Method for positioning pulse sound source under strong harmonic background
By using a microphone array to perform continuous wavelet transform and inverse transform under strong harmonic background, combined with the time-domain equivalent source method, the problem of inaccurate pulse sound source localization in traditional methods is solved, and high-precision pulse sound source identification and localization is achieved.
Patent Information
- Application Number
- CN202511760533.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-11-27
- Publication Date
- 2026-01-16
AI Technical Summary
Existing technologies struggle to accurately separate and locate weak pulse sound sources in the context of strong harmonics. Traditional time-domain near-field acoustic holography leads to confusion and information loss in reconstruction results, and existing post-processing methods are ineffective in recovering pulse signal characteristics.
A microphone array was arranged under a strong harmonic background, and continuous wavelet transform and inverse transform were performed. Based on the time-frequency distribution map, the characteristic frequency band was selected, and the wavelet time-domain sound pressure on the sound source surface was reconstructed using the time-domain equivalent source method. The location of the pulse sound source was identified by the sound pressure cloud map.
It significantly improves the ability to identify and locate weak pulse sound sources in the context of strong harmonics, reduces the amount of computation, is suitable for complex industrial sites, allows for flexible arrangement of microphone arrays, and overcomes the limitations of traditional methods.
Smart Images
Figure CN121348231A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application belongs to the technical field of sound source positioning and acoustic fault diagnosis, more specifically, a method for positioning a transient pulse sound source in a strong harmonic noise background. BACKGROUND
[0002] In the fault diagnosis of industrial equipment such as compressors and gearboxes, accurate positioning of transient pulse sound sources caused by faults such as discharge and impact is crucial. However, such pulse sound sources are often submerged in the strong harmonic noise background generated by the steady operation of the equipment, making effective separation and positioning a great challenge. Time-domain near-field acoustic holography technology can reconstruct the time-domain sound field of the sound source surface and is an effective tool for identifying transient sound sources. It has shown good application results in the positioning of ideal transient sound sources such as pulse spheres and pulse plates. However, the traditional time-domain near-field acoustic holography technology has limitations in positioning pulse sound sources in a strong harmonic background. The reason is that the input signal (microphone measurement signal) and the output result (reconstructed sound source surface sound pressure) of the traditional time-domain near-field acoustic holography method are full-band time-domain signals, resulting in the reconstructed sound field being a superposition of harmonic and pulse sound source signals, making it difficult to distinguish between the two types of sound sources on the final sound pressure cloud map. When the sound pressure level of the harmonic sound source is much higher than that of the pulse sound source, the weak fault pulse signal will be completely submerged in the strong harmonic background, resulting in failure to locate the fault. Although post-processing methods can be used to separate the sound sources, such as frequency domain filtering to separate different sound source components. However, practice shows that this method is difficult to accurately restore the original characteristics of the pulse, and the fundamental reason is that the characteristics of the pulse signal are damaged in the reconstruction process of the time-domain near-field acoustic holography. The damage to the signal mainly comes from two aspects: (a) transient pulse signals have wideband characteristics, while the limited sampling rate of the actual measurement system cannot fully meet the requirements of the Nyquist sampling theorem for high-frequency components, resulting in frequency aliasing in the reconstruction process, which causes the algorithm to misjudge false low-frequency components as real sound sources, ultimately resulting in pseudo-sources or causing the real source position to shift on the reconstructed image. (b) The inverse problem of time-domain near-field acoustic holography is essentially ill-posed, and small measurement errors can cause large deviations in the reconstructed solution. In order to obtain a stable solution, regularization filtering technology must be introduced. However, when the energy of the pulse signal is significantly lower than that of the harmonic component, the regularization process will inevitably suppress these weak, fault characteristic high-frequency pulse components, resulting in distortion of their waveforms or complete filtering, thus losing key information in the reconstruction result. In summary, the existing time-domain near-field acoustic holography technology is difficult to reliably and effectively separate and locate pulse fault sound sources in a strong harmonic background. SUMMARY
[0003] To overcome the shortcomings of the prior art, this invention provides a method for locating pulse sound sources in a strong harmonic background, aiming to fundamentally avoid the submergence and interference of strong harmonic components on weak pulse signals during the reconstruction process of the time-domain equivalent source method, and improve the identification ability and positioning accuracy of weak pulse sound sources in a strong harmonic background.
