Motor electromagnetic noise source positioning estimation method and system

Abstract

The invention provides a motor electromagnetic noise source positioning estimation method and system, and relates to the technical field of motor electromagnetic noise analysis and array signal processing. Comprising the steps of identifying the characteristic frequency of a motor electromagnetic signal, filtering out an interfering motor electromagnetic noise narrow-band signal based on spatial clustering and an energy threshold technology, conveniently and accurately identifying noise source characteristics, optimizing a narrow-band signal division method, improving the adaptability of the method to the environment, and improving the robustness of the method. According to the method and the system, the dependence on the number of motor electromagnetic noise source signal samples and the signal-to-noise ratio is reduced, finally, motor electromagnetic noise source positioning estimation is carried out on the residual motor electromagnetic noise narrow-band signal vectors, and based on the process, the accuracy of motor electromagnetic noise source azimuth angle estimation is improved, and the accuracy of motor electromagnetic noise source azimuth angle estimation is improved. And the electromagnetic noise source of the motor can be accurately positioned in a complex industrial environment.

Description

Technical Field

[0001] The present invention relates to the technical fields of motor electromagnetic noise analysis and array signal processing, and more specifically, to a method and system for locating and estimating motor electromagnetic noise sources. Background Art

[0002] In modern industrial applications, permanent magnet motors are widely used due to their high efficiency and high performance. However, they may generate electromagnetic noise during operation. These noises not only affect the accuracy of sensors and the service life of equipment, but may also cause interference to the surrounding environment. Electromagnetic noise is usually closely related to the design, manufacturing quality, and operating state of the motor. By locating the noise source, potential problems in motor design or manufacturing can be discovered, so as to carry out targeted optimization to improve the performance and reliability of the motor. Moreover, abnormal changes in electromagnetic noise often indicate that the motor may have faults or performance degradation. Accurately locating the noise source helps to timely discover and handle potential problems, and avoid the occurrence or expansion of faults.

[0003] At present, the research on the location of permanent magnet motor electromagnetic noise sources mainly focuses on how to accurately identify and locate electromagnetic noise sources, which are mainly divided into the incoherent signal-subspace method (ISSM) and the coherent signal-subspace method (CSM). CSM relies on estimating the direction angle of the electromagnetic noise source and selecting the best focusing frequency point to construct a focusing matrix, which requires relatively accurate prior knowledge and is difficult to obtain in practical applications. Moreover, the accuracy of its matrix directly affects the accuracy of the estimation result, and its adaptability to non-ideal conditions in practical applications is poor, which limits its universal application. In contrast, the ISSM method divides the broadband signal of motor electromagnetic noise into multiple non-overlapping narrowband signals, estimates the direction of arrival for each frequency band, and then obtains the final direction of arrival result through averaging, and locates the motor electromagnetic noise source according to the direction of arrival result. Because of its fast operation speed, high resolution, and strong compatibility, it has become the preferred solution for diagnosing and locating motor electromagnetic noise sources. However, when the existing ISSM method processes motor electromagnetic noise signals, it often fails to fully consider the significant characteristics of the noise sources in the signals, resulting in the divided narrowband signals may not have these significant characteristics, affecting the accuracy of location estimation.

[0004] In addition, currently, there is also a Multiple Signal Classification (MUSIC) algorithm used in the field of array signal processing to estimate the direction or frequency of signal sources. It is a spectral estimation method for signal processing. Based on matrix eigen-space decomposition, it decomposes the covariance matrix of the received signal into a signal subspace and a noise subspace, and uses the orthogonality of these two subspaces to estimate the direction of the signal source (DOA estimation, Direction of Arrival). It has high resolution, high accuracy, and stability, but has a high computational complexity and is sensitive to the accurate estimation of the number of signal sources and the calibration of the array.

[0005] Therefore, researchers are exploring improved methods to obtain more effective narrowband signals. In addition, during the analysis of motor electromagnetic noise signals, the construction of effective narrowbands is also a research hotspot, including the sideband electromagnetic noise theory of permanent magnet synchronous motors based on multi-physics field coupling and the channel equalization method based on response estimation frequency domain fitting. These studies provide new perspectives and tools for the diagnosis of motor electromagnetic noise, but further research and improvement are still needed to improve the accuracy and robustness of the motor electromagnetic noise source localization estimation method. Summary of the Invention

[0006] To solve the problem of low accuracy in current motor electromagnetic noise source localization, this application proposes a method and system for motor electromagnetic noise source localization estimation. By extracting effective narrowband signals of electromagnetic noise based on clustering algorithms and energy threshold techniques, it improves the accuracy of azimuth angle estimation of motor electromagnetic noise sources, making it more conducive to accurately locating motor electromagnetic noise sources in complex industrial environments.

