Stator assembly detection method, device, equipment and storage medium

Abstract

The present invention discloses a stator assembly detection method, device, equipment and storage medium, the method comprising: configuring a single-point excitation, multi-point collaborative excitation and sweep frequency excitation mode to a preset acoustic detection loop, and exciting each mode in order through orthogonal sequence planning to obtain a complete acoustic response signal; performing signal preprocessing on the acoustic response signal in the order of amplitude normalization, baseline drift correction and noise elimination, and extracting the acoustic wave propagation delay, attenuation coefficient and spectrum energy characteristics to obtain an acoustic characteristic spectrum; dividing the acoustic characteristic spectrum into a plurality of circumferential analysis units according to the stator slot position, calculating the correlation coefficient and spectrum similarity between adjacent units, and obtaining characteristic deviation data; comparing the characteristic deviation data with a preset evaluation standard, determining the air gap state level and uneven type of each circumferential analysis unit, and obtaining an air gap uniformity detection result. The technical solution of the present invention realizes non-destructive detection of the air gap uniformity between the winding and the iron core.

Description

Technical Field

[0001] The present invention relates to the technical field of assembled stator detection, and in particular to a stator assembly detection method, device, equipment and storage medium. Background Art

[0002] The stator is an important component in generators and motors, mainly composed of stator core, stator winding, insulation materials, etc. The stator core is made of multiple layers of silicon steel sheets, and is provided with several slots evenly distributed along the circumference inside, which are used to install the stator winding. The stator assembly process mainly includes steps such as winding embedding, lead connection and insulation treatment, among which the most critical is to accurately embed the winding into the stator slot. During the embedding process, it is necessary to ensure that an appropriate air gap is maintained between the winding and the slot wall to achieve good electromagnetic performance and heat dissipation.

[0003] After the stator is assembled, the air gap between the winding and the stator slot wall may be uneven due to human factors, material properties and process limitations during the assembly process. This unevenness can cause local overheating, increased vibration and reduced efficiency during motor operation. At present, the methods for detecting air gap uniformity mainly include visual inspection, mechanical probe measurement and electrical characteristics testing. Among them, visual inspection is limited by the depth and viewing angle of the slot, making it difficult to achieve comprehensive inspection; although mechanical probe measurement has high accuracy, it is easy to damage the insulation layer and has low inspection efficiency; although the electrical characteristics test method is non-destructive, it is difficult to accurately locate the uneven position and is easily affected by the external environment. Therefore, there is an urgent need for an air gap uniformity detection method that can achieve both non-destructive detection and accurate positioning. Summary of the invention

[0004] The main purpose of the present invention is to solve the technical problem that the existing detection method cannot realize non-destructive detection of the uniformity of the air gap between the winding and the iron core during the stator assembly process.

[0005] A first aspect of the present invention provides a stator assembly detection method, the stator assembly detection method comprising:

[0006] Input an acoustic wave excitation sequence into the preset acoustic detection loop, including configuring a single-point excitation mode, a multi-point collaborative excitation mode, and a sweep frequency excitation mode. Through orthogonal sequence planning, each mode is excited in sequence to obtain a complete acoustic response signal.

[0007] The acoustic response signal is preprocessed in the order of amplitude normalization, baseline drift correction and noise elimination, and the acoustic wave propagation delay, attenuation coefficient and spectrum energy characteristics are extracted to obtain an acoustic characteristic spectrum;

[0008] Dividing the acoustic characteristic spectrum into a plurality of circumferential analysis units according to the stator slot positions, calculating the correlation coefficient and spectrum similarity between adjacent units, and obtaining characteristic deviation data;

[0009] The characteristic deviation data is compared with a preset evaluation standard to determine the air gap state level and uneven type of each circumferential analysis unit, and obtain an air gap uniformity detection result.

[0010] Optionally, the arrangement process of the preset acoustic detection circuit includes:

[0011] Dividing the stator into a plurality of detection units, each detection unit corresponds to a stator slot, and each of the detection units includes an ultrasonic acoustic transducer and two acoustic sensors;

[0012] The plurality of detection units are evenly arranged along the outer circumference of the stator, so that the ultrasonic acoustic transducer of each detection unit is located at the top of the slot of the corresponding stator slot;

[0013] The two acoustic sensors in each of the detection units are respectively installed on both sides of the corresponding stator slots to receive ultrasonic signals reflected by the windings.

[0014] Optionally, the configuration of a single-point excitation mode, a multi-point collaborative excitation mode, and a sweep frequency excitation mode, and exciting each mode in sequence through orthogonal sequence planning to obtain a complete acoustic response signal, includes:

[0015] Based on a preset first orthogonal excitation sequence, the ultrasonic acoustic transducer is controlled to work alone at a preset time interval, and a single-point response signal corresponding to the acoustic sensor is synchronously collected;

[0016] Based on a preset second orthogonal excitation sequence, the adjacent ultrasonic acoustic transducers are controlled to work in combination according to a preset phase difference to generate a spatially directional acoustic beam, and spatial response signals of all the acoustic sensors are collected;

[0017] Based on a preset third orthogonal excitation sequence, the ultrasonic acoustic transducers are controlled to work synchronously according to a linear frequency modulation rule to generate a broadband excitation signal, and the frequency response signals of all the acoustic sensors are collected;

[0018] The collected single-point response signal, spatial response signal and frequency response signal are synthesized in time sequence through orthogonal sequence planning, and an anti-aliasing buffer is set between adjacent response signals to obtain the complete acoustic response signal.