[0004] The present invention solves the technical problem by adopting the following technical solution: The method for locating a pulse sound source under strong harmonic background is characterized by the following steps: A microphone array is arranged in the sound field containing the pulse sound source to be measured to collect time-domain sound pressure signals; a continuous wavelet transform is performed on each time-domain sound pressure signal to obtain a wavelet coefficient matrix; a characteristic frequency band with significant pulse signal energy is determined based on the time-frequency distribution map; an inverse continuous wavelet transform is performed on the wavelet coefficient matrix within the characteristic frequency band to obtain a reconstructed wavelet time-domain signal with enhanced pulse signal; the reconstructed wavelet time-domain signal is used as input, and the wavelet time-domain sound pressure on the sound source reconstruction surface is reconstructed using the time-domain equivalent source method; the target time of pulse occurrence is identified based on the reconstructed sound pressure, and a sound pressure cloud map is generated; the location of the pulse sound source is determined by the location of the maximum sound pressure amplitude, thereby effectively extracting and reconstructing pulse signal components from the mixed sound field, and thus locating the pulse sound source under strong harmonic background.
[0005] The method for locating a pulse sound source under a strong harmonic background according to the present invention includes the following steps: Step 1: Arrange a [structure / structure] in the radiated sound field of the sound source under test, which contains pulse signals. A measurement array consisting of several microphones; utilizing the measurement array Each microphone synchronously acquires the time-domain sound pressure signal of the radiated sound field, obtaining... The time-domain sound pressure signal of the path will be the first The time-domain sound pressure signal is denoted as ,in ; Step 2, regarding the above Each time-domain sound pressure signal in the path of the time-domain sound pressure signal is subjected to a continuous wavelet transform, and the results are obtained one-to-one. The wavelet coefficient matrix will be the first one. The wavelet coefficient matrix is denoted by equation (1). : (1); In formula (1): As a scale factor, The time shift factor, Describe the wavelet basis functions. express The complex conjugate; Step 3, based on the above Any one of the wavelet coefficient matrices generates a time-frequency distribution map, and a characteristic frequency band is selected according to energy distribution characteristics of the pulse signal and the harmonic signal in the time-frequency distribution map ; Step 4, performing inverse continuous wavelet transform on each of the wavelet coefficient matrices in the characteristic frequency band corresponding scale range to obtain a first reconstructed wavelet time-domain signal Step 5, using a time-domain equivalent source method, and taking the first reconstructed wavelet time-domain signal as an input of the time-domain equivalent source method to obtain reconstructed surface wavelet time-domain sound pressures at different times on a sound source reconstruction surface containing the pulse signal in the characteristic frequency band Step 6, in the reconstructed surface wavelet time-domain sound pressures , a time corresponding to a peak value of the reconstructed surface wavelet time-domain sound pressures is a target time of a pulse signal event, and based on a sound pressure amplitude on the sound source reconstruction surface at the target time, a sound pressure amplitude cloud map is generated, and a position corresponding to the pulse sound source on the sound source reconstruction surface is a maximum sound pressure amplitude in the sound pressure amplitude cloud map, so that positioning of the pulse sound source is completed. (2) In formula (2), a denotes a wavelet tolerance constant; b denotes a scale range corresponding to a frequency band; c denotes a wavelet center frequency; and d denotes a time sampling rate. Step 5, using a time-domain equivalent source method, and taking the first reconstructed wavelet time-domain signal as an input of the time-domain equivalent source method to obtain reconstructed surface wavelet time-domain sound pressures at different times on a sound source reconstruction surface containing the pulse signal in the characteristic frequency band Step 6, in the reconstructed surface wavelet time-domain sound pressures Step 6, in the reconstructed surface wavelet time-domain sound pressures , a time corresponding to a peak value of the reconstructed surface wavelet time-domain sound pressures is a target time of a pulse signal event, and based on a sound pressure amplitude on the sound source reconstruction surface at the target time, a sound pressure amplitude cloud map is generated, and a position corresponding to the pulse sound source on the sound source reconstruction surface is a maximum sound pressure amplitude in the sound pressure amplitude cloud map, so that positioning of the pulse sound source is completed.