[0007] To achieve the above technical effects, the technical solution of the present invention is as follows:

[0008] In a first aspect, this application proposes a method for motor electromagnetic noise source localization estimation, including:

[0009] S1: Obtain motor electromagnetic noise signals;

[0010] S2: Identify the characteristic frequencies of the motor electromagnetic noise signals, construct a characteristic frequency sample set based on the characteristic frequencies, perform spatial clustering on the characteristic frequency sample set, and extract different center frequencies after clustering;

[0011] S3: Based on different center frequencies, reconstruct different motor electromagnetic noise narrowband signals, and calculate the energy of each motor electromagnetic noise narrowband signal;

[0012] S4: Filter out motor electromagnetic noise narrowband signals with energy lower than the energy threshold based on the energy threshold technique to obtain a remaining motor electromagnetic noise narrowband signal vector;

[0013] S5: Perform motor electromagnetic noise source localization estimation on the remaining motor electromagnetic noise narrowband signal vectors.

[0014] Preferably, a co-prime sensor array is used to obtain the motor electromagnetic noise signal.

[0015] Preferably, the co-prime sensor array is a co-prime array composed of two sub-arrays, including a first sub-array and a second sub-array. The first sub-array has M sensors, and the spacing between adjacent sensors is Nd. The second sub-array has N sensors, and the spacing between adjacent sensors is Md. Among them, M and N are co-prime positive integers, d is the minimum unit sensor spacing, and taking the rightmost sensor as the reference position, the sensor positions of the co-prime sensor array are expressed as:

[0016] S = {mNd - M(N - 1)d | 0 ≤ m ≤ M - 1} ∪ {nMd - M(N - 1)d | 0 ≤ n ≤ N - 1}

[0017] Among them, m represents the element order of the co-prime sensor array;

[0018] The total number of sensors is: |S| = M + N - 1;

[0019] Suppose there are K motor electromagnetic noise signals arriving at the co-prime sensor array from different directions θ = [θ 1 , θ 2 , …, θ K . T At time t, the motor electromagnetic noise signal model received by the co-prime array is:

[0020]

[0021] Among them, represents the steering vector, λ is the signal wavelength, A s = [a s (θ 1 ), a s (θ 2 ), …, a s (θ K )] represents the manifold matrix, s(t) represents the reference signal waveform, α = [α 1 , α 2 , …, α K T represents the non-zero complex-valued fading coefficient vector, n s (t) represents a random additive white noise vector that follows a complex Gaussian distribution and is uncorrelated with the signal, represents the noise power;

[0022] For the first sub-array or the second sub-array in the relatively prime sensor array, K motor electromagnetic noise signals are incident from different directions θ = [θ 1 , θ 2 , …, θ K T The motor electromagnetic noise signal x m (t) received by the m-th array element at time t is expressed as:

[0023]

[0024] where s i (t) represents the waveform of the i-th motor electromagnetic noise signal source incident on the array element m at time t, is the unit incident vector of the i-th signal, is the position vector of the m-th array element, and n m (t) is the additive Gaussian white noise superimposed on the signal of the m-th array element.

[0025] Preferably, the obtained motor electromagnetic noise signal in the time domain is converted to the frequency domain through Fourier transform; after the motor electromagnetic noise signal is converted from the time domain to the frequency domain, prior frequency data is obtained based on the generation principle of the motor electromagnetic noise, the actual frequency domain signal is compared and matched with the prior frequency data, the characteristic frequencies of the motor electromagnetic noise signals received by each array element are identified, and a characteristic frequency sample set is constructed;

[0026] Based on the DBSCAN spatial clustering algorithm, the characteristic frequency sample set is clustered, and the characteristic frequency sample set is divided into J clusters with arbitrary shapes, and the center frequencies f j of each cluster are extracted, where j = 1, 2, …, J.

[0027] Preferably, according to the center frequencies f j of each cluster, J motor electromagnetic noise narrowband signals are reconstructed; the boundary frequencies of the J narrowband components with f j as the center frequency, k as the bandwidth coefficient, and D as the bandwidth are expressed as:

[0028]

[0029] D = k · f j

[0030] The calculation expression for the energy of each motor electromagnetic noise narrowband signal is:

[0031]

[0032] where x m (f j ) represents the narrowband component with the center frequency f j in the m-th array element.​

[0033] Preferably, the energy threshold E η is expressed as:

[0034]

[0035] Based on the energy threshold technology, narrowband signals of motor electromagnetic noise with energy lower than the energy threshold are filtered out, and J effective narrowband signals with energy greater than E η are retained to obtain the remaining motor electromagnetic noise narrowband signal vector X(f j ):

[0036] X(f j 0 = A(f j , θ)S(f j ) + N(f j ), j = 1, 2, …, J

[0037] where A(f j , θ) and S(f j ) are the array manifold matrix and the signal source at the frequency point f j respectively, and N(f j ) is the noise signal at the frequency point f j .

