Electrotome motor anomaly detection method

Abstract

The invention discloses an electrotome motor anomaly detection method, and relates to the technical field of electrotome motor anomaly detection, and the method comprises the steps: extracting the sample frequency domain features of a motor vibration signal sample of an electrotome, and obtaining a plurality of first feature sets based on the sample frequency domain features; obtaining any two first feature sets, and respectively obtaining features of the sets to obtain first features and second features; obtaining a distance threshold based on the distance between the first feature and the second feature; extracting a to-be-detected frequency domain feature of a to-be-detected vibration signal of the to-be-detected electrotome, and obtaining a plurality of third sets based on the to-be-detected frequency domain feature; respectively acquiring features of the third set and the first feature set to obtain a third feature and a fourth feature; obtaining a to-be-detected distance based on the distance between the third feature and the fourth feature; and the first detection result of the motor is obtained based on the distance threshold and the to-be-detected distance, so that the problem of low anomaly detection accuracy caused by only simple splicing in feature fusion of a traditional vibration detection method can be solved.

Description

Technical Field

[0001] The present invention relates to the field of abnormality detection of electric knife motors, and in particular to an abnormality detection method of electric knife motors. Background Art

[0002] As an important medical surgical instrument, the normal operation of the motor of the electrosurgical unit is crucial to the safety and effectiveness of the operation. During the operation of the electrosurgical unit, due to various factors, abnormal vibrations may occur, such as bearing wear, rotor imbalance, motor winding failure, etc. These abnormal vibrations will not only affect the cutting accuracy and performance of the electrosurgical unit, but may also increase the risk of surgery.

[0003] The traditional vibration detection method is to extract the features of the electrosurgical unit motor vibration data, normalize all the features and then superimpose them to obtain superimposed features, and then perform anomaly detection by identifying the superimposed features. However, this method simply splices the features, that is, the weight of each feature is the same or the weight is manually designed through experience, so the features cannot be fully integrated, resulting in low accuracy of anomaly detection. In addition, due to the influence of different surgical types and individual differences, this method has poor adaptability to complex working conditions and cannot effectively meet actual needs, which further leads to low accuracy of anomaly detection. Summary of the invention

[0004] In order to solve the problem that the feature fusion of the traditional vibration detection method is simply spliced ​​and the features cannot be fully fused, resulting in low accuracy of abnormality detection, the present invention provides an abnormality detection method for an electric knife motor, the method comprising: Acquire a plurality of motor vibration signal samples of the electric knife, extract the sample frequency domain features of each of the motor vibration signal samples, obtain a sample frequency domain feature matrix based on the sample frequency domain features, and obtain a plurality of first feature sets based on the sample frequency domain feature matrix, wherein the first feature sets all include a plurality of normalized sample frequency domain features; Obtain any two of the first feature sets to obtain a first set and a second set, obtain features of the first set and the second set respectively, and obtain a plurality of first features and a plurality of second features; Based on a first weight coefficient, obtaining a Jaccard distance between the first feature and the second feature, obtaining a plurality of first Jaccard distances, and obtaining a first distance threshold based on a maximum of the first Jaccard distances; Acquire a vibration signal to be detected of the electric knife to be detected, extract frequency domain features to be detected of the vibration signal to be detected, obtain a frequency domain feature matrix to be detected based on the frequency domain features to be detected, and obtain a plurality of third sets based on the frequency domain feature matrix to be detected, wherein the third sets all contain a plurality of normalized frequency domain features to be detected; Respectively acquiring features of the third set and the first feature set to obtain a plurality of third features and a plurality of fourth features; Based on the second weight coefficient, obtaining the Jaccard distance between the third feature and the fourth feature, obtaining a plurality of second Jaccard distances, and obtaining a first distance to be detected based on the maximum of the second Jaccard distances; Based on the first distance threshold and the first distance to be detected, a first detection result of the motor of the electric knife to be detected is obtained.

