A method for locating equipment fault sound sources based on sparse imaging technology

Abstract

The present invention discloses a method for locating the sound source of equipment faults based on sparse imaging technology, comprising the following steps: S1, collecting sound source signals through a microphone array MEMS sensor; S2, based on differential smoothing constraints, introducing a shift-invariant point spread function as a reconstruction dictionary to reconstruct the sound power map obtained by the delay-sum algorithm; S3, based on differential smoothing constraints, introducing a shift-invariant point spread function as a reconstruction dictionary to reconstruct the sound power map obtained by the delay-sum algorithm; S4, based on the differential smoothing constraint result, introducing a differential sparse constraint to further optimize and reconstruct the sound power map obtained by the delay-sum algorithm; S5, drawing an imaging cloud map of the optimized and reconstructed sound power map, and fusing and superimposing the imaging cloud map with the image collected by the camera, thereby realizing fault location. The present invention can ensure that the amount of calculation is fully reduced while obtaining a high-resolution and accurate reconstruction result, and solves the problems existing in the prior art.

Description

Technical Field

[0001] The present invention relates to the technical field of sound source localization, and in particular to a method for localizing the sound source of equipment failure based on sparse imaging technology. Background Art

[0002] The sound source localization method is a commonly used equipment fault location method in the power industry and industrial industry. The principle of this method is to estimate the wave direction of all sound sources in the scene relative to the microphone array by analyzing the signal collected by the microphone array. In recent years, some methods have emerged for sound source localization using the sparse characteristics of signals. They use the sparsity of signals in the time-frequency domain to transform the multi-source localization problem into a single-source localization problem, breaking through the limitations of traditional sound source localization methods and solving the sound source localization problem in underdetermined scenarios, that is, scenarios where the number of sound sources is greater than the number of microphones. However, these sound source localization methods usually produce low spatial resolution when performing delayed sum algorithms and beamforming imaging, and spurious sources will appear. Therefore, the positioning results based on beamforming are relatively rough and the accuracy needs to be further improved. Based on this, we provide a method for equipment fault sound source localization based on sparse imaging technology. Summary of the invention

[0003] The present invention provides a method for locating the sound source of equipment faults based on sparse imaging technology, the main purpose of which is to solve the problems existing in the prior art.

[0004] The present invention adopts the following technical solution:

[0005] A method for locating equipment fault sound sources based on sparse imaging technology comprises the following steps:

[0006] S1, collect sound source signals through microphone array MEMS sensor;

[0007] S2, performing Fourier transform on the collected sound source signal, and using a delay-sum algorithm to obtain a sound power diagram;

[0008] S3, based on Differential smoothing constraints are used to introduce a shift-invariant point spread function as a reconstruction dictionary to reconstruct the sound power map obtained by the delay-sum algorithm:

[0009]

[0010] Where: X represents the optimized sound power map, B represents the sound power map obtained by the delay-sum algorithm, P represents the point spread function, D represents the first-order difference matrix, λ1 represents the smoothing parameter, represents the Fourier transform, represents the inverse Fourier transform, represents the Hadamard product, represents a non-negative constraint;

[0011] S4, based on Difference smoothing constraint results, introducing The difference sparse constraint further optimizes the reconstruction of the sound power map obtained by the delay-sum algorithm:

[0012]

[0013] Where: represents sparse parameters;

[0014] S5. Draw an imaging cloud map with the optimized and reconstructed sound power map, and fuse and superimpose the imaging cloud map with the image collected by the camera to achieve fault location.

[0015] Furthermore, in step S3, the fast iterative threshold shrinkage algorithm framework is used to solve formula (1), thereby obtaining iterative solution formulas (1.1), (1.2) and (1.3) of formula (1):

[0016]

[0017]

[0018]

[0019] Where: L1 represents the Lipschitz constant, t1 represents the step size parameter, and Y1 represents the auxiliary variable.

[0020] Furthermore, for the iterative solution of equations (1.1), (1.2) and (1.3), first use The differentiability of the norm solves formula (1.1) and defines:

[0021]

[0022] You can get:

[0023]

[0024] Then, by iterating (1.1), (1.2) and (1.3) until convergence, the reconstructed sound power map is obtained.

[0025] Further, in step S4, the framework of the alternating direction multiplier method is adopted to solve formula (2) by splitting operators and alternating optimization variables, thereby obtaining iterative solution formulas (2.1), (2.2) and (2.3) of formula (2):

[0026]

[0027]

[0028]

[0029] Where: Z represents the auxiliary variable, p represents the penalty parameter, and U represents the dual variable after scale transformation.