[0019] Optionally, the acoustic characteristic spectrum is divided into a plurality of circumferential analysis units according to the stator slot positions, and the correlation coefficient and spectrum similarity between adjacent units are calculated to obtain characteristic deviation data, including:

[0020] The acoustic characteristic spectrum is converted into an analytical signal through Hilbert transform, and the instantaneous amplitude and instantaneous phase characteristics corresponding to each stator slot are extracted;

[0021] Based on the instantaneous amplitude and instantaneous phase characteristics, a circumferential acoustic characteristic matrix is ​​constructed, and the circumferential acoustic characteristic matrix is ​​divided into a plurality of circumferential analysis units according to the circumferential distribution of the stator slots;

[0022] Calculating the time domain correlation coefficient between the instantaneous features of adjacent circumferential analysis units, and performing spectrum analysis on the instantaneous features to obtain spectrum similarity;

[0023] Adaptive threshold processing is performed based on the time domain correlation coefficient and the frequency spectrum similarity to obtain the characteristic deviation data.

[0024] Optionally, performing spectrum analysis on the instantaneous features to obtain spectrum similarity includes:

[0025] Performing wavelet packet multi-layer decomposition on the instantaneous characteristics of the adjacent circumferential analysis units to obtain wavelet packet coefficients of different frequency bands;

[0026] The optimal decomposition layer number of each frequency band is determined based on the minimum Shannon entropy criterion to obtain the optimal decomposition frequency band;

[0027] The energy value of each frequency band is calculated according to the wavelet packet coefficient corresponding to the optimal decomposition frequency band, and the relative energy value of each frequency band is obtained by normalization processing;

[0028] The similarity of the relative energy values ​​of each frequency band between adjacent circumferential analysis units is calculated to obtain the frequency spectrum similarity.

[0029] Optionally, comparing the characteristic deviation data with a preset evaluation criterion to determine the air gap state level and uneven type of each circumferential analysis unit includes:

[0030] Constructing a multidimensional feature space, mapping the feature deviation data into a feature vector, wherein the feature vector includes a time domain correlation coefficient component and a frequency spectrum similarity component;

[0031] Calculate the degree of deviation between the characteristic vector and a preset standard characteristic vector based on the Mahalanobis distance to obtain a quantitative evaluation value of the air gap state;

[0032] According to the distribution law of the quantitative evaluation value of the air gap state, the air gap state is divided into different levels by using a fuzzy clustering method;

[0033] Based on the distribution pattern of the feature vector in the multi-dimensional feature space, the type of air gap non-uniformity is determined.

[0034] A second aspect of the present invention provides a stator assembly detection device, the stator assembly detection device comprising:

[0035] The excitation control module is used to input the acoustic wave excitation sequence into the preset acoustic detection loop, including configuring the single-point excitation mode, the multi-point collaborative excitation mode and the sweep frequency excitation mode, and to excite each mode in sequence through orthogonal sequence planning to obtain a complete acoustic response signal;

[0036] A signal processing module, used to perform signal preprocessing on the acoustic response signal in the order of amplitude normalization, baseline drift correction and noise elimination, and extract the acoustic wave propagation delay, attenuation coefficient and spectrum energy characteristics to obtain an acoustic characteristic spectrum;

[0037] A feature analysis module, used to divide the acoustic feature spectrum into a plurality of circumferential analysis units according to the stator slot positions, calculate the correlation coefficient and spectrum similarity between adjacent units, and obtain feature deviation data;

[0038] The state determination module is used to compare the characteristic deviation data with a preset evaluation standard, determine the air gap state level and uneven type of each circumferential analysis unit, and obtain an air gap uniformity detection result.

[0039] The third aspect of the present invention provides a stator assembly detection device, comprising: a memory and at least one processor, wherein instructions are stored in the memory, and the memory and the at least one processor are interconnected via lines; the at least one processor calls the instructions in the memory so that the stator assembly detection device performs the steps of the above-mentioned stator assembly detection method.

[0040] A fourth aspect of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores instructions, which, when executed on a computer, enable the computer to execute the steps of the above-mentioned stator assembly detection method.

[0041] The technical solution provided by the embodiments of the present application has at least the following advantages:

[0042] By setting a preset acoustic detection loop on the periphery of the stator and adopting three modes of acoustic wave excitation sequences, namely single-point excitation, multi-point collaborative excitation and swept frequency excitation, a full range of detection of the air gap state in the stator slot is achieved. Among them, the single-point excitation mode is used to obtain local precise information, the multi-point collaborative excitation mode enhances the detection sensitivity through spatial directional sound beams, and the swept frequency excitation mode provides broadband response characteristics. The three modes complement each other and overcome the defects of existing visual detection methods that are limited by slot depth and viewing angle. At the same time, through orthogonal sequence planning, each mode is excited in order, effectively avoiding mutual interference between the modes.

[0043] Secondly, the collected acoustic response signals are systematically processed, and the signal-to-noise ratio and reliability are improved through the preprocessing process of amplitude normalization, baseline drift correction and noise elimination. By extracting the acoustic wave propagation delay, attenuation coefficient and spectrum energy characteristics, a multi-dimensional acoustic characteristic spectrum is formed, which provides rich characteristic information for the accurate evaluation of the air gap state. This non-destructive testing method avoids the damage to the insulation layer that may be caused by mechanical probe measurement.

[0044] Furthermore, the acoustic characteristic spectrum is divided into multiple circumferential analysis units according to the stator slot position. By calculating the correlation coefficient and spectrum similarity between adjacent units, the characteristic deviation data is obtained, and the quantitative evaluation of the air gap uniformity is realized. This method based on the comparison of adjacent units overcomes the problem that the existing electrical characteristics test is difficult to accurately locate. Finally, by comparing the characteristic deviation data with the preset evaluation criteria, not only the air gap state level can be determined, but also the type of unevenness can be identified, providing a clear guiding basis for subsequent quality improvement. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the structures shown in these drawings without paying creative work.