[0005] Determining the weight of each feature can more accurately reflect the importance of different features in abnormality detection and improve the accuracy and reliability of detection; different vibration features have different importance in reflecting the operating status of the motor. The present invention introduces a weight coefficient. By assigning different weight coefficients to each vibration feature, the key features that have a greater impact on the operating status of the motor can be highlighted and the interference of secondary features can be suppressed; the weight coefficient can adjust the contribution of each feature in the fusion process according to actual conditions, so that the fused features can more accurately reflect the relationship between the operating status of the motor and each feature, so that the fused features can more accurately reflect the operating status of the motor. The Jaccard distance is used to measure the degree of difference between two sets. In vibration feature fusion, it can be used to measure the similarity or difference between different vibration features. By calculating the Jaccard distance, the present invention can understand the degree of correlation between each vibration feature and provide a basis for feature fusion. The present invention combines the weight coefficient with the Jaccard distance, which can capture the nonlinear relationship between different vibration characteristics while highlighting the key features that have a greater impact on the operating state of the motor. Even if there are complex interactions between the features, the Jaccard distance can reflect the intrinsic connection between them to a certain extent, thereby more comprehensively and fully fusing the feature information, so that the fused features can more accurately reflect the operating state of the motor, thereby improving the precision and accuracy of abnormality detection.

[0006] Among the multiple features of motor vibration, different features have different sensitivities to reflecting small changes. By setting the weight coefficients reasonably, those features that are more sensitive to small changes will be given higher weights. In this way, when fusing features, small changes in these key features will be highlighted and amplified, making them easier to detect. The Jaccard distance can consider the association between vibration features as a whole. When there is a small change in motor vibration, it may cause coordinated changes between multiple features. The Jaccard distance can capture the difference in feature sets brought about by this coordinated change, rather than just the change of a single feature, thereby reflecting the small changes in motor vibration more comprehensively and sensitively.

[0007] After the weight coefficient and Jaccard distance are combined, they can be adaptively adjusted according to the actual situation of the motor vibration characteristics. When capturing small changes, the weight coefficient will guide the focus on the changes in key features, while the Jaccard distance will make a comprehensive judgment from the perspective of the relationship between features. When a key feature changes slightly, the weight coefficient makes it occupy an important position in the fusion feature. At the same time, the Jaccard distance will examine the relationship between the change and other features, and comprehensively evaluate whether it is indeed an effective small change in motor vibration, rather than a pseudo change caused by noise or other interference, so as to achieve more accurate and sensitive capture of small changes in motor vibration, which is more conducive to the detection of early weak faults.

[0008] Furthermore, the method further comprises: Extracting sample time domain features of each motor vibration signal sample, obtaining a sample time domain feature matrix based on the sample time domain features, and obtaining a plurality of second feature sets based on the sample time domain feature matrix, wherein the second feature sets each include a plurality of normalized sample time domain features; Obtain any two of the second feature sets, obtain a fourth set and a fifth set, obtain features of the fourth set and the fifth set respectively, and obtain a plurality of fifth features and a plurality of sixth features; Based on the third weight coefficient, obtaining the Jaccard distance between the fifth feature and the sixth feature, obtaining a plurality of third Jaccard distances, and obtaining a second distance threshold based on the maximum third Jaccard distance; Extracting the time domain features to be detected of the vibration signal to be detected, obtaining a time domain feature matrix to be detected based on the time domain features to be detected, and obtaining a plurality of sixth sets based on the time domain feature matrix to be detected, wherein the sixth sets each include a plurality of normalized time domain features to be detected; Acquire the features of the sixth set and the second feature set respectively, and obtain a plurality of seventh features and a plurality of eighth features; acquire the Jaccard distance between the seventh feature and the eighth feature based on a fourth weight coefficient, and obtain a plurality of fourth Jaccard distances; and acquire a second distance to be detected based on the maximum fourth Jaccard distance; Based on the second distance threshold and the second distance to be detected, obtaining a second detection result of the motor of the electric knife to be detected; Based on the first detection result and the second detection result, a third detection result of the motor is obtained, and the first detection result of the motor is updated to the third detection result of the motor.

[0009] By obtaining the time domain characteristics of motor vibration and combining time domain and frequency domain analysis, the characteristic information of vibration signals can be obtained more comprehensively and abnormal information detection can be more accurate.

[0010] Furthermore, the first weight coefficient, the second weight coefficient, the third weight coefficient and the fourth weight coefficient are all obtained by: obtaining the first mean of the feature, and obtaining the first variance based on the first mean; obtaining the first covariance of any two features, and obtaining the first correlation coefficient based on the first covariance and the first variance; and obtaining the weight coefficient of each feature based on the first correlation coefficient and the first variance.

[0011] The variance reflects the degree of dispersion of each feature in different samples. The greater the dispersion, the richer the information contained in the feature. The correlation coefficient measures the linear correlation between two features. The closer it is to 1, the stronger the correlation between the two features, and the closer it is to 0, the weaker the correlation. The weights are determined based on the variance and correlation, and the information content of each feature is comprehensively considered and normalized, taking into account the various characteristics and historical data of the motor, ensuring the scientificity and rationality of the weights, making the entire detection method more adaptable and effective.

