Generator stator slot wedge tightness detection method

Abstract

The invention provides a method for detecting the tightness of a slot wedge of a generator stator. The method comprises the following steps: S101, collecting a sound signal for knocking the slot wedge; s102, carrying out preprocessing operation on the collected sound signals; s103, extracting fault feature parameters of the sound signals subjected to the preprocessing operation; and S104, based on the fault characteristic parameters of the sound signals, using a KNN algorithm to diagnose the tightness of the slot wedge. The generator slot wedge fault diagnosis method efficiently and accurately completes generator slot wedge fault diagnosis by collecting the sound signal at the position knocked by the knocking hammer and analyzing entropy information of the sound signal, and has the advantages of being simple and convenient in signal collection, high in diagnosis efficiency and accurate in diagnosis result.

Description

Technical Field

[0001] The invention relates to the technical field of generators, and in particular to a method for detecting tightness of a generator stator slot wedge. Background Art

[0002] The generator is an important part of the power system, and the generator is composed of a stator, a rotor and other components. The slot wedge is a structure used to fix the stator bars. When the generator is operating normally, the stator bars are energized, causing them to vibrate in the stator slots due to the radial electromagnetic force. This vibration will cause the slot wedge plate that fixes the stator bars to loosen due to the long-term operation of the generator, thereby aggravating the vibration of the stator bars and forming a vicious cycle. Long-term vibration causes the insulation layer to break, eventually causing the generator to fail to operate normally and even cause shutdown or other safety accidents.

[0003] Traditional manual inspection requires knocking on the slot wedges one by one. This inspection method has low inspection efficiency and cannot complete fault detection of large generators in a timely manner. The inspection results are based on the experience of the inspectors and are unstable. Currently, a large number of inspection methods analyze the sound signals emitted by knocking on the slot wedges and judge the status of the slot wedges, but the currently known analysis methods cannot guarantee extremely high accuracy. Summary of the invention

[0004] In view of this, an object of the present invention is to provide a method for detecting the tightness of a generator stator slot wedge, so as to improve the accuracy of detecting the tightness of a generator stator slot wedge.

[0005] To achieve the above-mentioned object of the invention, the present invention provides a method for detecting the tightness of a generator stator slot wedge, the method comprising the following steps:

[0006] S101, collecting sound signals of the wedge being struck;

[0007] S102, performing preprocessing operations on the collected sound signals;

[0008] S103, extracting fault characteristic parameters of the sound signal after the preprocessing operation;

[0009] S104. Based on the fault characteristic parameters of the sound signal, the tightness of the slot wedge is diagnosed using a KNN algorithm.

[0010] Furthermore, the collecting of the sound signal of the wedge hitting the slot includes the following steps:

[0011] S11, controlling the vibrator to strike the surface of the slot wedge;

[0012] S12, collecting sound signals generated by knocking through a sound sensor;

[0013] S13, the knocking sound signal collected by the sound sensor is transmitted through the radio frequency signal line, and the data is converted by the signal acquisition instrument and uploaded to the host computer.

[0014] Furthermore, the preprocessing operation includes the following steps:

[0015] S21, performing multi-scale wavelet decomposition on the noisy signal in the collected sound signal to obtain wavelet coefficients at different scales;

[0016] S22, applying a cross-validation threshold to perform soft threshold or hard threshold processing on the wavelet coefficients of the high-frequency part;

[0017] S23, performing inverse wavelet transform on the processed wavelet coefficients to reconstruct the signal.

[0018] Furthermore, after preprocessing, the multi-scale frequency band energy entropy spectrum of the sound signal is extracted, and the eigenvalue matrix is ​​constructed based on the multi-scale frequency band energy entropy spectrum, and the eigenvalue matrix is ​​used as the fault characteristic parameter for diagnosing the tightness of the slot wedge.