[0046] Figure 1 A schematic diagram of an embodiment of a stator assembly detection method in an embodiment of the present invention;

[0047] Figure 2 A schematic diagram of an embodiment of a stator assembly detection device in an embodiment of the present invention;

[0048] Figure 3 A schematic diagram of an embodiment of a stator assembly detection device in an embodiment of the present invention;

[0049] Figure 4 It is a schematic diagram of the structure of the preset acoustic detection circuit and stator.

[0050] Description of Figure Numbers:

[0051] 1. Detection unit; 11. Ultrasonic acoustic transducer; 12. Acoustic sensor; 2. Stator slot.

[0052] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0053] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0054] It should be noted that if the embodiments of the present invention involve directional indications (such as up, down, left, right, front, back...), the directional indications are only used to explain the relative position relationship, movement status, etc. between the components under a certain specific posture (as shown in the accompanying drawings). If the specific posture changes, the directional indication will also change accordingly.

[0055] In addition, the descriptions of "first", "second", etc. in the present invention are only used for descriptive purposes and cannot be understood as indicating or implying their relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In addition, "and / or" in the full text includes three solutions. Taking A and / or B as an example, it includes technical solution A, technical solution B, and technical solution that satisfies both A and B. In addition, the technical solutions between the various embodiments can be combined with each other, which must be based on the ability of ordinary technicians in the field to implement. When the combination of technical solutions is contradictory or cannot be implemented, it should be considered that such combination of technical solutions does not exist and is not within the scope of protection required by the present invention.

[0056] An embodiment of the present application provides a stator assembly detection method. Figure 1 A flow chart of a stator assembly detection method provided in an embodiment of the present application. In this embodiment, the method includes:

[0057] See also Figure 1 , input the acoustic wave excitation sequence into the preset acoustic detection loop, including configuring the single-point excitation mode, the multi-point collaborative excitation mode and the sweep frequency excitation mode, and through orthogonal sequence planning, each mode is excited in sequence to obtain a complete acoustic response signal;

[0058] In one embodiment of the present invention, see Figure 2 The arrangement process of the preset acoustic detection circuit includes: dividing the stator into multiple detection units, each detection unit corresponds to a stator slot, and each detection unit includes an ultrasonic acoustic transducer and two acoustic sensors; arranging the multiple detection units evenly along the outer circumference of the stator, so that the ultrasonic acoustic transducer of each detection unit is located at the top of the slot of the corresponding stator slot; installing the two acoustic sensors in each detection unit on both sides of the corresponding stator slot, respectively, to receive the ultrasonic signal reflected by the winding.

[0059] Specifically, the detection units need to be divided according to the structural characteristics of the stator. For example, if the stator has 18 slots, it is divided into 18 detection units, and each detection unit is responsible for the detection of one stator slot. Each detection unit contains an ultrasonic acoustic transducer and two acoustic sensors. This configuration can achieve comprehensive detection of a single slot. The ultrasonic acoustic transducer is made of piezoelectric ceramic material, with an operating frequency of 20-100kHz and a sound wave emission area of ​​10mm×10mm to ensure detection accuracy and avoid interference between adjacent slots.

[0060] During the arrangement of the detection units, 18 detection units are evenly arranged every 20 degrees along the outer circumference of the stator, so that the ultrasonic acoustic transducer of each detection unit is directly opposite the top of the corresponding stator slot. Taking a stator with a slot depth of 30mm as an example, the ultrasonic acoustic transducer maintains a fixed distance of 5mm from the slot. The setting of this distance takes into account the near-field effect of the ultrasonic wave and the divergence angle of the sound wave, ensuring that the sound wave can effectively propagate to the winding in the slot. The two acoustic sensors in each detection unit are installed on the left and right sides of the corresponding stator slot, respectively, at an angle of 30 degrees to the slot. This arrangement allows the sound waves to be effectively received by the acoustic sensors on both sides after being reflected by the winding and the slot wall, thereby obtaining acoustic information reflecting the air gap state.

[0061] The entire detection unit is fixed by a positioning device, which includes an outer ring with mounting holes that match the detection unit. The advantage of this arrangement is that it realizes independent detection of each slot, avoids signal interference between adjacent slots, and can effectively determine the uneven state of the air gap through the differential signal of the two acoustic sensors. For example, when the winding of a slot is biased, the reflected signals received by the sensors on both sides will show obvious asymmetric characteristics. By analyzing this asymmetry, the degree and direction of the uneven air gap can be determined.

[0062] In addition to the above-mentioned detection unit arrangement, the acoustic detection circuit can also be arranged in layers, that is, the ultrasonic acoustic transducer and the acoustic sensor are arranged on detection rings at different heights to form a three-dimensional detection network. Another arrangement is to use a mobile detection unit, which is driven by a stepper motor to move around the stator to achieve scanning detection of each slot. In addition, a flexible detection ring arrangement can also be used. The detection unit is installed on a flexible support structure, and the shape of the detection ring is adjusted by a pneumatic or hydraulic mechanism to achieve adaptive detection of stators of different sizes.