[0030] Furthermore, for the solution process of formula (2.1), we can use The norm can be solved by differentiability, that is,

[0031]

[0032] You can get:

[0033]

[0034]

[0035] Furthermore, for the solution process of formula (2.2), since formula (2.2) is The LASSO problem under regularization constraints is therefore solved using a fast iterative threshold shrinkage algorithm:

[0036]

[0037] Furthermore, by iterating (2.1), (2.2) and (2.3) until convergence, the optimized reconstructed sound power map is obtained.

[0038] Compared with the prior art, the present invention has the following beneficial effects:

[0039] The present invention optimizes the algorithm flow of the prior art. First, based on Differential smoothing constraints are used to introduce a shift-invariant point spread function as a reconstruction dictionary to reconstruct the sound power map obtained by the delay-sum algorithm; then based on Difference smoothing constraint results, introducing The differential sparse constraint further optimizes the reconstruction of the sound power map obtained by the delay-sum algorithm, thereby ensuring that the high-resolution reconstruction result is obtained while fully reducing the amount of calculation, cleverly making full use of Difference smoothing constraint calculation fast sum The advantage of accurate differential sparse constraint calculation solves the problems existing in the prior art. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 It is the algorithm flow chart of the present invention. DETAILED DESCRIPTION

[0041] The specific embodiments of the present invention will be described below with reference to the accompanying drawings.

[0042] Reference Figure 1 , a method for locating equipment fault sound sources based on sparse imaging technology, comprising the following steps:

[0043] S1. Acquiring sound source signals through a microphone array MEMS sensor. Specifically, this step collects multi-channel sound signals by moving a handheld microphone array MEMS sensor.

[0044] S2. Perform Fourier transform on the collected sound source signal, and use the delay-sum algorithm to obtain the sound power map. The Fourier transform and delay-sum algorithm can be implemented by conventional algorithms, which are not limited here. Specifically, the algorithm described in the patent application number 202010916621.X can be used.

[0045] S3, based on Differential smoothing constraints are used to introduce a shift-invariant point spread function as a reconstruction dictionary to reconstruct the sound power map obtained by the delay-sum algorithm:

[0046]

[0047] Where: X represents the optimized sound power map, B represents the sound power map obtained by the delay-sum algorithm, P represents the point spread function, D represents the first-order difference matrix, λ1 represents the smoothing parameter, represents the Fourier transform, represents the inverse Fourier transform, represents the Hadamard product, represents a non-negative constraint.

[0048] Generally speaking, for an ideal beam imaging diagram, the acoustic power between two adjacent grids should be similar, and the differential smoothing constraint is a powerful tool to constrain the power variation of adjacent grids. Considering the non-negativity and smoothness of the acoustic power diagram, the present invention introduces The difference smoothing constraint is used to reconstruct the sound power map obtained by the delay-sum algorithm (DAS), thereby improving the resolution and positioning accuracy of the sound source localization. Based on the difference smoothness constraint, the shift-invariant point spread function (PSF) is introduced as the reconstruction dictionary in order to effectively utilize the additional information to accurately locate the sound source.

[0049] In step S3, the fast iterative threshold shrinkage algorithm (FISTA) framework is used to solve formula (1), thereby obtaining iterative solution formulas (1.1), (1.2) and (1.3) of formula (1):

[0050]

[0051]

[0052]

[0053] Where: L1 represents the Lipschitz constant, t1 represents the step size parameter, and Y1 represents the auxiliary variable.

[0054] For formula (1.1), we can use The norm is differentiable to solve it, and the definition

[0055]

[0056] You can get:

[0057]

[0058] Formulas (1.2) and (1.3) are simple iterative equations, so by iterating (1.1), (1.2) and (1.3) until convergence, the reconstructed high-resolution sound power map can be obtained.

[0059] S4, based on Difference smoothing constraint results, introducing The difference sparse constraint further optimizes the reconstruction of the sound power map obtained by the delay-sum algorithm:

[0060]

[0061] Where: Represents a sparse parameter.

[0062] Considering the non-negativity and sparsity of the sound power map, we introduce Difference constraint The sound power map after differential smoothing constraint is optimized and reconstructed, so that full use can be made of Fast computation of difference smoothing constraints The difference sparse constraint has the advantage of accurate calculation, thus reducing the amount of calculation and speeding up the algorithm while maintaining high-resolution reconstruction results.