[0038] Preferably, using the inverse Fourier transform, the remaining motor electromagnetic noise narrowband signal vector X(f j ) is converted from the frequency domain to the time domain signal x s (t). The process of performing motor electromagnetic noise source localization estimation on the remaining motor electromagnetic noise narrowband signal vector includes:

[0039] Calculating the covariance matrix, the expression is:

[0040]

[0041] where T represents the snapshot index;

[0042] Based on the improved Capon beamformer algorithm, a highly focused beam is created to enhance the detection of coherent signal sources, following the weight w constraint, the expression is:

[0043]

[0044] The steering vector formula of the improved Capon beamformer is:

[0045] [g(θ 1 )a s (θ 1 )g(θ 2 )a s (θ 2 )…g(θM+N-1 )a s (θ M+N-1 )] T

[0046] The weight constraint of the improved Capon beamformer algorithm is expressed as The constraint condition is w H a s (θ) = g(θ); The weight vector is obtained by the Lagrange multiplier method. The weight vector is used as a weighting factor to enhance the motor electromagnetic noise signal in a specific direction and suppress or interfere with it;

[0047]

[0048] The power spectrum is:

[0049]

[0050] Preferably, the process of performing motor electromagnetic noise source localization estimation on the remaining motor electromagnetic noise narrowband signal vector further includes: weighting the time-domain signal x s (t) with the weight vector w to enhance the signal in the desired direction. The new signal form is: y(t) = w H x s (t), and calculating the covariance matrix R y again. The expression is:

[0051]

[0052] where, () H represents the conjugate transpose; When the number of motor electromagnetic noise sources is uncertain, to reduce the resources consumed by eigenvalue decomposition and speed up the operation, let U s = R y , and take U s as the signal subspace and calculate the noise subspace. Then the expression of the noise subspace is:

[0053]

[0054] where, μ introduces a hyperparameter. When μ is small enough, the influence on the and U n orthogonality is negligible;

[0055] Estimate and process the noise subspace U n according to the root-MUSIC algorithm, and define the polynomial as:

[0056]

[0057] where, P(z) = [1 z … z M-1 ​T , let \(f(z) = 0\), and solve for \(k\) roots \(z\) close to the unit circle 1 , \(z\) 2 , …, \(z\) K , obtain the azimuth angle of the motor electromagnetic noise source, and locate the motor electromagnetic noise source.

[0058] In a second aspect, the present application also proposes a motor electromagnetic noise source location and estimation system, which is used to implement the motor electromagnetic noise source location and estimation method described above, and includes:

[0059] A noise signal acquisition unit that acquires motor electromagnetic noise signals;

[0060] A clustering unit that is used to identify the characteristic frequencies of the motor electromagnetic noise signals, construct a characteristic frequency sample set based on the characteristic frequencies, perform spatial clustering on the characteristic frequency sample set, and extract different center frequencies after clustering;

[0061] An energy calculation unit that reconstructs different motor electromagnetic noise narrowband signals based on different center frequencies and calculates the energy of each motor electromagnetic noise narrowband signal;

[0062] A filtering unit that filters out motor electromagnetic noise narrowband signals with energy lower than the energy threshold based on the energy threshold technology to obtain a remaining motor electromagnetic noise narrowband signal vector;

[0063] A location and estimation unit that performs motor electromagnetic noise source location and estimation on the remaining motor electromagnetic noise narrowband signal vector.

[0064] Preferably, the noise signal acquisition unit is a co-prime sensor array, and the co-prime sensor array is a co-prime array composed of two sub-arrays, including a first sub-array and a second sub-array. The first sub-array has \(M\) sensors, and the spacing between adjacent two sensors is \(Nd\). The second sub-array has \(N\) sensors, and the spacing between adjacent two sensors is \(Md\). Among them, \(M\) and \(N\) are co-prime positive integers, \(d\) is the minimum unit sensor spacing, and taking the rightmost sensor as the reference position, the sensor positions of the co-prime sensor array are expressed as:

[0065] \(S=\{mNd - M(N - 1)d|0\leq m\leq M - 1\}\cup\{nMd - M(N - 1)d|0\leq n\leq N - 1\}\)

[0066] where \(m\) represents the element order of the co-prime sensor array;

[0067] The total number of sensors is: \(|S| = M + N - 1\).

[0068] Compared with the prior art, the beneficial effects of the technical solution of the present invention are:

[0069] The present invention provides a method and system for locating and estimating the electromagnetic noise source of an electric motor. First, the electromagnetic noise source signal of the electric motor is acquired, and then the signal is preprocessed, including identifying the characteristic frequencies of the electromagnetic signals of the electric motor and filtering out the interfering narrowband electromagnetic noise signals of the electric motor based on spatial clustering and energy threshold techniques, which facilitates more accurate identification of the noise source characteristics. Moreover, the narrowband signal division method is optimized, improving the adaptability of the method proposed in the present invention to the environment and reducing the dependence on the number of samples and signal-to-noise ratio of the electromagnetic noise source signal of the electric motor. Finally, the remaining narrowband electromagnetic noise signal vectors of the electric motor are used for locating and estimating the electromagnetic noise source of the electric motor. Based on the above process, the method and system proposed in the present invention improve the accuracy of the azimuth angle estimation of the electromagnetic noise source of the electric motor, which is more conducive to accurately locating the electromagnetic noise source of the electric motor in a complex industrial environment. BRIEF DESCRIPTION OF THE DRAWINGS

[0070] Figure 1 It is a schematic flow chart showing the method for locating and estimating the electromagnetic noise source of the electric motor proposed in the embodiment of the present invention;

[0071] Figure 2 It is a distribution diagram showing the co-prime sensor array proposed in the embodiment of the present invention;

[0072] Figure 3 It is a schematic diagram showing the acquisition of the electromagnetic noise source signal of the electric motor by using the co-prime sensor array proposed in the embodiment of the present invention;

[0073] Figure 4 It is a process diagram showing the improved Capon beamformer algorithm proposed in the embodiment of the present invention;

[0074] Figure 5 It is a composition structure diagram showing the electromagnetic noise source location system of the electric motor proposed in the embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0075] The drawings are only for illustrative purposes and should not be construed as limitations on this patent;

[0076] For better illustration of this embodiment, some parts of the drawings are omitted, enlarged or reduced, which do not represent the actual size;

[0077] For those skilled in the art, it is understandable that some well-known content descriptions in the drawings may be omitted.