[0063] In one embodiment of the present invention, the configuration of a single-point excitation mode, a multi-point collaborative excitation mode and a swept frequency excitation mode, and exciting each mode in sequence through orthogonal sequence planning to obtain a complete acoustic response signal, includes: based on a preset first orthogonal excitation sequence, controlling the ultrasonic acoustic transducer to work alone according to a preset time interval, and synchronously collecting the single-point response signal corresponding to the acoustic sensor; based on a preset second orthogonal excitation sequence, controlling the adjacent ultrasonic acoustic transducers to work in combination according to a preset phase difference to generate a spatial directional sound beam, and collecting the spatial response signals of all the acoustic sensors; based on a preset third orthogonal excitation sequence, controlling the ultrasonic acoustic transducers to work synchronously according to a linear frequency modulation rule to generate a wideband excitation signal, and collecting the frequency response signals of all the acoustic sensors; synthesizing the collected single-point response signal, spatial response signal and frequency response signal according to the time sequence through orthogonal sequence planning, and setting an anti-aliasing buffer between adjacent response signals to obtain the complete acoustic response signal.

[0064] Specifically, this link adopts a combination of three orthogonal excitation sequences to obtain comprehensive detection information. First, in the first orthogonal excitation sequence, each ultrasonic acoustic transducer (hereinafter described as a transducer) is controlled in turn to work individually at preset time intervals. Taking the 18-slot stator as an example, starting from transducer No. 1, each transducer works for 100 microseconds, and the working interval between adjacent transducers is 200 microseconds. Such a time setting ensures the integrity of signal acquisition and avoids interference from adjacent signals. When a transducer is working, the acoustic sensors on both sides of the corresponding slot synchronously collect the reflected signal to obtain a single-point response signal. This single-point excitation mode can clearly reflect the local characteristics of each slot and is suitable for discovering local air gap anomalies.

[0065] The second orthogonal excitation sequence adopts the mode of working in combination of adjacent transducers to form a directional sound beam by controlling the phase difference between adjacent transducers. In specific implementation, every three adjacent transducers constitute a working group, and the phase of the excitation signal of the transducers in the group is delayed by 120 degrees in sequence. For example, when the excitation signal of transducer No. 1 is 0 degree phase, transducer No. 2 is 120 degree phase, and transducer No. 3 is 240 degree phase. This phase difference setting enables the sound waves generated by the three transducers to form a directional superposition in space, thereby enhancing the energy of the sound waves in a specific direction. All acoustic sensors work simultaneously to collect the response signals generated by this spatial directional sound beam propagating in the stator. The advantage of this mode is that it improves the penetration ability of sound waves and can better detect the state of deep air gaps.

[0066] The third orthogonal excitation sequence uses linear frequency modulation to control all ultrasonic acoustic transducers to work synchronously. In the specific implementation, the frequency of the excitation signal is linearly scanned from 20kHz to 100kHz, and the scanning time is 500 microseconds. This broadband excitation method enables sound waves of different frequencies to excite different response characteristics of the stator structure and air gap. For example, sound waves in the low-frequency band (20-40kHz) mainly reflect the overall structural characteristics, the mid-frequency band (40-70kHz) can effectively detect the air gap state, and the high-frequency band (70-100kHz) is more sensitive to tiny defects.

[0067] After completing the signal acquisition of the three modes, these signals need to be synthesized. First, the single-point response signal, spatial response signal and frequency response signal are arranged according to the acquisition timing, and a 50-microsecond buffer is set between each two response signals. This buffer can effectively prevent aliasing between different types of signals. Then, the three signals are optimized and combined through the orthogonal sequence planning algorithm. For example, when an abnormality is detected in the single-point response signal of a certain slot, the spatial response signal and frequency response signal of the position are called for cross-validation to obtain a more reliable detection result. The final synthesized complete acoustic response signal contains local detail information, retains spatial distribution characteristics, and has rich characteristics in the frequency domain, providing comprehensive data support for subsequent signal analysis and air gap status judgment. This multi-mode combined excitation method gives full play to the advantages of various detection modes and significantly improves the accuracy and reliability of detection.

[0068] Please continue reading Figure 1 , performing signal preprocessing on the acoustic response signal in the order of amplitude normalization, baseline drift correction and noise elimination, and extracting the acoustic wave propagation delay, attenuation coefficient and spectrum energy characteristics to obtain an acoustic characteristic spectrum;

[0069] Specifically, the preprocessing of the acoustic response signal first performs amplitude normalization. In the specific operation, first find the maximum absolute value in the original signal sequence Then, each sample value in the original signal sequence is divided by the maximum absolute value. For example, if the original signal sequence is {-500mV, 200mV, 800mV, -300mV}, where the maximum absolute value is 800mV, the normalized signal sequence is {-0.625, 0.25, 1, -0.375}. This processing method ensures that all signal values ​​are mapped to the [-1, 1] interval, which is convenient for subsequent processing and comparison.

[0070] The baseline drift correction process is implemented by the sliding window method. The window length is set to 1000 sampling points (corresponding to a time of about 1ms), and the window moves 100 sampling points each time (the step length is 0.1ms). The average value of the signal is calculated in each window, and these average values ​​are interpolated through cubic spline to obtain a smooth baseline curve. Then the baseline curve is subtracted from the original signal to complete the baseline drift correction. This method can adapt to slowly changing baseline drifts, such as signal offsets caused by temperature changes.

[0071] The noise is eliminated by using the wavelet threshold denoising method. The specific steps are as follows: First, the signal is decomposed into 4 layers using the db4 wavelet basis function to obtain wavelet coefficients of different scales. Then the threshold is calculated based on the standard deviation of each layer of wavelet coefficients. The threshold calculation formula is: , where σ is the standard deviation of noise and N is the signal length. Soft threshold processing is performed on coefficients less than the threshold (i.e., these coefficients are set to 0), while coefficients greater than the threshold remain unchanged. Finally, the processed wavelet coefficients are reconstructed to obtain the denoised signal.