[0063] In step S4, the framework of the alternating direction method of multipliers (ADMM) is adopted to solve formula (2) by splitting operators and alternating optimization variables, thereby obtaining iterative solution formulas (2.1), (2.2) and (2.3) of formula (2):

[0064]

[0065]

[0066]

[0067] Where: Z represents the auxiliary variable, p represents the penalty parameter, and U represents the dual variable after scale transformation.

[0068] For formula (2.1), we can use The norm is differentiable to solve it, and the definition

[0069]

[0070] You can get:

[0071]

[0072]

[0073] For formula (2.2), after analysis, it can be seen that it is The LASSO problem under regularization constraints can be solved by the fast iterative threshold shrinkage algorithm (FISTA). The specific solution formula is:

[0074]

[0075] Where: represents the soft threshold function, which is defined as:

[0076]

[0077] The above definition of the soft threshold function is interpreted element-wise.

[0078] Formula (2.3) is a simple iterative equation, so by iterating (2.1), (2.2) and (2.3) until convergence, the optimized reconstructed sound power map can be obtained.

[0079] It can be seen that by solving the sparse reconstruction problem of the sound power map through methods such as ADMM and FISTA, high-resolution and accurate reconstruction of the sound power map can be achieved.

[0080] S5, draw an imaging cloud map with the optimized and reconstructed sound power map, and fuse and superimpose the imaging cloud map with the image collected by the camera, so as to realize fault location. The present invention can realize equipment power-off diagnosis, spacing portability and safety through long-distance contactless measurement.

[0081] The above is only a specific implementation of the present invention, but the design concept of the present invention is not limited thereto. Any non-substantial changes to the present invention using this concept shall be deemed as an infringement of the protection scope of the present invention.

Claims

1. A method for locating equipment fault sound sources based on sparse imaging technology, characterized in that: The steps include: S1, collect sound source signals through microphone array MEMS sensor; S2, performing Fourier transform on the collected sound source signal, and using a delay-sum algorithm to obtain a sound power diagram; S3, based on Differential smoothing constraints are used to introduce a shift-invariant point spread function as a reconstruction dictionary to reconstruct the sound power map obtained by the delay-sum algorithm: ， Where: X represents the optimized sound power map, B represents the sound power map obtained by the delay-sum algorithm, P represents the point spread function, D represents the first-order difference matrix, λ1 represents the smoothing parameter, represents the Fourier transform, represents the inverse Fourier transform, represents the Hadamard product, represents a non-negative constraint; S4, based on Difference smoothing constraint results, introducing The difference sparse constraint further optimizes the reconstruction of the sound power map obtained by the delay-sum algorithm: ， Where: represents sparse parameters; S5. Draw an imaging cloud map with the optimized and reconstructed sound power map, and fuse and superimpose the imaging cloud map with the image collected by the camera to achieve fault location.

2. The method for locating equipment fault sound sources based on sparse imaging technology according to claim 1, characterized in that: In step S3, the fast iterative threshold shrinkage algorithm framework is used to solve formula (1), thereby obtaining iterative solution formulas (1.1), (1.2) and (1.3) of formula (1): ， ， ， Where: L1 represents the Lipschitz constant, t1 represents the step size parameter, and Y1 represents the auxiliary variable.

3. A method for locating equipment fault sound sources based on sparse imaging technology as claimed in claim 2, characterized in that: For the iterative solution method of formulas (1.1), (1.2) and (1.3), first use The differentiability of the norm solves formula (1.1) and defines: ， You can get: ， Then, by iterating (1.1), (1.2) and (1.3) until convergence, the reconstructed sound power map is obtained.

4. The method for locating equipment fault sound sources based on sparse imaging technology according to claim 1, characterized in that: In step S4, the framework of the alternating direction multiplier method is adopted to solve formula (2) by splitting operators and alternating optimization variables, thereby obtaining iterative solution formulas (2.1), (2.2) and (2.3) of formula (2): ， ， ， Where: Z represents the auxiliary variable, represents the penalty parameter, and U represents the scale-transformed dual variable.

5. The method for locating the sound source of equipment faults based on sparse imaging technology according to claim 4, characterized in that: For the solution process of formula (2.1), we can use The norm can be solved by differentiability, that is, ， You can get: 。 6. The method for locating the sound source of equipment faults based on sparse imaging technology according to claim 5, characterized in that: For the solution process of formula (2.2), since formula (2.2) is The LASSO problem under regularization constraints is therefore solved using a fast iterative threshold shrinkage algorithm: 。 7. The method for locating equipment fault sound sources based on sparse imaging technology according to claim 6, characterized in that: By iterating (2.1), (2.2) and (2.3) until convergence, the optimized reconstructed sound power map is obtained.

Images