[0006] The method for positioning a pulse sound source in a strong harmonic background has the following characteristics: in step 5, the time-domain equivalent source method is used to obtain the sound source reconstruction surface wavelet time-domain sound pressures as follows: First, a plurality of equivalent sources are arranged, so that the sound source reconstruction surface is located in a sound field region between the equivalent sources and a measurement array; the i-th equivalent source is denoted as is represented by formula (3) The superposition of sound pressure of the equivalent sources: (3) ; In formula (3) : Let represent the sound pressure acquisition time, let represent the distance between the th equivalent source and the th microphone; Let Let represent the delay time, ; let represent the sound speed; Let represent the source strength of the th equivalent source at the th time point; Then, the sound pressure acquisition time is discretized into each sampling time point, let represent the th sampling time point, and have: , wherein: is the initial time, that is, when , the sound source does not emit sound, let represent the time sampling rate; let represent the number of discrete time points contained in the total acquisition time, that is, ; Further, the delay time in formula (3) is discretized into , , the th reconstructed wavelet time domain signal is discretized into , and the source strength is discretized into ; Then, interpolation processing is performed: the discretized source strength is written into the form shown in formula (4) through linear interpolation: (4); In formula (4) : let represent the th time interpolation node, , ; wherein: , is the minimum value in all ; Let represent the Lagrange linear interpolation function, which is represented by formula (5) : (5); In formula (5): ; After the completion of the discrete and interpolation, according to the first wavelet time domain signal after the reconstruction of the road obtained by formula (6) represents the discrete first wavelet time domain signal after the reconstruction of the road : (6); traverse the measurement array on measurement points, get the matrix equation as formula (7): (7); In formula (7): to represent the column vector composed of time domain sound pressure collected by measurement points at time, namely: , with the right upper corner mark "T" to represent the transpose of the vector to represent the source intensity column vector composed of equivalent source intensity at time of equivalent sources, namely: to represent the transfer matrix of equivalent sources to measurement points, the size is ; The element of the first row, the first column of the transfer matrix is shown in formula (8): (8); According to the source intensity column vector at each time, the reconstruction surface wavelet time domain sound pressure of the sound source reconstruction surface is calculated by formula (9): (9); In formula (9): to represent the column vector composed of reconstruction surface wavelet time domain sound pressure of reconstruction points on the sound source reconstruction surface at time, namely: in, For the reconstruction of the sound source surface, the first The reconstruction points are at The reconstructed surface wavelet time-domain sound pressure at time, with This indicates the total number of reconstruction points on the sound source reconstruction surface. ; by express An equivalent source onto the sound source reconstruction surface The transfer matrix of the reconstructed points is of size . Transfer matrix The Middle line, number The elements of the column are shown in equation (10): (10); In formula (10): Indicates the first The first equivalent source and sound source reconstruction surface The distance between the reconstruction points; Delay time for: .
[0007] The method for locating a pulse sound source under strong harmonic background is also characterized by the following: the sound source reconstruction surface is the actual surface of the sound source or a set plane or curved surface located between the measurement array and the actual surface of the sound source to be measured; the set curved surface fully or partially surrounds the sound source to be measured.
[0008] The method for locating pulse sound sources under strong harmonic backgrounds in this invention is characterized by the following: the sound source under test containing pulse signals refers to transient signals generated by physical phenomena such as discharge, impact, explosion, or leakage; the time-domain waveform of the transient signal is a non-periodic transient peak with a duration between 0.1 milliseconds and 200 milliseconds; and the transient signal exhibits short duration, vertical stripes, and wideband energy concentration characteristics on the time-frequency distribution diagram, which can be distinguished from harmonic signals, which exhibit long duration and narrow band on the time-frequency diagram.
[0009] Compared with the prior art, the beneficial effects of the present invention are reflected in: 1. This invention utilizes continuous wavelet transform and inverse transform to actively separate and reconstruct the frequency band signal dominated by pulse characteristics from the mixed sound field before reconstruction. This fundamentally avoids the submergence and interference of strong harmonic components on weak pulse signals during the reconstruction process using the time-domain equivalent source method. Compared to conventional time-domain equivalent source methods that directly process the full-band signal or methods that post-process the reconstruction results, this invention significantly improves the ability to identify and locate weak pulse sound sources in a strong harmonic background.
[0010] 2、The application adopts a "pre-processing" strategy, that is, before sound field reconstruction, only a small number of microphone sampling points are subjected to continuous wavelet transform and inverse transform. In contrast, the traditional "post-processing" strategy needs to perform the same time-frequency analysis operation on the sound source surface reconstruction points which are much more than the sampling points after sound field reconstruction. Since sound field reconstruction is the core of the application calculation, and the time-frequency analysis calculation overhead is relatively fixed, applying this fixed overhead to a small number of sampling points instead of a large number of reconstruction points can fundamentally reduce the total calculation amount. This efficiency optimization based on the front movement of the processing link makes the application particularly suitable for high spatial resolution sound source cloud maps.