[0072] In the feature extraction stage, the cross-correlation method is used to calculate the acoustic wave propagation delay. The transmitted signal x(t) and the received signal y(t) are cross-correlated, and the calculation formula is: ,in, To transmit the signal, To receive the signal, is the time-delayed variable. Find the cross-correlation function The time delay τ corresponding to the maximum value is the sound wave propagation delay. The attenuation coefficient is obtained by calculating the frequency domain amplitude ratio of the received signal and the transmitted signal. Specifically, the two signals are fast Fourier transformed and the amplitude ratio is calculated at the main frequency point. The spectrum energy feature is to perform a 512-point fast Fourier transform on the signal, and then calculate the energy proportion in the three frequency bands of 20-40kHz, 40-70kHz and 70-100kHz. The energy of each frequency band is calculated by the sum of the squares of the amplitudes of each frequency point in the band.

[0073] Please continue reading Figure 1 , dividing the acoustic characteristic spectrum into a plurality of circumferential analysis units according to the stator slot positions, calculating the correlation coefficient and spectrum similarity between adjacent units, and obtaining characteristic deviation data;

[0074] In one embodiment of the present invention, the acoustic feature spectrum is divided into multiple circumferential analysis units according to the stator slot positions, the correlation coefficient and the spectrum similarity between adjacent units are calculated, and the feature deviation data is obtained, including: converting the acoustic feature spectrum into an analytical signal through Hilbert transform, and extracting the instantaneous amplitude and instantaneous phase characteristics corresponding to each stator slot position; constructing a circumferential acoustic feature matrix based on the instantaneous amplitude and instantaneous phase characteristics, and dividing the circumferential acoustic feature matrix into multiple circumferential analysis units according to the circumferential distribution of the stator slots; calculating the time domain correlation coefficient between the instantaneous characteristics of adjacent circumferential analysis units, and performing spectrum analysis on the instantaneous characteristics to obtain spectrum similarity; performing adaptive threshold processing based on the time domain correlation coefficient and spectrum similarity to obtain the feature deviation data.

[0075] Specifically, after obtaining the acoustic characteristic spectrum, it is first necessary to convert it into an analytical signal through Hilbert transform. Perform Hilbert transform to obtain its conjugate component ,Right now ,in, represents the Hilbert transform operator, and then the original signal As the real part, the transformed signal As the imaginary part, construct the complex analytical signal . Where j represents an imaginary unit. Taking the 18-slot stator as an example, assuming that the acoustic feature time domain sequence collected from a certain slot is {1.2, 0.8, -0.5, -0.9, 1.5}, the conjugate sequence obtained after Hilbert transform is {0.6, 1.1, 0.3, -1.2, 0.7}, thus obtaining the analytical signal of the slot. By processing the analytical signal, the instantaneous amplitude feature is extracted and instantaneous phase characteristics The advantage of this transformation method is that it can obtain the amplitude and phase information of the signal at the same time, providing a more comprehensive feature description for subsequent analysis.

[0076] Based on the acquired instantaneous amplitude and phase features, a circumferential acoustic feature matrix is ​​constructed. For an 18-slot stator, an 18×2 matrix is ​​constructed, where each row of the matrix corresponds to a slot and the two columns store the instantaneous amplitude and phase features respectively. For example, the matrix element M(i, 1) stores the instantaneous amplitude mean of the i-th slot, and M(i, 2) stores the instantaneous phase mean of the i-th slot. Considering the circumferential symmetry of the stator, the matrix is ​​divided into a circumferential analysis unit according to every three adjacent slots, forming a total of 6 analysis units. This grouping method takes into account the continuity of local features while maintaining an appropriate analysis scale.

[0077] For the instantaneous characteristics of adjacent circumferential analysis units, the time domain correlation coefficient and spectrum similarity are calculated. The time domain correlation coefficient is calculated by the Pearson correlation coefficient. Assuming that the instantaneous amplitude sequences of two adjacent units are X and Y respectively, the correlation coefficient is:

[0078] ;

[0079] Among them, cov(X, Y) represents the covariance of sequences X and Y, and Represent the standard deviation of sequence X and Y respectively, and the value range of correlation coefficient r is [-1,1]. When r is close to 1, it means that the two sequences are highly positively correlated; when r is close to -1, it means that the two sequences are highly negatively correlated; when r is close to 0, it means that the correlation between the two sequences is weak. In the present invention, the closer the r value is to 1, the more similar the instantaneous characteristics of adjacent units are, and the more uniform the air gap state is.

[0080] For example, when the amplitude sequence of the first unit is {1.2, 1.3, 1.1} and the amplitude sequence of the second unit is {1.25, 1.28, 1.15}, the calculated correlation coefficient is close to 1, indicating that the features of the two units are highly similar. For the calculation of spectral similarity, firstly, the instantaneous features are fast Fourier transformed to obtain the spectrum F(ω), and then the cosine similarity of the adjacent unit spectra is calculated. The calculation of spectral similarity not only considers the main frequency component, but also includes the energy distribution characteristics of the sidebands.

[0081] Finally, the adaptive threshold method is used to process the time domain correlation coefficient and spectrum similarity to obtain the feature deviation data. The adaptive threshold is calculated based on the OTSU algorithm, which determines the optimal threshold by maximizing the inter-class variance. For example, when the calculated correlation coefficient is distributed in the interval [0.8, 0.99], the OTSU algorithm may obtain 0.92 as the threshold, and the correlation coefficient below this threshold is marked as abnormal. For each circumferential analysis unit, the two indicators of time domain correlation coefficient and spectrum similarity are considered at the same time, and the final feature deviation data is obtained by weighted averaging. This adaptive threshold processing method overcomes the disadvantage that fixed thresholds are difficult to adapt to different working conditions, and improves the adaptability and accuracy of detection.