[0019] Furthermore, the extraction of fault characteristic parameters of the sound signal includes the following steps:

[0020] S31, performing multi-scale decomposition on the sound signal after the preprocessing operation to obtain a series of sound signals at different scales;

[0021] S32. Use Fourier transform to decompose the sound signals at different scales into multiple frequency bands. Assume there are q frequency bands, and the energy E of frequency band q jq , which is expressed as follows:

[0022]

[0023] Among them, f q1 and f q2 are the lower and upper frequencies of the frequency band q, respectively, and X (j) (f) is the representation of the sound signal at the jth scale in the frequency domain, (f) is the frequency domain, and df is the frequency change in the frequency domain;

[0024] S33, calculate the proportion of each frequency band energy to the total energy, expressed as follows:

[0025]

[0026] Among them, p jq is the ratio of each frequency band energy to the total energy, q is the index variable, the starting value is 1, and the ending value is Q;

[0027] S34. Calculate the energy entropy for each scale and frequency band respectively, as follows:

[0028]

[0029] Among them, H jq is the energy entropy;

[0030] S35. Combine the energy entropy values ​​at all scales and frequency bands to extract the multi-scale frequency band energy entropy spectrum

[0031] H MSE =[H 1q , H 2q , ..., H Jq ]

[0032] Among them, H MSE is the multi-scale frequency band energy entropy spectrum, H 1q is the energy entropy value of the qth frequency band at the first scale, H 2q is the energy entropy value of the qth frequency band at the second scale, J is the total number of scales;

[0033] S36. Taking each entropy value in the multi-scale frequency band energy entropy spectrum as a feature, constructing an eigenvalue matrix, which is expressed as follows:

[0034] x i =[H 1qi , H 2qi , ..., H Jqi ]

[0035] Among them, x i is the eigenvalue matrix, H 1qi is the energy entropy value of the qth frequency band on the first scale of the i-th sound signal, H 2qi is the energy entropy value of the qth frequency band on the second scale of the i-th sound signal.

[0036] Furthermore, the method of diagnosing the tightness of the slot wedge using the KNN algorithm includes the following steps:

[0037] S41, determine the value in the KNN algorithm, and consider the nearest K neighbors to judge the new sample;

[0038] S42. Use the Euclidean distance method to calculate the distance between the new sample and each sample in the training set, expressed as follows:

[0039]

[0040] Among them, d(x m , x i ) is the distance between the mth sample and the ith sample, x m is the mth sample, x i is the i-th sample, x mj and jare the values ​​of the mth sample and the i-th sample on the jth feature, respectively, and Q is the number of frequency bands;

[0041] S43. Based on the distance between the new sample and each sample in the training set, determine the K nearest neighbors of the new sample, calculate and compare the number of samples in each state of the K samples, and determine the category of the new sample.

[0042] Compared with the prior art, the present invention has the following beneficial effects:

[0043] The present invention can efficiently and accurately complete the generator slot wedge fault diagnosis by performing a preprocessing operation on the collected sound signal of the slot wedge knocking, extracting the fault characteristic parameters of the knocking sound signal after the preprocessing operation, and performing parameter analysis on the knocking sound signal. The present invention has the advantages of simple signal acquisition, high diagnostic efficiency, accurate diagnostic results, etc. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only preferred embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0045] Figure 1 A schematic flow chart of a method for detecting tightness of a generator stator slot wedge provided in an embodiment of the present invention;

[0046] Figure 2 A schematic diagram of the structure of a generator stator slot wedge provided in an embodiment of the present invention;

[0047] Figure 3 A schematic diagram of the analysis of the 2000ms slot wedge tightening state provided by an embodiment of the present invention;

[0048] Figure 4 A schematic diagram of the 1000ms slot wedge tightening state analysis provided by an embodiment of the present invention;

[0049] Figure 5 A schematic diagram of the 500ms slot wedge tightening state analysis provided by an embodiment of the present invention;

[0050] Figure 6 A schematic diagram of the 300ms slot wedge tightening state analysis provided by an embodiment of the present invention;

[0051] Figure 7 A schematic diagram of the 2000ms slot wedge loose state analysis provided by an embodiment of the present invention;

[0052] Figure 8 A schematic diagram of the 1000ms slot wedge loose state analysis provided by an embodiment of the present invention;

[0053] Fig. 9 A schematic diagram of the 500ms slot wedge loose state analysis provided by an embodiment of the present invention;

[0054] Fig.10 A schematic diagram of a 300ms slot wedge loose state analysis provided by an embodiment of the present invention;

[0055] Fig.11 A schematic diagram of 100K sampling frequency analysis provided by an embodiment of the present invention;