[0078] The technical solutions of the present invention will be further described below with reference to the drawings and embodiments.

[0079] The description of the positional relationship in the drawings is only for illustrative purposes and should not be construed as limitations on this patent;

[0080] Embodiment 1

[0081] This embodiment proposes a method for estimating the location of the motor electromagnetic noise source. The flowchart of this method is shown in Figure 1 , as Figure 1 shown, and it includes the following steps:

[0082] S1: Obtain the motor electromagnetic noise signal;

[0083] S2: Identify the characteristic frequencies of the motor electromagnetic noise signal, construct a characteristic frequency sample set based on the characteristic frequencies, perform spatial clustering on the characteristic frequency sample set, and extract the different center frequencies after clustering;

[0084] S3: Based on the different center frequencies, reconstruct different motor electromagnetic noise narrowband signals, and calculate the energy of each motor electromagnetic noise narrowband signal;

[0085] S4: Filter out the motor electromagnetic noise narrowband signals with energy lower than the energy threshold based on the energy threshold technology to obtain the remaining motor electromagnetic noise narrowband signal vector;

[0086] S5: Perform location estimation of the motor electromagnetic noise source on the remaining motor electromagnetic noise narrowband signal vector.

[0087] In this embodiment, first, the motor electromagnetic noise source signal is obtained, and then the signal is preprocessed, including identifying the characteristic frequencies of the electromagnetic signal and filtering out the interfering motor electromagnetic noise narrowband signals based on spatial clustering and energy threshold technology, which is convenient for more accurately identifying the characteristics of the noise source. By adopting advanced signal processing technologies, the accuracy of signal preprocessing and feature extraction is improved. These technologies can better identify and suppress noise, retain the key characteristics of the noise source, thereby improving the accuracy and reliability of the motor electromagnetic noise source signal location, reducing the errors introduced by the preprocessing steps, improving the accuracy of feature recognition, making the subsequent motor electromagnetic noise source signal location estimation more reliable, and providing a solid foundation for the accurate diagnosis of motor electromagnetic noise. Moreover, the narrowband signal division method is optimized, the adaptability to the environment is improved, and the dependence on the number of samples and the signal-to-noise ratio is reduced. Finally, the improved MUSIC-AP algorithm is adopted to improve the accuracy of the motor electromagnetic noise source azimuth estimation, which is more conducive to accurately locating the motor electromagnetic noise source in a complex industrial environment.

[0088] Embodiment 2

[0089] In this embodiment, a coprime sensor array is used to obtain the motor electromagnetic noise signal. The coprime sensor array proposed in this embodiment is shown in Figure 2 , as shown in Figure 2, the co-prime sensor array is a co-prime array composed of two sub-arrays, including a first sub-array and a second sub-array. The first sub-array has M sensors, and the spacing between adjacent sensors is Nd. The second sub-array has N sensors, and the spacing between adjacent sensors is Md. Among them, M and N are co-prime positive integers, and d is the minimum unit sensor spacing. Taking the rightmost sensor as the reference position, the sensor positions of the co-prime sensor array are expressed as:

[0090] S = {mNd - M(N - 1)d | 0 ≤ m ≤ M - 1} ∪ {nMd - M(N - 1)d | 0 ≤ n ≤ N - 1}

[0091] Among them, m represents the element order of the co-prime sensor array;

[0092] The total number of sensors is: |S| = M + N - 1;

[0093] In this embodiment, there are K motor electromagnetic noise signals arriving at the co-prime sensor array from different directions θ = [θ 1 , θ 2 , …, θ K . T At time t, the motor electromagnetic noise signal model received by the co-prime array is:

[0094]

[0095] Among them, represents the steering vector, λ is the signal wavelength, A s = [a s (θ 1 ), a s (θ 2 ), …, a s (θ K )] represents the manifold matrix, s(t) represents the reference signal waveform, α = [α 1 , α 2 , …, α K T represents the non-zero complex-valued fading coefficient vector, n s (t) represents a random additive white noise vector that follows a complex Gaussian distribution and is uncorrelated with the signal, represents the noise power;

[0096] For the first sub-array or the second sub-array in the co-prime sensor array, as Figure 3 shown, taking the first sub-array as an example, assuming that K motor electromagnetic noise signals ( Figure 3 only shows one of them) come from different directions θ = [θ 1 , θ 2,…,θ K T Incident, the motor electromagnetic noise signal x received by the m-th array element at time t m is expressed as:

[0097]

[0098] where s i (t) represents the waveform of the i-th motor electromagnetic noise signal source incident on the array element m at time t, is the unit incident vector of the i-th signal, is the position vector of the m-th array element, n m (t) is the additive Gaussian white noise superimposed on the signal of the m-th array element.