[0011] 3、The application is based on the time-domain equivalent source method for sound field reconstruction, which has no special requirements for the geometric shape of the microphone array. Therefore, the measurement array of the application can be flexibly arranged in any shape (such as irregular curved surface, non-uniformly distributed array, etc.), effectively overcoming the dependence of traditional acoustic holography and other methods on regular shaped measurement surfaces (such as plane, cylindrical surface, spherical surface), greatly improving the applicability and flexibility of the method in complex industrial sites. BRIEF DESCRIPTION OF DRAWINGS
[0012] Figure 1 is a flow chart of the method of the application; Figure 2 is a spatial distribution diagram of the spherical simulation model used in the embodiment; Figure 3 is a position identification schematic diagram of one harmonic point sound source and two pulse point sound sources in the embodiment; Figure 4 is a time-frequency diagram obtained by continuous wavelet transform of the time-domain sound pressure signal of No. 1 measurement point in the embodiment; Figure 5 is a positioning result schematic of the first pulse point sound source by the method of the application; Figure 6 is a positioning result schematic of the first pulse point sound source by the time-domain equivalent source near-field acoustic holography method in the prior art as a comparative example; Figure 7 is a positioning result schematic of the first pulse point sound source by the time-domain equivalent source near-field acoustic holography method in the prior art combined with wavelet post-processing as a comparative example; Figure 8 is a positioning result schematic of the second pulse point sound source by the method of the application; Figure 9 is a positioning result schematic of the second pulse point sound source by the time-domain equivalent source near-field acoustic holography method in the prior art as a comparative example; Figure 10is the positioning result of the second pulse point sound source by the time domain equivalent source near-field acoustic holography method in the prior art combined with wavelet post-processing as a comparative example; Figure label: a1 is a measurement point, a2 is a reconstruction point, a3 is an equivalent source point, a4 is a harmonic point sound source, a5 is a first pulse point sound source, and a6 is a second pulse point sound source. DETAILED DESCRIPTION
[0013] Reference Figure 1 In this embodiment, the method for positioning the pulse sound source in the strong harmonic background comprises the following steps:
[0014] In this embodiment, the method for positioning the pulse sound source in the strong harmonic background comprises the following steps: Step 1, arranging a measurement array composed of transducers in the radiation sound field of the to-be-measured sound source containing the pulse signal; the measurement array can be of any shape, for example but not limited to a planar array, a cylindrical array, a spherical array, or an irregular array arranged according to the space constraints on site, etc.; using the transducers of the measurement array to synchronously collect the time domain sound pressure signals of the radiation sound field, obtaining time domain sound pressure signals, the first time domain sound pressure signal is denoted as , wherein .
[0015] Step 2, performing continuous wavelet transform on each of the time domain sound pressure signals, one by one, to obtain wavelet coefficient matrices, the first wavelet coefficient matrix is denoted as characterized by formula (1): (1) ; In formula (1): is a scale factor, is a time translation factor, indicates a wavelet base function, express The complex conjugate; These are complex-valued wavelet basis functions, such as the classic Amor wavelet, Bump wavelet, Morse wavelet, etc.
[0016] Step 3, based on Generate a time-frequency distribution map from any wavelet coefficient matrix in the given wavelet coefficient matrices, or... The time-frequency distribution map is generated by averaging the wavelet coefficient matrices. The characteristic frequency band is then selected based on the energy distribution characteristics of the pulse and harmonic signals in the time-frequency distribution map. .
[0017] Step 4: In the characteristic frequency band Within the corresponding scale range, for Perform an inverse continuous wavelet transform on each of the wavelet coefficient matrices to obtain the values in the characteristic frequency band. The internal pulse signal is enhanced. The purpose of reconstructing wavelet time-domain signals is to discard characteristic frequency bands. In addition to the frequency bands, to suppress background harmonic noise while preserving and enhancing the signal-to-noise ratio of the target pulse signal, the first The reconstructed wavelet time-domain signal is denoted as and characterized by equation (2). : (2); In formula (2): Denotes the wavelet tolerance constant; Indicates frequency band The corresponding scale range; And there are: , , Indicates the wavelet center frequency. Indicates the time sampling rate; In order to effectively implement wavelet transform and its inverse transform in the frequency domain, in equation (1) It should be uniform sampling, and the signal's time sampling rate should be... It should be at least twice the maximum frequency to be analyzed.
[0018] Step 5: Employ the time-domain equivalent source method, and... The reconstructed wavelet time-domain signal is used as the input to the time-domain equivalent source method to obtain the signal in the characteristic frequency band. Wavelet time-domain sound pressure levels on the reconstructed surface of a sound source containing pulse signals at different times. The specific procedure is as follows: First, arrange A number of equivalent sources are used to locate the sound source reconstruction surface in the sound field region between the equivalent sources and the measurement array; represent the first equivalent source; then the wavelet time-domain signal after the fourth road reconstruction is represented by the superposition of the sound pressure radiated by the first equivalent source as shown in equation (3): (3) ; In equation (3): Let represent the sound pressure acquisition time, let represent the distance between the first equivalent source and the first microphone; Let Let represent the delay time, ; let represent the sound speed; Let represent the source strength of the first equivalent source at time ; and let Then, in order to realize numerical calculation, the above continuous physical model needs to be converted into a calculable discrete matrix equation, that is, the following discrete processing is performed: the sound pressure acquisition time is discretized into each sampling time, let represent the first sampling time, and have: where: is the initial time, that is, when the sound source does not emit sound, let represent the time sampling rate; let represent the number of discrete time points contained in the total acquisition time, that is, , ; further, the delay time in equation (3) is discretized into , , the wavelet time-domain signal after the fourth road reconstruction is discretized into , and the source strength is discretized into .