[0056] Fig.12 A schematic diagram of 50K sampling frequency analysis provided by an embodiment of the present invention;

[0057] Fig.13 A schematic diagram of 20K sampling frequency analysis provided by an embodiment of the present invention;

[0058] Fig.14 A schematic diagram of 10K sampling frequency analysis provided by an embodiment of the present invention;

[0059] Fig.15 A schematic diagram of the analysis of the second excitation frequency of the excitation provided by an embodiment of the present invention;

[0060] Fig.16 A schematic diagram of three excitation frequency analyses provided in an embodiment of the present invention;

[0061] Fig.17 A schematic diagram of the analysis of the four excitation frequencies of the excitation provided by the embodiment of the present invention;

[0062] Fig.18 A schematic diagram of six excitation frequency analyses provided by an embodiment of the present invention;

[0063] Fig.19 A schematic diagram of non-offset excitation point analysis provided by an embodiment of the present invention;

[0064] Fig. 20 A schematic diagram of an analysis of an excitation point offset by 2 mm provided in an embodiment of the present invention;

[0065] Fig.21 A schematic diagram of an analysis of an excitation point offset by 4 mm provided in an embodiment of the present invention;

[0066] Fig. 22 A schematic diagram of an analysis of an excitation point offset by 6 mm provided in an embodiment of the present invention;

[0067] Fig.23 A schematic diagram of a slot wedge tightness test analysis provided by an embodiment of the present invention; DETAILED DESCRIPTION

[0068] The principles and features of the present invention are described below in conjunction with the accompanying drawings. The enumerated embodiments are only used to explain the present invention and are not used to limit the scope of the present invention.

[0069] like Figure 1 As shown, the present invention provides a flow chart of a method for detecting tightness of a generator stator slot wedge.

[0070] This embodiment provides a method for detecting the tightness of a generator stator slot wedge, the method comprising the following steps:

[0071] S101, collecting sound signals of the wedge being struck, specifically comprising the following steps:

[0072] S11, controlling the vibrator to strike the surface of the slot wedge;

[0073] S12, collecting sound signals generated by knocking through a sound sensor;

[0074] S13, the knocking sound signal collected by the sound sensor is transmitted through the radio frequency signal line, and the data is converted by the signal acquisition instrument and uploaded to the host computer.

[0075] S102, preprocessing the collected sound signal, specifically including the following steps:

[0076] S21, selecting Daubechies wavelet according to the characteristics of the knocking sound signal, performing multi-scale wavelet decomposition on the noisy signal in the collected sound signal, and obtaining wavelet coefficients at different scales;

[0077] S22, applying a cross-validation threshold to determine a threshold, and performing soft threshold or hard threshold processing on the wavelet coefficients of the high-frequency part above the threshold to reduce the influence of noise, thereby achieving the effect of noise reduction filtering;

[0078] S23, performing inverse wavelet transform on the processed wavelet coefficients to reconstruct the signal.

[0079] S103, extracting fault characteristic parameters of the knocking sound signal after the pre-processing operation, specifically comprising the following steps:

[0080] After preprocessing, the multi-scale frequency band energy entropy spectrum of the sound signal is extracted, and the eigenvalue matrix is ​​constructed based on the multi-scale frequency band energy entropy spectrum. The eigenvalue matrix is ​​used as the fault characteristic parameter for diagnosing the tightness of the slot wedge.

[0081] S31, multi-scale decomposition: performing multi-scale decomposition on the sound signal after the preprocessing operation to obtain a series of sound signals at different scales;

[0082] S32. Frequency band division: Use Fourier transform to decompose sound signals at different scales into multiple frequency bands. Assume there are q frequency bands, and the energy E of frequency band q jq , which is expressed as follows:

[0083]

[0084] Among them, f q1 and f q2 are the lower and upper frequencies of the frequency band q, respectively, and X (j) (f) is the representation of the sound signal in the frequency domain at the jth scale, describing the amplitude and phase of the sound signal at different frequencies; (f) is the frequency domain, df is the frequency change in the frequency domain;

[0085] S33, energy normalization: calculate the ratio of each frequency band energy to the total energy, expressed as follows:

[0086]

[0087] Among them, p jq is the ratio of each frequency band energy to the total energy, q is the index variable, the starting value is 1, and the ending value is Q;