[0099] In this embodiment, the obtained motor electromagnetic noise signal in the time domain is converted to the frequency domain through Fourier transform. The Fourier transform (FFT) plays a key role in this step. It converts the time-domain signal into a frequency-domain signal, enabling the clear presentation of the distribution of the signal in the frequency domain. In the frequency domain, the characteristic frequencies can be observed and analyzed more intuitively. Combining with the prior frequency data, the characteristic frequencies related to the electromagnetic noise source can be extracted more accurately, and at the same time, it provides the necessary frequency-domain data basis for subsequent frequency-based processing (such as narrowband reconstruction, etc.).

[0100] After the motor electromagnetic noise signal is converted from the time domain to the frequency domain, based on the principle of motor electromagnetic noise generation, prior frequency data is obtained. The actual frequency-domain signal is compared and matched with the prior frequency data to identify the characteristic frequencies of the motor electromagnetic noise signals received by each array element, and a characteristic frequency sample set is constructed. In this embodiment, these prior frequency data are obtained based on the principle of motor electromagnetic noise generation. For example, the characteristic frequency f = 2k 1 f o (k 1 = 1, 2, 3…, f o is the motor operating frequency), the characteristic frequency f = cf k ± df c (c and d are positive integers with the same parity, f k is the inverter switching frequency) and the characteristic frequency f = uf e / p ± kf k ± f o (u is the correlation coefficient, f e is the rotational frequency of dynamic eccentricity), etc. By comparing and matching the actual frequency-domain signal with these theoretical characteristic frequencies, a sample set G is constructed, and these characteristic frequencies are important identifiers of the motor electromagnetic noise source.

[0101] ​Then, based on the DBSCAN spatial clustering algorithm, the feature frequency sample set is clustered. This algorithm divides the data set into several clusters of arbitrary shapes according to density. In this embodiment, the feature frequency sample set is divided into J clusters of arbitrary shapes, and the central frequency f of each cluster is extracted j , j = 1, 2, …, J. It can effectively identify the core points, boundary points, and noise points in the set, thereby detecting and processing outliers. In the diagnosis of electromagnetic noise sources, it can handle the situation where the signal feature frequencies of each array element are different in the measured environment. In the frequency-domain signal, the feature frequency usually corresponds to the frequency component with a larger amplitude in the spectrum. These feature frequencies can be found by searching for the peaks in the spectrum. For several feature frequencies, the amplitude and phase information of each feature frequency need to be extracted respectively. After extracting several feature frequencies, the central frequency can be obtained by calculating the arithmetic mean of these feature frequencies. That is, taking the amplitude of each feature frequency as the weight, the feature frequencies are weighted averaged, and the result is the central frequency. This method can ensure that the central frequency can accurately reflect the overall distribution of the signal in the spectrum. In this embodiment, through the above steps, D = k·f is accurately extracted j The central frequency f of each cluster j (j = 1, 2, …, J).

[0102] In this embodiment, according to the central frequency f of each cluster j , J narrowband signals of motor electromagnetic noise are reconstructed; specifically, a band-pass filter can be designed according to the central frequency and the expected bandwidth. The purpose of the band-pass filter is to extract the narrowband signal components corresponding to the central frequency from the original signal. The design of the filter needs to consider parameters such as the type of filter (such as Butterworth filter, Chebyshev filter, etc.), the order, and the cut-off frequency. Then, the original signal is passed through the designed band-pass filter to obtain the filtered narrowband signal. The filtering process will remove the signal components that are not related to the central frequency and retain the required narrowband signal. Finally, in some cases, further reconstruction processing may be required for the filtered narrowband signal. This may be because the filtering process introduces distortion or noise, or because it is necessary to adjust parameters such as the amplitude and phase of the signal to meet specific requirements. The methods of signal reconstruction may include techniques such as interpolation, extrapolation, and denoising.

[0103] With f j as the central frequency, k as the bandwidth coefficient (in this embodiment, k = 0.1), and D as the bandwidth, the boundary frequencies of the J narrowband components are expressed as:

[0104]

[0105] D = k·f j

[0106] The calculation expression for the energy of each narrowband signal of the motor electromagnetic noise is as follows:

[0107]

[0108] Among them, x m (f j ) represents the narrowband component with a center frequency of f j in the m-th array element.

[0109] Preferably, the energy threshold E η is expressed as:

[0110]

[0111] Based on the energy threshold technology, filter out the narrowband signals of the motor electromagnetic noise with energy lower than the energy threshold, and retain J effective narrowband signals with energy greater than E η to obtain the remaining motor electromagnetic noise narrowband signal vector X(f j ):

[0112] X(f j 0 = A(f j , θ)S(f j ) + N(f j ), j = 1, 2, …, J

[0113] Among them, A(f j , θ) and S(f j ) are the array manifold matrix and the signal source at the frequency point f j respectively, and N(f j ) is the noise signal at the frequency point f j . Through the above reconstruction steps, a noise narrowband signal with more significant frequency characteristics can be obtained to overcome the adverse effects of frequency dispersion and environmental interference on DOA azimuth estimation.