[0019] Since the discrete time of the equivalent source is usually not an integer multiple of the sampling time interval, the equivalent source strength needs to be interpolated as follows: the discretized source strength is written in the form shown in equation (4) through linear interpolation: (4) ; In equation (4): Let Indicates the first Time interpolation nodes , ;in, , For all The minimum value in; by The Lagrange linear interpolation function is represented by equation (5): (5); In equation (5): ; The interpolation function is not limited to the Lagrange linear interpolation of equation (5), but can also use quadratic interpolation, cubic interpolation, or spline interpolation, etc. After discretization and interpolation are completed, the first interpolation function is used based on the first interpolation function represented by equation (3). Wavelet time-domain signal after path reconstruction We obtain the discrete first term represented by equation (6). Wavelet time-domain signal after path reconstruction : (6); Traversing the measurement array With 1 measurement point, the matrix equation is obtained as shown in equation (7): (7); In equation (7): by express Each measurement point is at Real-time data collection A column vector composed of time-domain sound pressures, namely: The superscript "T" indicates the transpose of the vector. by express An equivalent source in Moment A source strength column vector composed of equivalent source strengths, namely: by express An equivalent source to The transfer matrix for each measurement point is of size [value]. ; Transfer matrix The line, number The elements of the column are shown in equation (8): (8); According to each Source strength column vector at time step The wavelet time-domain sound pressure of the reconstructed surface of the sound source is calculated using equation (9): (9); In equation (9): by Represents the surface of sound source reconstruction The reconstruction points are at Moment A column vector composed of wavelet time-domain sound pressure levels of the reconstructed surface. Right now: in, For the reconstruction of the sound source surface, the first The reconstruction points are at The reconstructed surface wavelet time-domain sound pressure at time, with This indicates the total number of reconstruction points on the sound source reconstruction surface. ; by express An equivalent source onto the sound source reconstruction surface The transfer matrix of the reconstructed points is of size . Transfer matrix The Middle line, number The elements of the column are shown in equation (10): (10); In formula (10): Indicates the first The first equivalent source and sound source reconstruction surface The distance between the reconstruction points; Delay time for: .
[0020] Step 6: Reconstruct the wavelet time-domain sound pressure level. middle, The peak value corresponds to the target time of the pulse signal event. Based on the sound pressure amplitude on the sound source reconstruction surface at the target time, a sound pressure amplitude cloud map is generated. The maximum sound pressure amplitude in the sound pressure amplitude cloud map is the corresponding position of the pulse sound source on the sound source reconstruction surface, thus completing the localization of the pulse sound source.
[0021] In a specific embodiment, the sound source reconstruction surface is the actual surface of the sound source or a set plane or curved surface located between the measurement array and the actual surface of the sound source under test; the set curved surface fully or partially surrounds the sound source under test. The sound source under test containing pulse signals refers to transient signals generated by physical phenomena such as discharge, impact, explosion, or leakage; the time-domain waveform of the transient signal is a non-periodic transient peak with a duration between 0.1 milliseconds and 200 milliseconds; and the transient signal exhibits short duration, vertical stripes, and wideband energy concentration characteristics on the time-frequency distribution diagram, which can be distinguished from harmonic signals, which exhibit long duration and narrow band on the time-frequency diagram.
[0022] Simulation Experiment Simulation objective: To verify the performance of the method of the present invention in locating pulse sound sources under strong harmonic background. Specifically, it verifies that the time-domain equivalent source near-field acoustic holography method combined with wavelet preprocessing has better reliability and accuracy in pulse sound source localization compared with the conventional time-domain equivalent source near-field acoustic holography method and the time-domain equivalent source near-field acoustic holography method combined with wavelet postprocessing.