[0088] S34. Calculate energy entropy: Calculate energy entropy for each scale and frequency band respectively, as shown below:

[0089]

[0090] Among them, H jq is the energy entropy;

[0091] S35, extracting multi-scale frequency band energy entropy spectrum: combining the energy entropy values ​​on all scales and frequency bands, extracting the multi-scale frequency band energy entropy spectrum, which is expressed as follows:

[0092] H MSE =[H 1q , H 2q , ..., H Jq ]

[0093] Among them, H MSE is the multi-scale frequency band energy entropy spectrum, H 1q is the energy entropy value of the qth frequency band at the first scale, H 2q is the energy entropy value of the qth frequency band at the second scale, J is the total number of scales;

[0094] S36. Taking each entropy value in the multi-scale frequency band energy entropy spectrum as a feature, constructing an eigenvalue matrix, which is expressed as follows:

[0095] x i =[H 1qi , H2qi , ..., H Jqi ]

[0096] Among them, x i is the eigenvalue matrix, H 1qi is the energy entropy value of the qth frequency band on the first scale of the i-th sound signal, H 2qi is the energy entropy value of the qth frequency band on the second scale of the i-th sound signal.

[0097] S104, based on the fault characteristic parameters of the sound signal, using the KNN algorithm to diagnose the tightness of the slot wedge, specifically including the following steps:

[0098] The eigenvalue matrix constructed based on the multi-scale frequency band energy entropy spectrum in step S103 is normalized to ensure that the contribution weight of each eigenvalue to the classification is the same;

[0099] The eigenvalue matrix is ​​divided into a test data set and a training data set. The training data set is divided into N equal subsets through cross-validation. Each subset is used as a validation set in turn, and the rest are used as training sets. Finally, the average value of the performance evaluation index of the N validation results is used as the performance evaluation of the model. By comparing the performance evaluation indicators of the model under different K values, the best K value is selected, the optimal training model is determined, and the trained model is tested on the test data set to verify the effectiveness of the KNN algorithm.

[0100] S41, based on step S41, determine the value in the KNN algorithm, and consider the nearest K neighbors to judge the new sample;

[0101] S42. Use the Euclidean distance method to calculate the distance between the new sample and each sample in the training set, expressed as follows:

[0102]

[0103] Among them, d(x m ,x i ) is the distance between the mth sample and the ith sample, x m is the mth sample, x i is the i-th sample, x mj and x ij are the values ​​of the mth sample and the i-th sample on the jth feature, respectively, and Q is the number of frequency bands;

[0104] S43. Based on the distance between the new sample and each sample in the training set, determine the K nearest neighbors of the new sample, calculate and compare the number of samples in each state of the K samples, and determine the category of the new sample.

[0105] like Figure 2As shown in the figure, when the stator slot wedge is fixed very tightly, the slot wedge and the entire stator slot can be regarded as a whole. After a certain impact load is applied, the slot wedge will not vibrate itself; when the stator slot wedge is loose, the slot wedge and the slot at the loose part are no longer a tightly combined whole. Under the action of the impact load, the loose part of the slot wedge will inevitably produce local vibration, thus forming a sound wave with a specific frequency, which is essentially the vibration problem of a thin plate acoustic resonator. Combined with the material, width and thickness parameters of the stator slot wedge, a thin plate acoustic resonator model can be established based on the thin plate theory, and then the main characteristic frequency of the stator slot wedge after loosening can be calculated. In addition, the tightness of the slot wedge is related to the amplitude of the main characteristic frequency component. In comparison, the spectral peak amplitude of the main characteristic frequency component of the acoustic signal radiated by the loose slot wedge is relatively large, while the spectral peak amplitude of the main characteristic frequency component of the acoustic signal radiated by the tightened slot wedge is relatively small.