[0114] Through this series of operations, from the original electromagnetic noise signal, after processing such as clustering and narrowband reconstruction, the original broadband noise signal can be reconstructed into a narrowband signal with a narrower bandwidth and more distinct frequency characteristics. These narrowband signals have more distinct frequency characteristics, reducing the influence of factors such as environmental interference in DOA estimation.

[0115] Then, using the inverse Fourier transform, convert the remaining motor electromagnetic noise narrowband signal vector X9f j ) from the frequency domain to the time domain signal x s (t). The process of performing motor electromagnetic noise source localization estimation on the remaining motor electromagnetic noise narrowband signal vector includes:

[0116] Calculate the covariance matrix, and the expression is:

[0117]

[0118] Among them, T represents the snapshot index;

[0119] Based on the improved Capon beamformer algorithm, a highly focused beam is created to enhance the detection of coherent signal sources, so as to improve the signal-to-noise ratio and spatial resolution. Following the weight w constraint, the expression is:

[0120]

[0121] The steering vector formula of the improved Capon beamformer is:

[0122] [g(θ 1 )a s (θ 1 )g(θ 2 )a s (θ 2 )…g(θ M+N-1 )a s (θ M+N-1 )] T

[0123] In the co-prime sensor array, the array gain g(θ i ) has a specific form when i = 1, 2, …, M + N - 1, which will significantly affect the value of a(θ i ), and since the gain pattern directly affects the response of the array in the "observation direction", the weight constraint of the improved Capon beamformer algorithm is expressed as The constraint condition is w H a s (θ) = g(θ); as Figure 4 shown, the weight vector is obtained by the Lagrange multiplier method, and the weight vector, as a weighting factor, enhances the motor electromagnetic noise signal in a specific direction and suppresses or interferes with it;

[0124]

[0125] The power spectrum is:

[0126]

[0127] The process of performing motor electromagnetic noise source localization estimation on the remaining motor electromagnetic noise narrowband signal vector also includes:

[0128] Weight the time-domain signal x s (t) with the weight vector w to enhance the signal in the desired direction, and the new signal form is: y(t) = w H x s (t), and calculate the covariance matrix R y again, and the expression is:

[0129]

[0130] wherein, () H represents conjugate transpose; in the case of uncertainty about the number of motor electromagnetic noise sources, to reduce the resources consumed by eigenvalue decomposition and speed up the operation, let U s = R y , and take U s as the signal subspace, calculate the noise subspace, then the expression of the noise subspace is:

[0131]

[0132] wherein, μ introduces a hyperparameter. When μ is small enough, the influence on the and U n orthogonality is negligible;

[0133] According to the root-MUSIC algorithm, estimate and process the noise subspace U n , and define the polynomial as:

[0134]

[0135] wherein, P(z) = [1 z... z M-1 T , let f(z) = 0, and solve for k roots z 1 , z 2 , …, z K close to the unit circle, and obtain the azimuth angles of the motor electromagnetic noise sources to locate the motor electromagnetic noise sources. Here, the roots of the polynomial f(z) are complex numbers z that satisfy f(z) = 0. In the complex plane, these roots can be represented as points. The unit circle is a circle in the complex plane centered at the origin with a radius of 1. Its equation is |z| = 1. The unit circle has a special position in complex analysis because it is closely related to problems such as the roots of polynomials and stability analysis. The statement that the roots are close to the unit circle means that the roots may lie on the unit circle or be very close to the boundary of the unit circle. The roots on the unit circle are used to measure the stability of the sought azimuth angles to accurately locate the motor electromagnetic noise sources.

[0136] Through the above steps, the estimation of the direction of arrival of coherent signals in a co-prime sensor array demonstrates unique advantages and effectiveness, providing strong technical support for signal processing in related fields. The entire process starts from signal acquisition, goes through preprocessing, characteristic frequency extraction, and finally narrowband signal reconstruction and clustering, forming a complete signal processing flow in the motor electromagnetic noise diagnosis system, which can effectively process motor electromagnetic noise signals and provide a reliable data basis for the analysis and location of noise sources.

[0137] Embodiment 3 ​

[0138] See Figure 5 , this embodiment proposes a system for locating the electromagnetic noise source of a motor, which is used to implement the method for estimating the location of the electromagnetic noise source of the motor, and includes:

[0139] A noise signal acquisition unit that acquires the electromagnetic noise signal of the motor;

[0140] A clustering unit, which is used to identify the characteristic frequencies of the electromagnetic noise signals of the motor, construct a set of characteristic frequency samples based on the characteristic frequencies, perform spatial clustering on the set of characteristic frequency samples, and extract different center frequencies after clustering;

[0141] An energy calculation unit, which reconstructs different narrowband electromagnetic noise signals of the motor based on different center frequencies and calculates the energy of each narrowband electromagnetic noise signal of the motor;

[0142] A filtering unit, which filters out the narrowband electromagnetic noise signals of the motor with energy lower than the energy threshold based on the energy threshold technology to obtain a vector of remaining narrowband electromagnetic noise signals of the motor;

[0143] A location estimation unit that estimates the location of the electromagnetic noise source of the motor for the vector of remaining narrowband electromagnetic noise signals of the motor.