[0023] Simulation process: Taking the classic spherical model as an example, numerical simulation of weak pulse sound source localization is performed. The spatial distribution of the spherical simulation model is as follows: Figure 2 As shown. Measurement point a1 is arranged at a radius of... On the surface of a sphere, and The intervals in both directions are A total of 366 points. The sound source reconstruction surface is the radius. The sphere, with reconstruction point a2 arranged on the sphere, in and The intervals in both directions are A total of 3321 points. The equivalent source point a3 is arranged in a radius of... On the sphere, their distribution pattern is consistent with the measurement points. The simulation includes one harmonic frequency source and two pulse point sources, with their positions (in spherical coordinates)... (Representation): The harmonic frequency sound source a4 (circular) is located at... At this location, the first pulse point sound source a5 (equilateral triangle) is positioned relatively close to the harmonic point sound source a4, located in... At this location, the second pulse point sound source a6 (inverted triangle) is positioned relatively far from the harmonic point sound source a4, located in... The locations of the three sound sources are indicated as follows: Figure 3 As shown.
[0024] To simulate noise in a real industrial scenario that includes a fundamental frequency and its harmonic components related to rotational speed, the signal from harmonic frequency source a4 is derived from the fundamental frequency. It is composed of its harmonic components, with the highest frequency cutoff at... The maximum harmonic order is Fundamental frequency amplitude The harmonic amplitude attenuation coefficient is .by Let represent the harmonic signal, and its mathematical expression is shown in equation (11). Let the time sampling frequency be . The signal duration is 0.2s.
[0025] (11); To simulate weak fault pulse signals under a strong harmonic background, the amplitudes of two pulse point sound sources were set to be lower than those of the harmonic point sound sources. Both emitted sawtooth pulse signals within a time interval of 0-0.2s, with a time sampling frequency consistent with the aforementioned harmonic signals. The specific parameters of the pulse signals were set as follows: the first pulse of the first pulse point sound source a5 started at 1 / 33s, the pulse interval was 1 / 30s (i.e., 30Hz), and the pulse width was 1ms; the first pulse of the second pulse point sound source a6 started at 1 / 47s, the pulse interval was 1 / 60s (i.e., 60Hz), and the pulse width was 2ms. The signal amplitudes of both pulse sound sources were set to increase linearly with time (amplitude increasing from 0.04 to 0.12) to simulate the gradual manifestation of the fault.
[0026] In the spherical measurement array, the time-domain sound pressure signals acquired at each measurement point are the superposition result of the combined effects of the three point sources, with a signal acquisition duration of 0.2 seconds. First, a continuous wavelet transform is performed on the original time-domain signal obtained from the measurements to acquire its time-frequency distribution, thereby determining the target analysis frequency band suitable for pulse extraction. Taking the signal from measurement point 1 as an example, its time-frequency distribution after continuous wavelet transform is as follows... Figure 4 As shown in the figure. For ease of illustration and analysis, the horizontal axis in the figure uses "number of time points" (number of time points = time × time sampling rate) instead of the actual time unit. The vertical stripes shown in the figure are typical characteristics of pulse signals in the time-frequency diagram. Based on this, a frequency band with clear pulse characteristics and concentrated energy is selected as the analysis object; in this example, the range is 5990Hz to 6010Hz. Next, the wavelet coefficients corresponding to this frequency band are extracted, and the wavelet time-domain signal dominated by the pulse component at all measurement points in this frequency band is reconstructed through inverse continuous wavelet transform. Finally, the reconstructed wavelet time-domain signal is used as input, and the wavelet time-domain sound pressure at each reconstruction point on the sound source reconstruction surface is calculated using the time-domain equivalent source method. The reconstruction results of two time point numbers are compared below: (1) Select time 4195, at this time, the nearest reconstruction point to the first pulse point sound source a5 (a regular triangle) will occur pulse, compare the positioning effect of the following three methods for the pulse point sound source: (1) the wavelet preprocessing combined time domain equivalent source near-field acoustic holography method of the application, (2) the conventional time domain equivalent source near-field acoustic holography method, (3) the wavelet post-processing combined time domain equivalent source near-field acoustic holography method. The sound pressure distribution of the reconstruction surface calculated by the three methods is shown in Figure 5 , Figure 6 , Figure 7 respectively. As can be seen from the figure, the method of the application ( Figure 5 ) effectively filters out the interference of the harmonic sound source, and clearly locates the position of the first pulse point sound source a5 (a regular triangle). The conventional method ( Figure 6 ) cannot identify the position of the harmonic point sound source a4 (a circle) because the weak pulse signal is not extracted and the harmonic sound pressure is dominant. The wavelet post-processing combined method ( Figure 7 ) can filter out the influence of the harmonic sound source, but the weak pulse signal is distorted due to the error introduced by the time domain equivalent source near-field acoustic holography method, and false sound source artifacts are generated at the position where the pulse signal should not appear (an inverted triangle).