[0082] In one embodiment of the present invention, the spectral analysis of the instantaneous characteristics to obtain spectral similarity includes: performing multi-layer wavelet packet decomposition on the instantaneous characteristics of the adjacent circumferential analysis units to obtain wavelet packet coefficients of different frequency bands; determining the optimal number of decomposition layers for each frequency band based on the Shannon entropy minimum criterion to obtain the optimal decomposition frequency band; calculating the energy value of each frequency band according to the wavelet packet coefficient corresponding to the optimal decomposition frequency band, and normalizing it to obtain the relative energy value of each frequency band; calculating the similarity of the relative energy values ​​of each frequency band between adjacent circumferential analysis units to obtain the spectral similarity.

[0083] Specifically, wavelet packet decomposition of the instantaneous characteristics of adjacent circumferential analysis units is a key step in obtaining spectral similarity. First, the db4 wavelet basis is used to perform wavelet packet decomposition on the instantaneous feature sequence, and the maximum number of decomposition layers is initially set to 5. Taking the instantaneous feature signal with a sampling rate of 1024Hz as an example, the first layer of decomposition divides the signal into two frequency bands of 0-512Hz and 512-1024Hz, and the second layer divides the signal into four frequency bands of 0-256Hz, 256-512Hz, 512-768Hz and 768-1024Hz. By analogy, the fifth layer will obtain 32 frequency bands. This multi-layer decomposition method can perform more refined time-frequency analysis of the signal, and has better time-frequency localization characteristics than the traditional Fourier transform.

[0084] When determining the optimal number of decomposition layers, the Shannon entropy minimum criterion is used. The Shannon entropy is calculated for each frequency band of each decomposition layer, and the calculation formula is: ,in is the energy proportion of the i-th wavelet packet coefficient. For example, if the wavelet packet coefficient sequence of a certain frequency band is {0.5, 0.3, 0.2}, then =0.5, =0.3, =0.2, and substitute it into the formula to calculate the Shannon entropy of the frequency band. Compare the total entropy values ​​of different decomposition layers, and select the layer with the smallest entropy value as the optimal decomposition layer. This selection method based on information entropy can find the best balance between time and frequency resolution, ensuring the effectiveness of feature extraction.

[0085] After obtaining the optimal decomposition layer, calculate the energy value of each frequency band. Assuming the optimal decomposition layer is the third layer, the signal is divided into 8 frequency bands, and the wavelet packet coefficient sequence of each frequency band is , calculate the energy value The calculated energy value of each frequency band is divided by the total energy to obtain the normalized relative energy value. This normalization process eliminates the influence of signal intensity differences and makes the frequency band energies of different detection units comparable.

[0086] Finally, the similarity of the relative energy values ​​of each frequency band between adjacent circumferential analysis units is calculated. The cosine similarity formula is used to calculate the similarity of the corresponding frequency bands of two units. Assuming that the relative energy value vectors of a frequency band of the two units are A and B respectively, the similarity is:

[0087] .

[0088] The final spectrum similarity is obtained by averaging the similarities of all frequency bands. If the similarity is close to 1, it indicates that the spectrum characteristics of the two units are highly similar; if the similarity is significantly lower than 1, it indicates that there is an anomaly. This spectrum similarity calculation method based on wavelet packet decomposition not only considers the local characteristics of the signal in the time domain, but also retains the global information in the frequency domain, and can accurately reflect the differences between adjacent units.

[0089] Please continue reading Figure 1 , the characteristic deviation data is compared with the preset evaluation criteria, the air gap state level and uneven type of each circumferential analysis unit are determined, and the air gap uniformity detection result is obtained.

[0090] In one embodiment of the present invention, the characteristic deviation data is compared with a preset evaluation criterion to determine the air gap state level and uneven type of each circumferential analysis unit, including: constructing a multidimensional feature space, mapping the characteristic deviation data into a feature vector, the feature vector including a time domain correlation coefficient component and a spectrum similarity component; calculating the degree of deviation between the feature vector and a preset standard feature vector based on the Mahalanobis distance to obtain a quantitative evaluation value of the air gap state; according to the distribution law of the quantitative evaluation value of the air gap state, using a fuzzy clustering method to divide the air gap state into different levels; based on the distribution pattern of the feature vector in the multidimensional feature space, determining the type of air gap unevenness.

[0091] Specifically, the evaluation of the air gap state first requires the construction of a multidimensional feature space. For each detection unit, the time domain correlation coefficient and the spectrum similarity are combined to form a two-dimensional feature vector. Taking the 18-slot stator as an example, a feature vector is generated between two adjacent detection units. For example, when the time domain correlation coefficient is 0.95 and the spectrum similarity is 0.92, a feature vector [0.95, 0.92] is formed. For the complete stator, a total of 18 such feature vectors are formed to form a complete feature space. This multidimensional feature representation method combines the information in the time domain and the frequency domain, and can fully reflect the characteristics of the air gap state.

[0092] In the feature space, the Mahalanobis distance is used to calculate the degree of deviation between the feature vector and the preset standard feature vector. The standard feature vector comes from the statistical analysis of normal stator samples. For example, by measuring 100 qualified stators, the standard feature vector [0.98, 0.96] is obtained. The calculation of the Mahalanobis distance takes into account the correlation between features. The calculation formula is: , where x is the eigenvector to be tested, μ is the standard eigenvector (i.e. the mean vector of normal samples), Σ is the covariance matrix, Represents the inverse matrix of the covariance matrix. When the calculated Mahalanobis distance D value is small (such as D<2), it indicates that the air gap state is close to the standard state; when the D value is large, it indicates that there is a significant deviation. This evaluation method based on Mahalanobis distance overcomes the disadvantage of Euclidean distance that does not consider feature correlation and provides a more accurate deviation measurement.