[0144] In this embodiment, the noise signal acquisition unit is a co-prime sensor array, and the co-prime sensor array is a co-prime array composed of two sub-arrays, including a first sub-array and a second sub-array. The first sub-array has M sensors, and the distance between adjacent two sensors is Nd. The second sub-array has N sensors, and the distance between adjacent two sensors is Md. Among them, M and N are co-prime positive integers, d is the minimum unit sensor spacing, and taking the rightmost sensor as the reference position, the sensor positions of the co-prime sensor array are expressed as:

[0145] S = {mNd - M(N - 1)d|0 ≤ m ≤ M - 1} ∪ {nMd - M(N - 1)d|0 ≤ n ≤ N - 1}

[0146] Among them, m represents the element order of the co-prime sensor array;

[0147] The total number of sensors is: |S| = M + N - 1.

[0148] The embodiments are only examples given to clearly illustrate the present invention and are not limitations on the implementation manners of the present invention. For those of ordinary skill in the art, other different forms of changes or modifications can be made based on the above description. It is not necessary and impossible to list all the implementation manners here. Any modifications, equivalent replacements, and improvements made within the spirit and principle of the present invention shall be included in the protection scope of the claims of the present invention.

Claims

1. A method for locating and estimating a motor electromagnetic noise source, characterized in that: include: S1: Obtain the electromagnetic noise signal of the motor; S2: Identify the characteristic frequency of the motor electromagnetic noise signal, construct a characteristic frequency sample set based on the characteristic frequency, spatially cluster the characteristic frequency sample set, and extract different center frequencies after clustering; S3: Based on different center frequencies, different motor electromagnetic noise narrowband signals are reconstructed, and the energy of each motor electromagnetic noise narrowband signal is calculated; S4: filtering out the motor electromagnetic noise narrowband signal with energy lower than the energy threshold based on the energy threshold technology, and obtaining the remaining motor electromagnetic noise narrowband signal vector; S5: Estimate the location of the motor electromagnetic noise source based on the remaining motor electromagnetic noise narrowband signal vector.

2. The method for locating and estimating the electromagnetic noise source of a motor according to claim 1, characterized in that: The electromagnetic noise signal of the motor is obtained by using a mutual prime sensor array.

3. The method for locating and estimating the electromagnetic noise source of a motor according to claim 2, characterized in that: The mutually prime sensor array is a coprime array composed of two subarrays, including a first subarray and a second subarray, wherein the first subarray has M sensors, and the spacing between two adjacent sensors is Nd, and the second subarray has N sensors, and the spacing between two adjacent sensors is Md, wherein M and N are mutually prime positive integers, d is the minimum unit sensor spacing, and the rightmost sensor is used as the reference position. The sensor position of the mutually prime sensor array is expressed as: S={mNd-M(N-1)d|0≤m≤M-1}∪{nMd-M(N-1)d|0≤n≤N-1}, where m represents the order of the elements of the coprime sensor array; The total number of sensors is: |S|=M+N-1; Suppose K motor electromagnetic noise signals come from different directions θ=[θ1,θ2,…,θ K ] T Arriving at the mutually prime sensor array, at time t, the motor electromagnetic noise signal model received by the mutually prime array is: in, represents the steering vector, λ is the signal wavelength, A s =[a s (θ1),a s (θ2),…,a s (θ K )] represents the manifold matrix, s(t) represents the reference signal waveform, α=[α1,α2,…,α K ] T represents the non-zero complex-valued fading coefficient vector, n s (t) indicates that it follows a complex Gaussian distribution and a random additive white noise vector that is uncorrelated with the signal, represents the noise power; For the first subarray or the second subarray in the mutual prime sensor array, K motor electromagnetic noise signals are transmitted from different directions θ=[θ1,θ2,…,θ K ] T Incident, the motor electromagnetic noise signal x received by the mth array element at time t m (t) is expressed as: Among them, s i (t) represents the waveform of the electromagnetic noise signal source of the i-th motor incident on the array element m at time t, is the unit incident vector of the ith signal, is the position vector of the Mth array element, n m (t) is the additive white Gaussian noise superimposed on the signal of the mth array element.

4. The method for locating and estimating the electromagnetic noise source of a motor according to claim 3, characterized in that: The acquired time domain electromagnetic noise signal of the motor is converted to the frequency domain through Fourier transform; after the motor electromagnetic noise signal is converted from the time domain to the frequency domain, the prior frequency data is obtained based on the principle of motor electromagnetic noise generation, the actual frequency domain signal is compared and matched with the prior frequency data, the characteristic frequency of the motor electromagnetic noise signal received by each array element is identified, and a characteristic frequency sample set is constructed; Based on the DBSCAN spatial clustering algorithm, the characteristic frequency sample set is clustered, the characteristic frequency sample set is divided into J clusters of arbitrary shapes, and the central frequency f of each cluster is extracted. j , j=1,2,…,J.