[0012] Furthermore, the specific steps of obtaining the sample frequency domain feature matrix include: Normalizing the sample frequency domain features, and obtaining the sample frequency domain feature matrix based on the normalized sample frequency domain features; The specific steps of obtaining the frequency domain feature matrix to be detected include: normalizing the frequency domain features to be detected, and obtaining the frequency domain feature matrix to be detected based on the normalized frequency domain features to be detected; The specific steps of obtaining the sample time domain feature matrix include: normalizing the sample time domain features, and obtaining the sample time domain feature matrix based on the normalized sample time domain features; The specific steps of obtaining the time domain feature matrix to be detected include: normalizing the time domain features to be detected, and obtaining the time domain feature matrix to be detected based on the normalized time domain features to be detected.

[0013] Normalization processing unifies the value range of each feature parameter to the interval [0,1].

[0014] Furthermore, the calculation formula for normalizing the sample frequency domain features is: in, Indicates The normalized sample frequency domain features, Indicates The frequency domain features of samples before normalization, and Respectively The minimum and maximum values ​​of .

[0015] Furthermore, the calculation formula of the first Jaccard distance is: in, express The first Jaccard distance between the first feature set A and the first feature set B in , Represents the sample frequency domain characteristics The first weight coefficient of The first feature set A represents Features, represents the first feature set B Features, represents the number of features in the first feature set, Representation characteristics With features The number of intersection elements, Representation characteristics With features The number of elements in the union of .

[0016] Furthermore, the calculation formula of the first correlation coefficient is: in, Represents the sample frequency domain characteristics The first correlation coefficient of Represents the sample frequency domain characteristics And the sample frequency domain characteristics The covariance of and Represent the sample frequency domain characteristics And the sample frequency domain characteristics The mean square error, Indicates The normalized sample frequency domain features, Represents an integer greater than or equal to 1.

[0017] Furthermore, the calculation formula of the first weight coefficient is: in, Represents the sample frequency domain characteristics The first weight coefficient of and Represent the sample frequency domain characteristics And the sample frequency domain characteristics The mean square error, Indicates the number of features in the first feature set.

[0018] Furthermore, the method further comprises: Classifying the motor vibration signal samples to obtain a number of classified sample signals, and obtaining the category of the vibration signal to be detected; Based on the category, obtaining audio signals corresponding to the classified sample signal and the vibration signal to be detected, respectively obtaining a first audio signal and a second audio signal, and obtaining an abnormal audio signal based on the first audio signal and the second audio signal; Based on the second audio signal and the classified sample signal, respectively acquiring a first abnormal vibration signal corresponding to the abnormal audio signal and a second abnormal vibration signal corresponding to the classified sample signal; The first waveform graph and the second waveform graph of the first abnormal vibration signal and the second abnormal vibration signal are respectively obtained, the abnormal similarity of the first waveform graph and the second waveform graph is obtained, the abnormal value is obtained based on the abnormal similarity, and the usage time is obtained based on the abnormal value and the category.

[0019] The motor will make a specific sound when running. If there is a fault (such as bearing damage, rotor imbalance, etc.) and vibration, the frequency, pitch or volume of the sound will change. The abnormal vibration signal is obtained based on the audio characteristics, and a comprehensive judgment is made whether it is abnormal. According to the degree of abnormality and the type of surgery, it is determined how long the electric knife can be used normally. Therefore, during the operation, the user can arrange the replacement time more reasonably.

[0020] Furthermore, the specific steps of obtaining the abnormal similarity include: Acquire an extreme value of the second waveform graph, and adjust the first waveform graph based on the extreme value to obtain a third waveform graph; Based on a preset period, first data and second data of the second waveform graph and the third waveform graph are respectively obtained, waveform similarity is obtained based on the fluctuation directions of the first data and the second data, and a ratio value is obtained based on the values ​​of the first data and the second data; Based on the waveform similarity and the ratio value, the abnormality similarity is obtained.

[0021] By adjusting the two waveforms according to the extreme values ​​and expanding the difference between the two waveform graphs, the similarity and ratio values ​​can be obtained more accurately, thereby making anomaly detection more accurate and using time more accurate.

[0022] One or more technical solutions provided by the present invention have at least the following technical effects or advantages: 1. Assigning different weight coefficients to each vibration feature can highlight the key features that have a greater impact on the motor's operating state and suppress the interference of minor features; adjust the contribution of each feature in the fusion process according to actual conditions, so that the fused features can more accurately reflect the relationship between the motor's operating state and each feature, and more accurately reflect the motor's operating state.