[0093] For the calculated quantitative evaluation value of the air gap state, the fuzzy C-means clustering method is used to divide the state level. First, the number of clusters is set to 3, corresponding to the three levels of excellent, qualified and unqualified. Each evaluation value will get a membership vector, indicating the degree to which it belongs to each level. For example, the membership vector of a certain evaluation value is [0.8, 0.15, 0.05], indicating that the air gap state at this location is mainly excellent. This fuzzy clustering method avoids the rigidity of hard threshold division and can better reflect the continuous change characteristics of the air gap state.

[0094] Finally, the specific type of air gap unevenness is determined by analyzing the distribution pattern of the feature vector in multidimensional space. Three basic distribution patterns are defined: when the feature vector shows local aggregation and deviates from the standard vector, it is determined as local unevenness; when the feature vector shows periodic changes, it is determined as periodic unevenness; when the feature vector shows random distribution, it is determined as random unevenness. By matching with the pre-established pattern library, the specific unevenness type is obtained. This determination method based on distribution pattern not only gives a qualitative description of the air gap state, but also provides clear direction guidance for subsequent quality improvement.

[0095] The above describes the stator assembly detection method in the embodiment of the present invention. The following describes the stator assembly detection device in the embodiment of the present invention. Figure 2 In one embodiment of the present invention, a stator assembly detection device includes:

[0096] The excitation control module 201 is used to input an acoustic wave excitation sequence into a preset acoustic detection loop, including configuring a single-point excitation mode, a multi-point collaborative excitation mode, and a sweep frequency excitation mode, and to excite each mode in sequence through orthogonal sequence planning to obtain a complete acoustic response signal;

[0097] The signal processing module 202 is used to perform signal preprocessing on the acoustic response signal in the order of amplitude normalization, baseline drift correction and noise elimination, and extract the acoustic wave propagation delay, attenuation coefficient and spectrum energy characteristics to obtain an acoustic characteristic spectrum;

[0098] A feature analysis module 203 is used to divide the acoustic feature spectrum into a plurality of circumferential analysis units according to the stator slot positions, calculate the correlation coefficient and spectrum similarity between adjacent units, and obtain feature deviation data;

[0099] The state determination module 204 is used to compare the characteristic deviation data with a preset evaluation standard, determine the air gap state level and uneven type of each circumferential analysis unit, and obtain an air gap uniformity detection result.

[0100] above Figure 2 The stator assembly detection device in the embodiment of the present invention is described in detail from the perspective of modular functional entities. The stator assembly detection device in the embodiment of the present invention is described in detail from the perspective of hardware processing.

[0101] Figure 3 It is a structural schematic diagram of a stator assembly detection device provided by an embodiment of the present invention. The stator assembly detection device 300 may have relatively large differences due to different configurations or performances, and may include one or more processors (central processing units, CPU) 310 (for example, one or more processors) and a memory 320, and one or more storage media 330 (for example, one or more mass storage device terminals) storing application programs 333 or data 332. Among them, the memory 320 and the storage medium 330 can be short-term storage or permanent storage. The program stored in the storage medium 330 may include one or more modules (not shown in the figure), and each module may include a series of instruction operations in the stator assembly detection device 300. Furthermore, the processor 310 can be configured to communicate with the storage medium 330, and execute a series of instruction operations in the storage medium 330 on the stator assembly detection device 300 to implement the steps of the above-mentioned stator assembly detection method.

[0102] The stator assembly detection device 300 may also include one or more power supplies 340, one or more wired or wireless network interfaces 350, one or more input and output interfaces 360, and / or one or more operating systems 331, such as Windows Serve, Mac OS X, Unix, Linux, FreeBSD, etc. It will be appreciated by those skilled in the art that Figure 3 The structure of the stator assembly detection device shown does not constitute a limitation on the stator assembly detection device provided by the present invention, and may include more or fewer components than shown in the figure, or a combination of certain components, or a different arrangement of components.

[0103] The present invention also provides a computer-readable storage medium, which may be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium. Instructions are stored in the computer-readable storage medium. When the instructions are executed on a computer, the computer executes the steps of the stator assembly detection method.

[0104] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the above-described system, device, or unit can refer to the corresponding process in the aforementioned method embodiment and will not be repeated here.

[0105] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art or the whole or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk and other media that can store program code.

[0106] The above description is only a preferred embodiment of the present invention, and does not limit the patent scope of the present invention. All equivalent structural changes made by using the contents of the present invention specification and drawings under the inventive concept of the present invention, or directly / indirectly applied in other related technical fields are included in the patent protection scope of the present invention.

