propagation coefficient from the equivalent source to the observation surface corresponding to the array, thereby obtaining the abnormal noise sound field distribution of the target process.

[0178] In the several embodiments provided by this invention, it should be understood that the disclosed methods and systems can be implemented in other ways. For example, the system embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and other division methods may be used in actual implementation.

[0179] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0180] Furthermore, the functional modules in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or in the form of hardware plus software functional modules.

[0181] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0182] The embodiments of this application can acquire and process relevant data based on artificial intelligence technology. Artificial intelligence is the theory, method, technology, and application system that uses digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to obtain optimal results.

[0183] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for measuring the abnormal noise sound field of industrial equipment based on an acoustic array, characterized in that, The method includes: S1. Based on the target process, a microphone array synchronously collects the sound waves generated by industrial equipment to obtain the multi-channel time-domain sound pressure of the target process; S2. Perform a short-time Fourier transform on the multi-channel time-domain sound pressure to obtain the time-frequency complex sound pressure vector of the target process; S3. Construct the cross-spectral matrix of the time-frequency points in the target process based on the time-frequency complex sound pressure vector, and perform eigenvalue decomposition on the cross-spectral matrix to obtain the dominant sound propagation vector of the target process; S4. Calculate the guiding vector of the scanning grid points in the target process according to the geometric position of the microphone array, and filter the abnormal sound time-frequency point set of the target process based on the median of the spatial correlation coefficient between the dominant sound propagation vector and the guiding vector. S5. Based on the complex sound pressure vector of the abnormal noise at the time-frequency point and the sound field transmission response from the equivalent source to the array in the target process, the equivalent source intensity of the target process is inverted. S6. Based on the equivalent source intensity and the sound propagation coefficient from the equivalent source to the observation surface corresponding to the array, reconstruct the sound pressure and particle velocity on the observation surface to obtain the abnormal sound field distribution of the target process.

2. The method for measuring the abnormal noise sound field of industrial equipment based on an acoustic array as described in claim 1, characterized in that, The microphone array based on the target process synchronously acquires sound waves generated by industrial equipment to obtain the multi-channel time-domain sound pressure of the target process, including: The microphone array is triggered by a synchronous clock signal to collect the equipment sound waves of the target process, thereby obtaining the multi-channel analog voltage signal of the target process; The multi-channel analog voltage signal is converted from analog to digital to obtain the time-domain sound pressure sampling sequence of the target process; The time-domain sound pressure sampling sequence is combined according to the microphone channel number to obtain the multi-channel time-domain sound pressure of the target process.

3. The method for measuring the abnormal sound field of industrial equipment based on an acoustic array as described in claim 1, characterized in that, The step of performing a short-time Fourier transform on the multi-channel time-domain sound pressure to obtain the time-frequency complex sound pressure vector of the target process includes: Cosine square windows are applied to the channel time-domain sound pressures of the multi-channel time-domain sound pressures to perform frame division, thereby obtaining the windowed time-domain sound pressure sequence of the target process; The windowed time-domain sound pressure sequence is subjected to a discrete Fourier transform to obtain the complex sound pressure spectrum of the target process; Arrange the complex sound pressure spectrum in channel order to obtain the time-frequency complex sound pressure vector of the target process.

4. The method for measuring the abnormal noise sound field of industrial equipment based on an acoustic array as described in claim 1, characterized in that, The process of constructing the cross-spectral matrix of time-frequency points in the target process based on the time-frequency complex sound pressure vector, and performing eigenvalue decomposition on the cross-spectral matrix to obtain the dominant acoustic propagation vector of the target process includes: The cross-spectral matrix of the target process is obtained by performing an outer product operation between the time-frequency complex sound pressure vector and the corresponding conjugate transpose row vector; The cross-spectral matrix is ​​decomposed into eigenvalues ​​to obtain the feature vector set of the target process, and the feature vector corresponding to the largest eigenvalue in the feature vector set is taken as the dominant acoustic propagation vector of the target process.

5. The method for measuring the abnormal noise sound field of industrial equipment based on an acoustic array as described in claim 1, characterized in that, The step of calculating the steering vector of the scanning grid points during the target process based on the geometric position of the microphone array, and filtering the set of abnormal sound time-frequency points of the target process based on the median spatial correlation coefficient between the dominant sound propagation vector and the steering vector, includes: Using the three-dimensional coordinates of the microphones in the microphone array in the spatial coordinate system as the reference origin, the three-dimensional spatial region around the industrial equipment in the target process is scanned to obtain the scanning grid points of the target process. The steering vector of the target process is calculated based on the spatial distance from the scanning grid points to the microphone and the corresponding angular frequency; The time-frequency spatial correlation of the target process is calculated based on the dominant acoustic propagation vector and the steering vector; Using the median of the maximum spatial correlation among the time-frequency points as the screening criterion, time-frequency points with a maximum spatial correlation greater than the median are retained to obtain the abnormal time-frequency point set of the target process.