5. The method for locating and estimating the electromagnetic noise source of a motor according to claim 4, characterized in that: According to the center frequency f of each cluster j , reconstruct J motor electromagnetic noise narrowband signals; with f j The boundary frequencies of J narrowband components with center frequency, k as bandwidth coefficient and D as bandwidth are expressed as: D=k·f j The calculation expression of the energy of each motor electromagnetic noise narrowband signal is: Among them, x m (f j ) indicates that the center frequency of the mth array element is f j narrowband component.

6. The method for locating and estimating the electromagnetic noise source of a motor according to claim 5, characterized in that: The energy threshold E η It is expressed as: Based on the energy threshold technology, the motor electromagnetic noise narrowband signal with energy lower than the energy threshold is filtered out, and J signals with energy greater than E are retained. η The effective narrowband signal of the residual motor electromagnetic noise narrowband signal vector X(f j ): X(f j )=A(f j ,θ)S(f j )+N(f j ),j=1,2,…,J Among them, A(f j ,θ) and S(f j ) are the frequency points f j The array flow matrix and the signal source at N(f j ) is the frequency f j The noise signal at .

7. The method for locating and estimating the electromagnetic noise source of a motor according to claim 6, characterized in that: Using inverse Fourier transform, the residual motor electromagnetic noise narrowband signal vector X(f j ) is converted from frequency domain to time domain signal x s (t), the process of locating and estimating the motor electromagnetic noise source based on the remaining motor electromagnetic noise narrowband signal vector includes: Calculate the covariance matrix, the expression is: Where T represents the snapshot index; Based on the improved Capon beamformer algorithm, a highly focused beam is created to enhance the detection of coherent signal sources, subject to the weight w constraint, expressed as: The steering vector formula of the improved Capon beamformer is: [g(θ1)a s (θ1)g(θ2)a s (θ2)…g(θ M+N-1 )a s (i M+N-1 )] T The constraints of the improved Capon beamformer algorithm are expressed as The constraint condition is w H a s (θ) = g(θ); a weight vector is obtained by Lagrange multiplier method, and the weight vector is used as a weighting factor to enhance the electromagnetic noise signal of the motor in a specific direction and suppress or interfere; The power spectrum is:

8. The method for locating and estimating the electromagnetic noise source of a motor according to claim 7, characterized in that: The process of locating and estimating the motor electromagnetic noise source based on the remaining motor electromagnetic noise narrowband signal vector also includes: The time domain signal x is conditioned on the weight vector w. s (t) is weighted to enhance the signal in the desired direction. The new signal is in the form of: y(t) = w H x s (t), calculate the covariance matrix R again y , the expression is: in,() H represents the conjugate transpose; in order to reduce the resources consumed by eigenvalue decomposition and speed up the calculation when the number of motor electromagnetic noise sources is uncertain, let U s =R y , will U s As the signal subspace, calculate the noise subspace, then the noise subspace expression is: Among them, μ introduces a hyperparameter. When μ is small enough, and U n The influence of orthogonality is ignored; according to the root-MUSIC algorithm, the noise subspace U n Perform estimation processing and define the polynomial as: Where P(z)=[1 z… z M-1 ] T , let f(z) = 0, and solve for the k roots z1,z2,…,z close to the unit circle K , obtain the azimuth of the motor electromagnetic noise source and locate the motor electromagnetic noise source.

9. A motor electromagnetic noise source location estimation system, characterized in that: The system is used to implement the method for locating and estimating the electromagnetic noise source of a motor according to any one of claims 1 to 8, comprising: A noise signal acquisition unit, for acquiring an electromagnetic noise signal of the motor; A clustering unit is used to identify the characteristic frequency of the motor electromagnetic noise signal, construct a characteristic frequency sample set based on the characteristic frequency, spatially cluster the characteristic frequency sample set, and extract different center frequencies after clustering; An energy calculation unit reconstructs different motor electromagnetic noise narrowband signals based on different center frequencies and calculates the energy of each motor electromagnetic noise narrowband signal; A filtering unit, which filters out the motor electromagnetic noise narrowband signal whose energy is lower than the energy threshold based on the energy threshold technology, and obtains the remaining motor electromagnetic noise narrowband signal vector; The positioning estimation unit performs motor electromagnetic noise source positioning estimation on the remaining motor electromagnetic noise narrowband signal vector.

10. The motor electromagnetic noise source location estimation system according to claim 9, characterized in that: The noise signal acquisition unit is a mutually prime sensor array, which is a coprime array composed of two subarrays, including a first subarray and a second subarray, wherein the first subarray has M sensors, and the spacing between two adjacent sensors is Nd, and the second subarray has N sensors, and the spacing between two adjacent sensors is Md, wherein M and N are mutually prime positive integers, d is the minimum unit sensor spacing, and the rightmost sensor is used as the reference position. The sensor position of the mutually prime sensor array is expressed as: S={mNd-M(N-1)d|0≤m≤M-1}∪{nMd-M(N-1)d|0≤n≤N-1} Wherein, m represents the order of the elements of the mutually prime sensor array; the total number of sensors is: |S|=M+N-1.

Images