[0027] (2) Select time 4390, at this time, the nearest reconstruction point to the second pulse point sound source a6 (an inverted triangle) will occur pulse, and the positioning effect of the above three methods for the pulse sound source is compared. The sound pressure distribution of the reconstruction surface calculated by the three methods is shown in Figure 8 , Figure 9 , Figure 10 respectively. As can be seen from the figure, the conclusion is basically similar to the foregoing simulation, the method of the application ( Figure 8 ) effectively filters out the interference of the harmonic sound source, and clearly locates the position of the second pulse point sound source a6 (an inverted triangle). The conventional method ( Figure 9 ) can hardly identify the weak ghost at the position corresponding to the pulse sound source 2, but the bright spot of the harmonic sound source (a circle) is still dominant, and it cannot be determined whether the pulse sound source is located only at the inverted triangle position. The wavelet post-processing combined method ( Figure 10 ) can locate the second pulse point sound source a6, but false sound source artifacts are still generated at the non-sound source position (the circular position representing the harmonic sound source a4 and the left and right sides thereof) in the reconstruction result.
[0028] Therefore, compared with the conventional time domain equivalent source near-field acoustic holography method and the wavelet post-processing combined time domain equivalent source near-field acoustic holography method, the wavelet preprocessing combined time domain equivalent source near-field acoustic holography method proposed by the application has better reliability and accuracy in pulse sound source positioning.
Claims
1. A method of locating a pulsed acoustic source in the presence of strong harmonic background, characterized by: A microphone array is arranged in an acoustic field containing a pulse sound source to be measured to collect time-domain sound pressure signals; each time-domain sound pressure signal is subjected to continuous wavelet transform to obtain a wavelet coefficient matrix; Based on the time-frequency distribution, a characteristic frequency band with significant pulse signal energy is determined; inverse continuous wavelet transform is performed on the wavelet coefficient matrix in the characteristic frequency band to obtain a reconstructed wavelet time-domain signal with enhanced pulse signal; the reconstructed wavelet time-domain signal is taken as input to reconstruct a wavelet time-domain sound pressure on a sound source reconstruction surface by using a time-domain equivalent source method; the target time of pulse generation is identified according to the reconstructed sound pressure, and a sound pressure nephogram is generated; the position of the pulse sound source is determined by the maximum sound pressure amplitude, so as to effectively extract and reconstruct the pulse signal component from the mixed acoustic field, and thus locate the pulse sound source in a strong harmonic background.
2. The method of locating a pulsed acoustic source in a strong harmonic background according to claim 1, characterized in that The method comprises the following steps: Step 1: Arrange a [structure / structure] in the radiated sound field of the sound source under test, which contains pulse signals. A measurement array consisting of several microphones; utilizing the measurement array Each microphone synchronously acquires the time-domain sound pressure signal of the radiated sound field, obtaining... The time-domain sound pressure signal of the path will be the first The time-domain sound pressure signal is denoted as ,in ; Step 2, performing continuous wavelet transform on each of the plurality of time-domain sound pressure signals respectively to obtain a one-to-one corresponding plurality of wavelet coefficient matrices Step 2, performing continuous wavelet transform on each of the plurality of time-domain sound pressure signals respectively to obtain a one-to-one corresponding plurality of wavelet coefficient matrices Step 2, performing continuous wavelet transform on each of the plurality of time-domain sound pressure signals respectively to obtain a one-to-one corresponding plurality of wavelet coefficient matrices Step 2, performing continuous wavelet transform on each of the plurality of time-domain sound pressure signals respectively to obtain a one-to-one corresponding plurality of wavelet coefficient matrices : （1）； In formula (1): is a scale factor, is a time shift factor, denotes a wavelet basis function, denotes the complex conjugate of Step 3, based on the above A time-frequency distribution map is generated using any one of the wavelet coefficient matrices. A characteristic frequency band is then selected based on the energy distribution characteristics of the pulse and harmonic signals within the time-frequency distribution map. ; Step 4, in the characteristic frequency band for each of the wavelet coefficient matrices : （2）； In formula (2): denotes a wavelet admissible constant; denotes a frequency band corresponding to a scale range; And also: , , denotes the wavelet center frequency, denotes the time sampling rate; Step 5: Employ the time-domain equivalent source method and convert the... The reconstructed wavelet time-domain signal is used as the input to the time-domain equivalent source method to obtain the signal in the characteristic frequency band. Wavelet time-domain sound pressure levels on the reconstructed surface of a sound source containing pulse signals at different times. ; Step 6, the peak value of the sound pressure in the reconstructed wavelet time domain , The moment corresponding to the peak value of the sound pressure in the reconstructed wavelet time domain is the target moment of the occurrence of the pulse signal event; based on the sound pressure amplitude on the sound source reconstruction surface at the target moment, a sound pressure amplitude cloud chart is generated, and the position corresponding to the pulse sound source on the sound source reconstruction surface is the maximum sound pressure amplitude in the sound pressure amplitude cloud chart, thereby completing the positioning of the pulse sound source.