6. The method for measuring the abnormal sound field of industrial equipment based on an acoustic array as described in claim 5, characterized in that, The step of calculating the steering vector for the target process based on the spatial distance from the scanning grid points to the microphone and the corresponding angular frequency includes: The ratio of the spatial distance from the scanning grid point to the microphone to the speed of sound is used as the sound wave propagation time value of the target process; The product of the sound wave propagation time value and the corresponding angular frequency is used as the phase delay value of the target process; The cosine of the phase delay value is used as the real part and the sine of the phase delay value is used as the imaginary part to construct the complex value of the phase delay of the target process, and the complex value of the phase delay is arranged in channel order to form the guide vector of the target process.

7. The method for measuring the abnormal sound field of industrial equipment based on an acoustic array as described in claim 5, characterized in that, The formula for calculating the spatial correlation of time and frequency points includes: in, The spatial correlation of the time-frequency points, This represents the total number of microphone channels. This is the sequence index of the microphone channel. The first of the dominant sound propagation vectors One element, The first guide vector One element, As a regularization factor, The dominant sound propagation vector, The guide vector is described above.

8. The method for measuring the abnormal noise sound field of industrial equipment based on an acoustic array as described in claim 1, characterized in that, The method of retrieving the equivalent source intensity of the target process based on the complex sound pressure vector of the abnormal noise at the concentrated abnormal noise time-frequency point and the sound field transmission response from the equivalent source to the array in the target process includes: The inverse acoustic field response matrix of the target process is obtained by performing a conjugate transpose on the equivalent source-to-array acoustic field transfer response matrix. The sound field inverse response matrix is ​​regularized to obtain the stable transfer matrix of the target process, and the stable transfer matrix is ​​inverted to obtain the inversion matrix of the target process. Multiplying the inversion matrix by the abnormal sound pressure vector yields the equivalent source intensity of the target process.

9. The method for measuring the abnormal sound field of industrial equipment based on an acoustic array as described in claim 1, characterized in that, The step of reconstructing the sound pressure and particle velocity on the observation surface based on the equivalent source intensity and the sound propagation coefficient from the equivalent source to the observation surface corresponding to the array, to obtain the abnormal sound field distribution of the target process, includes: An observation surface is defined within the spatial region surrounding the industrial equipment during the target process, and observation grid points for the target process are arranged on the observation surface. The product of the equivalent source intensity and the sound propagation coefficient of the corresponding observation grid point is taken as the equivalent source sound pressure component. Then, the equivalent source sound pressure component is summed to obtain the complex sound pressure value of the target process. The real part of the complex sound pressure value is taken as the sound pressure amplitude, and the sound pressure amplitude is combined to obtain the sound pressure amplitude distribution of the target process; The ratio of the difference in sound pressure amplitude between adjacent observation grid points in the target process to the spatial distance between the adjacent observation grid points is taken as the sound pressure spatial gradient of the target process, and the ratio of the sound pressure spatial gradient to the air inertial acoustic impedance in the target process is taken as the particle velocity component of the target process. The complex sound pressure value and the particle velocity component are cross-spectral processed to obtain the sound intensity vector of the target process, and the sound pressure amplitude distribution and the sound intensity vector are combined to form the abnormal sound field distribution of the target process.

10. A sound field measurement system for industrial equipment abnormal noise based on an acoustic array, used to implement the sound field measurement method for industrial equipment abnormal noise based on an acoustic array as described in any one of claims 1-9, characterized in that, The system includes: A multi-channel time-domain sound pressure module synchronously acquires sound waves generated by industrial equipment based on a microphone array of the target process to obtain the multi-channel time-domain sound pressure of the target process; The time-frequency complex sound pressure vector module performs a short-time Fourier transform on the multi-channel time-domain sound pressure to obtain the time-frequency complex sound pressure vector of the target process; The dominant sound propagation vector module constructs a cross-spectral matrix of time-frequency points in the target process based on the time-frequency complex sound pressure vector, and performs eigenvalue decomposition on the cross-spectral matrix to obtain the dominant sound propagation vector of the target process; The abnormal noise time-frequency point set module calculates the guide vector of the scanning grid points in the target process based on the geometric position of the microphone array, and filters the abnormal noise time-frequency point set of the target process based on the median of the spatial correlation coefficient between the dominant sound propagation vector and the guide vector. The equivalent source strength module, based on the complex sound pressure vector of the abnormal noise at the time-frequency point and the sound field transmission response from the equivalent source to the array in the target process, inversely calculates the equivalent source strength of the target process. The abnormal noise sound field distribution module reconstructs the sound pressure and particle velocity on the observation surface based on the equivalent source intensity and the sound propagation coefficient from the equivalent source to the observation surface corresponding to the array, thereby obtaining the abnormal noise sound field distribution of the target process.

Images