Claims

1. A stator assembly detection method, characterized in that: include: Input an acoustic wave excitation sequence into the preset acoustic detection loop, including configuring a single-point excitation mode, a multi-point collaborative excitation mode, and a sweep frequency excitation mode. Through orthogonal sequence planning, each mode is excited in sequence to obtain a complete acoustic response signal. The arrangement process of the preset acoustic detection circuit includes: dividing the stator into a plurality of detection units, each detection unit corresponds to a stator slot, and each detection unit includes an ultrasonic acoustic transducer and two acoustic sensors; arranging the plurality of detection units uniformly along the outer circumference of the stator, so that the ultrasonic acoustic transducer of each detection unit is located at the top of the slot of the corresponding stator slot; installing the two acoustic sensors in each detection unit on both sides of the corresponding stator slot, respectively, to receive the ultrasonic signal reflected by the winding; The configuration of the single-point excitation mode, the multi-point collaborative excitation mode and the swept frequency excitation mode, and the orthogonal sequence planning are used to sequentially excite each mode to obtain a complete acoustic response signal, including: based on a preset first orthogonal excitation sequence, the ultrasonic acoustic transducer is controlled to work alone according to a preset time interval, and the single-point response signal corresponding to the acoustic sensor is synchronously collected; based on a preset second orthogonal excitation sequence, the adjacent ultrasonic acoustic transducers are controlled to work in combination according to a preset phase difference to generate a spatial directional sound beam, and the spatial response signals of all the acoustic sensors are collected; based on a preset third orthogonal excitation sequence, the ultrasonic acoustic transducers are controlled to work synchronously according to a linear frequency modulation rule to generate a wideband excitation signal, and the frequency response signals of all the acoustic sensors are collected; the collected single-point response signal, spatial response signal and frequency response signal are synthesized according to the time sequence through orthogonal sequence planning, and an anti-aliasing buffer is set between adjacent response signals to obtain the complete acoustic response signal; The acoustic response signal is preprocessed in the order of amplitude normalization, baseline drift correction and noise elimination, and the acoustic wave propagation delay, attenuation coefficient and spectrum energy characteristics are extracted to obtain an acoustic characteristic spectrum; Dividing the acoustic characteristic spectrum into a plurality of circumferential analysis units according to the stator slot positions, calculating the correlation coefficient and spectrum similarity between adjacent units, and obtaining characteristic deviation data; The characteristic deviation data is compared with a preset evaluation standard to determine the air gap state level and uneven type of each circumferential analysis unit, and obtain an air gap uniformity detection result.

2. The stator assembly detection method according to claim 1, characterized in that: The acoustic characteristic spectrum is divided into a plurality of circumferential analysis units according to the stator slot positions, and the correlation coefficient and spectrum similarity between adjacent units are calculated to obtain characteristic deviation data, including: The acoustic characteristic spectrum is converted into an analytical signal through Hilbert transform, and the instantaneous amplitude and instantaneous phase characteristics corresponding to each stator slot are extracted; Based on the instantaneous amplitude and instantaneous phase characteristics, a circumferential acoustic characteristic matrix is ​​constructed, and the circumferential acoustic characteristic matrix is ​​divided into a plurality of circumferential analysis units according to the circumferential distribution of the stator slots; Calculating the time domain correlation coefficient between the instantaneous features of adjacent circumferential analysis units, and performing spectrum analysis on the instantaneous features to obtain spectrum similarity; Adaptive threshold processing is performed based on the time domain correlation coefficient and the frequency spectrum similarity to obtain the characteristic deviation data.

3. The stator assembly detection method according to claim 2, characterized in that: The performing spectrum analysis on the instantaneous features to obtain spectrum similarity includes: Performing wavelet packet multi-layer decomposition on the instantaneous characteristics of the adjacent circumferential analysis units to obtain wavelet packet coefficients of different frequency bands; The optimal decomposition layer number of each frequency band is determined based on the minimum Shannon entropy criterion to obtain the optimal decomposition frequency band; The energy value of each frequency band is calculated according to the wavelet packet coefficient corresponding to the optimal decomposition frequency band, and the relative energy value of each frequency band is obtained by normalization processing; The similarity of the relative energy values ​​of each frequency band between adjacent circumferential analysis units is calculated to obtain the frequency spectrum similarity.

4. The stator assembly detection method according to claim 1, characterized in that: The characteristic deviation data is compared with a preset evaluation standard to determine the air gap state level and uneven type of each circumferential analysis unit, including: Constructing a multidimensional feature space, mapping the feature deviation data into a feature vector, wherein the feature vector includes a time domain correlation coefficient component and a frequency spectrum similarity component; Calculate the degree of deviation between the characteristic vector and a preset standard characteristic vector based on the Mahalanobis distance to obtain a quantitative evaluation value of the air gap state; According to the distribution law of the quantitative evaluation value of the air gap state, the air gap state is divided into different levels by using a fuzzy clustering method; Based on the distribution pattern of the feature vector in the multi-dimensional feature space, the type of air gap non-uniformity is determined.

5. A stator assembly detection device, characterized in that: The stator assembly detection device adopts the stator assembly detection method according to any one of claims 1 to 4, and the stator assembly detection device comprises: The excitation control module is used to input the acoustic wave excitation sequence into the preset acoustic detection loop, including configuring the single-point excitation mode, the multi-point collaborative excitation mode and the sweep frequency excitation mode, and to excite each mode in sequence through orthogonal sequence planning to obtain a complete acoustic response signal; A signal processing module, used to perform signal preprocessing on the acoustic response signal in the order of amplitude normalization, baseline drift correction and noise elimination, and extract the acoustic wave propagation delay, attenuation coefficient and spectrum energy characteristics to obtain an acoustic characteristic spectrum; A feature analysis module, used to divide the acoustic feature spectrum into a plurality of circumferential analysis units according to the stator slot positions, calculate the correlation coefficient and spectrum similarity between adjacent units, and obtain feature deviation data; The state determination module is used to compare the characteristic deviation data with a preset evaluation standard, determine the air gap state level and uneven type of each circumferential analysis unit, and obtain an air gap uniformity detection result.

6. A stator assembly detection device, characterized in that: The stator assembly detection device comprises: a memory and at least one processor, wherein instructions are stored in the memory; The at least one processor calls the instructions in the memory to enable the stator assembly detection device to perform the steps of the stator assembly detection method according to any one of claims 1-4.

7. A computer-readable storage medium having instructions stored thereon, characterized in that: When the instructions are executed by the processor, the steps of the stator assembly detection method according to any one of claims 1 to 4 are implemented.

Images