[0023] 2. By calculating the Jaccard distance between features, we can understand the degree of correlation between the vibration features. In vibration feature fusion, we can measure the similarity or difference between different vibration features and provide a basis for feature fusion.

[0024] 3. Combining the weight coefficient and the Jaccard distance can highlight the key features that have a greater impact on the operating state of the motor, while capturing the nonlinear relationship between different vibration features. It can more comprehensively and fully integrate the information of multiple vibration features, so that the fused features can more accurately reflect the operating state of the motor, thereby improving the precision and accuracy of anomaly detection.

[0025] 4. Combining the weight coefficient and the Jaccard distance, when capturing small changes, the weight coefficient will guide the focus on the changes in key features, while the Jaccard distance will make a comprehensive judgment from the perspective of the relationship between features to confirm whether it is an effective small change in motor vibration, thereby achieving more accurate and sensitive capture of small changes in motor vibration, which is more conducive to the detection of early weak faults.

[0026] 5. The weights are determined based on the variance and correlation, and the information content of each feature is comprehensively considered and normalized, taking into account the various characteristics and historical data of the motor, ensuring the scientificity and rationality of the weights, making the entire detection method more adaptable and effective.

[0027] 6. Obtain the usage time based on abnormal values ​​and categories, obtain abnormal vibration signals based on audio features, and comprehensively judge whether it is abnormal. According to the degree of abnormality and the type of surgery, determine how long the electrosurgical unit can be used normally, so that during the operation, the user can arrange the replacement time more reasonably.

[0028] 7. Based on the waveform similarity and the ratio value, the abnormal similarity is obtained. The two waveforms are adjusted according to the extreme values ​​to expand the difference between the two waveforms. The similarity and ratio values ​​can be obtained more accurately, thereby making the abnormal detection more accurate and the use time more accurate. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] The drawings described herein are used to provide a further understanding of the embodiments of the present invention, constitute a part of the present invention, and do not constitute a limitation on the embodiments of the present invention; Figure 1 The present invention is a schematic diagram of a detection process of a method for detecting abnormality of an electric knife motor. DETAILED DESCRIPTION

[0030] In order to more clearly understand the above-mentioned purpose, features and advantages of the present invention, the present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be noted that the embodiments of the present invention and the features in the embodiments can be combined with each other without conflict.

[0031] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those within the scope of this description. Therefore, the protection scope of the present invention is not limited to the specific embodiments disclosed below.

[0032] Example 1

[0033] refer to Figure 1 This embodiment provides a method for detecting abnormality of an electric knife motor, the method comprising: Acquire several motor vibration signal samples of the electric knife, extract the sample frequency domain features of each motor vibration signal sample, such as obtaining the frequency component by fast Fourier transform and other methods, the frequency component may include the frequency of each order of harmonics and its amplitude, etc., so as to obtain the sample frequency domain features; Obtaining a sample frequency domain feature matrix based on the sample frequency domain features, and obtaining a plurality of first feature sets based on the sample frequency domain feature matrix, wherein the first feature sets each include a plurality of normalized sample frequency domain features; For example, the sample frequency domain feature matrix is ​​defined as ,in is the number of features, For the The vector of features, m is the number of samples. Similarly, the sample frequency domain feature matrix after normalization is defined as ;

[0034] Obtain any two of the first feature sets to obtain a first set and a second set, obtain features of the first set and the second set respectively, and obtain a plurality of first features and a plurality of second features; As the first set and , respectively represent the vibration characteristic vector set of the motor in the normal state, both belong to A sample of and The first set A and the second set B correspond to feature vector

[0035] Based on the first weight coefficient, the Jaccard distance between the first feature and the second feature is obtained to obtain a plurality of first Jaccard distances, and a first distance threshold is obtained based on the maximum first Jaccard distance, such as calculating and The Jaccard distance of is obtained, and the maximum value of the Jaccard distance is obtained to obtain the first distance threshold; Obtain a vibration signal to be detected of the electric knife to be detected, extract the frequency domain features to be detected of the vibration signal to be detected, obtain a frequency domain feature matrix to be detected based on the frequency domain features to be detected, and obtain a plurality of third sets based on the frequency domain feature matrix to be detected, wherein the third sets all contain a plurality of normalized frequency domain features to be detected; the processing method of the frequency domain feature matrix to be detected is the same as the processing method of the sample frequency domain feature matrix; Respectively acquiring features of the third set and the first feature set to obtain a plurality of third features and a plurality of fourth features; Based on the second weight coefficient, obtaining the Jaccard distance between the third feature and the fourth feature, obtaining a plurality of second Jaccard distances, and obtaining a first distance to be detected based on the maximum of the second Jaccard distances; Based on the first distance threshold and the first distance to be detected, a first detection result of the motor of the electric knife to be detected is obtained. That is, if the first distance to be detected is greater than the first distance threshold, the motor of the electric knife to be detected is judged to be abnormal, otherwise it is normal.