[0106] Experimental test analysis: Analyze the sampling time, sampling frequency, excitation frequency, and excitation point, including:

[0107] (1) Sampling time analysis: Figure 3-Figure 6 As shown, the tight state of the slot wedge is analyzed, and the sampling time is 2000ms, 1000ms, 500ms, and 300ms respectively;

[0108] like Figure 7-Figure 10 As shown, the wedge loosening state is analyzed, and the sampling time is 2000ms, 1000ms, 500ms, and 300ms respectively;

[0109] By collecting data at the sampling frequency and observing and analyzing the images, it can be determined that the sampling time of 2000ms, 1000ms, 500ms, and 300ms can collect all the sound signals generated by the knocking. By observing the time domain image, it can be found that when the sampling time is short, all the required sound signals can be collected and the collection of other interference signals can be reduced at the same time. Combined with the literature on slot wedge vibration, the sampling time is selected as 500ms.

[0110] (2) Sampling frequency analysis: Fig.11 As shown, the sampling frequency of 100K is analyzed, and the sampling time is 2S;

[0111] like Fig.12 As shown, the sampling frequency of 50K is analyzed, and the sampling time is 0.5S;

[0112] like Fig.13 As shown, the 20K sampling frequency is analyzed, and the sampling time is 0.5S;

[0113] like Fig.14 As shown, the 10K sampling frequency is analyzed, and the sampling time is 0.5S;

[0114] By observing and comparing the images, it is found that the characteristic frequency of the slot wedge appears near 6000Hz. If the complete frequency band of the slot wedge knocking signal is to be collected, based on the sampling theorem, the sampling frequency should be at least 2.56-4 times the highest frequency in the collected signal. Therefore, among the sampling frequencies tested, 50k and 100k can meet this requirement. However, considering that the sampling time is determined, the higher the sampling frequency, the more complicated the collected data is, and the longer the data cycle of the later algorithm analysis and processing is. At the same time, there is no difference in the judgment of the slot wedge fault characteristics, so the sampling frequency is selected as 50k;

[0115] (3) Excitation frequency analysis: Fig.15 As shown, the sampling time is 500ms, the sampling frequency is 50K, the extension time is 50ms, and the excitation is 2 times within 500ms;

[0116] like Fig.16 As shown, the sampling time is 500ms, the sampling frequency is 50K, the extension time is 50ms, and the excitation is 3 times within 500ms;

[0117] like Fig.17 As shown, the sampling time is 500ms, the sampling frequency is 50K, the extension time is 50ms, and the excitation is 4 times within 500ms;

[0118] like Fig.18 As shown, the sampling time is 500ms, the sampling frequency is 50K, the extension time is 50ms, and the excitation is 6 times within 500ms;

[0119] The frequency settings of the signal generator are 2, 3, 4, and 6 knocks / 500ms to determine the excitation frequency during the signal generator test. Analysis and comparison show that different excitation frequencies have little effect on the integrity of the collected signal, but considering that the energy in the interval where the characteristic frequency of the collected signal is located is more significant, and at the same time, to prevent the temperature and wear rate of the knocking head from being increased due to excessively high excitation frequency, 3 times should be selected as the final excitation frequency.

[0120] (4) Excitation point analysis: Analysis of the influence of the degree of deviation of the excitation point from the center line. After selecting the striking force, test the signal response when the striking point deviates from the center line to different degrees;

[0121] like Fig.19 As shown, the sampling frequency is 50K, the extension time is 50ms, and the offset state is no offset;

[0122] like Fig. 20 As shown, the sampling frequency is 50K, the extension time is 50ms, and the offset state is 2mm;

[0123] like Fig.21As shown, the sampling frequency is 50K, the extension time is 50ms, and the offset state is 4mm;

[0124] like Fig. 22 As shown, the sampling frequency is 50K, the extension time is 50ms, and the offset state is 6mm;

[0125] Image comparison and analysis found that the influence of radial offset of the excitation position on the knocking sound signal can be ignored, that is, for the same slot wedge, the signals collected by knocking at different positions can diagnose the state of the slot wedge and will not affect the diagnosis result.

[0126] (5) Experimental test results: Fig.23 As shown, when the sound signal is collected, data analysis extracts the frequency domain peak frequency of the signal. When it is greater than 5375 Hz, the slot wedge is in a tight state. Conversely, when it is lower than 5375 Hz, the slot wedge is in a loose state.

[0127] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for detecting the tightness of a generator stator slot wedge, characterized in that: The method comprises the following steps: S101, collecting sound signals of the wedge being struck; S102, performing preprocessing operations on the collected sound signals; S103, extracting fault characteristic parameters of the sound signal after the preprocessing operation; S104. Based on the fault characteristic parameters of the sound signal, the tightness of the slot wedge is diagnosed using a KNN algorithm.