3. The method of locating a pulsed acoustic source in a strong harmonic background according to claim 1, wherein: The step 5 is to obtain the sound source reconstruction surface wavelet time domain sound pressure by using time domain equivalent source method according to the following process : First, arrange An equivalent source is used, such that the sound source reconstruction surface is located in the sound field region between the equivalent source and the measurement array; Indicates the first There are one equivalent source; then the first one mentioned in step 4... Wavelet time-domain signal after path reconstruction It is characterized by equation (3). The superposition of sound pressure radiated from the equivalent sources: (3) ； In formula (3), with denotes the sound pressure acquisition time, with denotes the distance between the th equivalent source and the th microphone; with with a delay time, ; with denotes the speed of sound; with representing the source strength of the th equivalent source at time t. Then the discrete processing is carried out: is the sound pressure acquisition time Discrete for each sampling time, to represent the first sampling time, and there is: , wherein: is the initial time, that is, when the sound source does not sound, to represent the time sampling rate; to represent the number of discrete time points contained in the total time length of acquisition, that is , , ; further, the delay time in formula (3) is discrete , , the first way reconstructed wavelet time domain signal is discrete , the source intensity is discrete ; Interpolation processing is performed again: the discrete source intensity is linearly interpolated is written as shown in expression (4): （4）； In formula (4): with denotes the first time interpolation node, , ; wherein, , is the minimum value of all . In denotes the Lagrange linear interpolation function, as characterized by equation (5): （5）； In formula (5): ; After the completion of the discretization and interpolation, the first wavelet time domain signal after road reconstruction discretized first wavelet time domain signal represented by equation (6) wavelet time domain signal after road reconstruction : （6）； The matrix equation as formula (7) is obtained by traversing the measurement points on the measurement array. The matrix equation as formula (7) is obtained by traversing the measurement points on the measurement array. （7）； In formula (7), In indicates a column vector consisting of the time-domain sound pressures measured at the time instant, i.e. : The transpose of a vector is denoted by a superscript "T" in the upper right corner In denotes one equivalent source in time, the source strength column vector composed of one equivalent source intensity, namely: with representing a transfer matrix of measuring points, of size ; The transfer matrix The element in the first row, first column is given by equation (8): （8）； According to each Source strength column vector at time instant The reconstructed surface wavelet time domain sound pressure of the sound source reconstruction surface is obtained by formula (9): （9）； In formula (9), with a column vector representing the wavelet domain sound pressure of the reconstruction point on the reconstruction surface at the time instant, That is, wherein, is the reconstructed wavelet domain sound pressure of the reconstructed surface at the time instant t for the reconstructed point on the reconstructed surface with the index i, is the reconstructed wavelet domain sound pressure of the reconstructed surface at the time instant t for the reconstructed point on the reconstructed surface with the index i, is the reconstructed wavelet domain sound pressure of the reconstructed surface at the time instant t for the reconstructed point on the reconstructed surface with the index i, denotes the total number of reconstructed points on the reconstructed surface, ; with denotes the equivalent sources to the sound source reconstruction surface the transfer matrix of the reconstruction point, size ; the transfer matrix the element in the row, the column is shown in equation (10): （10）； In formula (10), d represents the distance between the i-th equivalent source and the j-th reconstruction point on the sound source reconstruction surface. the distance between the i-th equivalent source and the j-th reconstruction point on the sound source reconstruction surface. delay time is: .
4. The method of locating a pulsed acoustic source in a strong harmonic background according to claim 1, wherein, The sound source reconstruction surface is an actual surface of the sound source or a set plane or curved surface located between the measurement array and the actual surface of the sound source to be measured; the set curved surface fully or partially surrounds the sound source to be measured.
5. The method of locating a pulsed acoustic source in a strong harmonic background according to claim 1, wherein, The pulse sound source to be measured contains a pulse signal, which is a transient signal generated by physical phenomena such as discharge, impact, explosion or leakage; the time-domain waveform of the transient signal is a non-periodic transient peak with a duration of 0.1 milliseconds to 200 milliseconds; and the transient signal shows short-time, vertical stripe and wide-band energy aggregation characteristics on a time-frequency distribution, which can be distinguished from long-time, narrow-band harmonic signals on a time-frequency distribution.