[0036] The specific steps of obtaining the sample frequency domain feature matrix include: normalizing the sample frequency domain features, and obtaining the sample frequency domain feature matrix based on the normalized sample frequency domain features; The specific steps of obtaining the frequency domain feature matrix to be detected include: normalizing the frequency domain features to be detected, and obtaining the frequency domain feature matrix to be detected based on the normalized frequency domain features to be detected.

[0037] For example, the minimum-maximum normalization is used to process the sample frequency domain feature matrix and the frequency domain feature matrix to be detected.

[0038] Example 2

[0039] On the basis of implementation 1, in this embodiment, the method further includes: Extracting sample time domain features of each motor vibration signal sample, obtaining a sample time domain feature matrix based on the sample time domain features, and obtaining a plurality of second feature sets based on the sample time domain feature matrix, wherein the second feature sets each include a plurality of normalized sample time domain features; in this embodiment, the sample time domain features may include mean, standard deviation, root mean square, peak value, peak-to-peak value, effective value, skewness, peak factor, kurtosis index, etc.; Obtain any two of the second feature sets, obtain a fourth set and a fifth set, obtain features of the fourth set and the fifth set respectively, and obtain a plurality of fifth features and a plurality of sixth features; Based on the third weight coefficient, obtaining the Jaccard distance between the fifth feature and the sixth feature, obtaining a plurality of third Jaccard distances, and obtaining a second distance threshold based on the maximum third Jaccard distance; Extracting the time domain features to be detected of the vibration signal to be detected, obtaining a time domain feature matrix to be detected based on the time domain features to be detected, and obtaining a plurality of sixth sets based on the time domain feature matrix to be detected, wherein the sixth sets each include a plurality of normalized time domain features to be detected; Acquire the features of the sixth set and the second feature set respectively, and obtain a plurality of seventh features and a plurality of eighth features; acquire the Jaccard distance between the seventh feature and the eighth feature based on a fourth weight coefficient, and obtain a plurality of fourth Jaccard distances; and acquire a second distance to be detected based on the maximum fourth Jaccard distance; Based on the second distance threshold and the second distance to be detected, a second detection result of the motor of the electric knife to be detected is obtained; that is, if the second distance to be detected is greater than the second distance threshold, the motor of the electric knife to be detected is judged to be abnormal, otherwise it is normal.

[0040] Based on the first detection result and the second detection result, the third detection result of the motor is obtained, and the first detection result of the motor is updated to the third detection result of the motor. That is, if both detection results are normal, the motor of the electric knife to be detected is judged to be normal, otherwise it is abnormal.

[0041] The specific steps of obtaining the sample time domain feature matrix include: normalizing the sample time domain features, and obtaining the sample time domain feature matrix based on the normalized sample time domain features; The specific steps of obtaining the time domain feature matrix to be detected include: normalizing the time domain features to be detected, and obtaining the time domain feature matrix to be detected based on the normalized time domain features to be detected.

[0042] In this embodiment, the second detection result of the motor is obtained in the same manner as the first detection result of the motor in the first embodiment, except that the former is a time domain feature and the latter is a frequency domain feature.

[0043] Example 3

[0044] Based on the above implementation, in this embodiment, the first weight coefficient, the second weight coefficient, the third weight coefficient and the fourth weight coefficient are all obtained in the following manners: Obtaining a first mean of the feature, and obtaining a first variance based on the first mean; Obtaining a first covariance of any two features, and obtaining a first correlation coefficient based on the first covariance and the first variance; A weight coefficient of each feature is obtained based on the first correlation coefficient and the first variance.

[0045] Example 4

[0046] Based on the above implementation, in this embodiment, the calculation formula for normalizing the sample frequency domain features is: in, Indicates The normalized sample frequency domain features, Indicates The frequency domain features of samples before normalization, and Respectively The minimum and maximum values ​​of .