2. A method for detecting tightness of a generator stator slot wedge according to claim 1, characterized in that: In step S101, collecting the sound signal of the wedge hitting the slot includes the following steps: S11, controlling the vibrator to strike the surface of the slot wedge; S12, collecting sound signals generated by knocking through a sound sensor; S13, transmitting the knocking sound signal collected by the sound sensor through the radio frequency signal line, and converting the data through the signal acquisition instrument and uploading it to the host computer.

3. A method for detecting tightness of a generator stator slot wedge according to claim 1, characterized in that: In step S102, the preprocessing operation includes the following steps: S21, performing multi-scale wavelet decomposition on the noisy signal in the collected sound signal to obtain wavelet coefficients at different scales; S22, applying a cross-validation threshold to perform soft threshold or hard threshold processing on the wavelet coefficients of the high-frequency part; S23, performing inverse wavelet transform on the processed wavelet coefficients to reconstruct the signal.

4. A method for detecting tightness of a generator stator slot wedge according to claim 1, characterized in that: The method for extracting fault characteristic parameters of the sound signal after the preprocessing operation includes: extracting the multi-scale frequency band energy entropy spectrum of the sound signal after the preprocessing operation, constructing an eigenvalue matrix based on the multi-scale frequency band energy entropy spectrum, and using the eigenvalue matrix as the fault characteristic parameters for diagnosing the tightness of the slot wedge.

5. A method for detecting tightness of a generator stator slot wedge according to claim 4, characterized in that: The extraction of fault characteristic parameters of the sound signal comprises the following steps: S31, performing multi-scale decomposition on the sound signal after the preprocessing operation to obtain a series of sound signals at different scales; S32. Use Fourier transform to decompose the sound signals at different scales into multiple frequency bands. Assume there are Q frequency bands, and the energy E of frequency band q jq , which is expressed as follows: Among them, f q1 and f q2 are the lower and upper frequencies of the frequency band q, respectively, and X (j) (f) is the representation of the sound signal at the jth scale in the frequency domain, (f) is the frequency domain, and df is the frequency change in the frequency domain; S33, calculate the ratio of each frequency band energy to the total energy, expressed as follows: Among them, p jq is the ratio of each frequency band energy to the total energy, q is the index variable, the starting value is 1, and the ending value is Q; S34. Calculate the energy entropy for each scale and frequency band respectively, as follows: Among them, H jq is the energy entropy; S35, combining the energy entropy values ​​at all scales and frequency bands, extracting the multi-scale frequency band energy entropy spectrum, expressed as follows: H MSE =[H 1q ,H 2q ,...,H Jq ] Among them, H MSE is the multi-scale frequency band energy entropy spectrum, H 1q is the energy entropy value of the qth frequency band at the first scale, H 2q is the energy entropy value of the qth frequency band at the second scale, J is the total number of scales; S36. Taking each entropy value in the multi-scale frequency band energy entropy spectrum as a feature, constructing an eigenvalue matrix, which is expressed as follows: x i =[H 1qi ,H 2qi ,...,H Jqi ] Among them, x i is the eigenvalue matrix, H 1qi is the energy entropy value of the qth frequency band on the first scale of the i-th sound signal, H 2qi is the energy entropy value of the qth frequency band on the second scale of the i-th sound signal.

6. A method for detecting tightness of a generator stator slot wedge according to claim 1, characterized in that: The method of diagnosing the tightness of the slot wedge using the KNN algorithm includes the following steps: S41, determine the value in the KNN algorithm, and consider the nearest K neighbors to judge the new sample; S42. Use the Euclidean distance method to calculate the distance between the new sample and each sample in the training set, expressed as follows: Among them, d(x m ,x i ) is the distance between the mth sample and the ith sample, x m is the mth sample, x i is the i-th sample, x mj and x ij are the values ​​of the mth sample and the i-th sample on the jth feature, respectively, and Q is the number of frequency bands; S43. Based on the distance between the new sample and each sample in the training set, determine the K nearest neighbors of the new sample, calculate and compare the number of samples in each state of the K samples, and determine the category of the new sample.

Images