[0047] The calculation formula of the first Jaccard distance is: in, express The first Jaccard distance between the first feature set A and the first feature set B in , Represents the sample frequency domain characteristics The first weight coefficient of The first feature set A represents Features, represents the first feature set B Features, represents the number of features in the first feature set, Representation characteristics With features The number of intersection elements, Representation characteristics With features The number of elements in the union of .

[0048] The calculation formula of the first correlation coefficient is: in, Represents the sample frequency domain characteristics The first correlation coefficient of Represents the sample frequency domain characteristics And the sample frequency domain characteristics The covariance of and Represent the sample frequency domain characteristics And the sample frequency domain characteristics The mean square error, Indicates The normalized sample frequency domain features, Represents an integer greater than or equal to 1.

[0049] The calculation formula of the first weight coefficient is: in, Represents the sample frequency domain characteristics The first weight coefficient of and Represent the sample frequency domain characteristics And the sample frequency domain characteristics The mean square error, Indicates the number of features in the first feature set.

[0050] In this formula, the denominator comprehensively considers and normalizes the information of each feature by summing the product of the variance of all features and (1-square of the correlation coefficient) so that the sum of the weights is 1.

[0051] Example 5

[0052] Based on the above implementation, in this embodiment, the method further includes: Classifying the motor vibration signal samples, such as classifying them according to the type of surgery, obtaining a number of classified sample signals, and obtaining the type of the vibration signal to be detected; Based on the category, audio signals corresponding to the classified sample signal and the vibration signal to be detected are obtained to obtain a first audio signal and a second audio signal respectively, and an abnormal audio signal is obtained based on the first audio signal and the second audio signal, such as by comparing the frequency, pitch and volume of the sound to obtain an audio signal that exceeds a preset range, thereby obtaining an abnormal audio signal; Based on the second audio signal and the classified sample signal, respectively acquiring a first abnormal vibration signal corresponding to the abnormal audio signal and a second abnormal vibration signal corresponding to the classified sample signal; The first waveform graph and the second waveform graph of the first abnormal vibration signal and the second abnormal vibration signal are obtained respectively, the abnormal similarity of the first waveform graph and the second waveform graph is obtained, the abnormal value is obtained based on the abnormal similarity, and the use time is obtained based on the abnormal value and the category. For example, the use time of the electric knife under different abnormal values ​​and different surgical categories is preset, so as to obtain the current use time of the electric knife according to the current abnormal value.

[0053] The specific steps of obtaining the abnormal similarity include: Obtaining an extreme value of the second waveform graph, and adjusting the first waveform graph based on the extreme value to obtain a third waveform graph, such as adjusting the extreme value of the first waveform graph to the same value according to the extreme value of the second waveform graph; Based on a preset period, first data and second data of the second waveform graph and the third waveform graph are respectively obtained, such as obtaining the first data and second data within a waveform period, and obtaining waveform similarity based on the fluctuation direction of the first data and the second data, such as judging whether the fluctuation direction is upward or downward according to the value, and then obtaining the tangent point according to the derivative of the value, so as to calculate the slope of the waveform according to the value of the tangent point, if the fluctuation directions are different, the waveform similarity can be directly defined as -1, and if they are the same, the waveform similarity can be obtained according to the slope; Obtaining a ratio value based on the values ​​of the first data and the second data, such as ratio value=first data / second data; Based on the waveform similarity and the ratio value, the abnormal similarity is obtained. If different weights are assigned to them, the abnormal similarity is obtained after adding them.

[0054] Although the preferred embodiments of the present invention have been described, those skilled in the art may make other changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.

[0055] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A method for detecting abnormality of an electric knife motor, characterized in that: The method comprises: Acquire a plurality of motor vibration signal samples of the electric knife, extract the sample frequency domain features of each of the motor vibration signal samples, obtain a sample frequency domain feature matrix based on the sample frequency domain features, and obtain a plurality of first feature sets based on the sample frequency domain feature matrix, wherein the first feature sets all include a plurality of normalized sample frequency domain features; Obtain any two of the first feature sets to obtain a first set and a second set, obtain features of the first set and the second set respectively, and obtain a plurality of first features and a plurality of second features; Based on a first weight coefficient, obtaining a Jaccard distance between the first feature and the second feature, obtaining a plurality of first Jaccard distances, and obtaining a first distance threshold based on a maximum of the first Jaccard distances; Acquire a vibration signal to be detected of the electric knife to be detected, extract frequency domain features to be detected of the vibration signal to be detected, obtain a frequency domain feature matrix to be detected based on the frequency domain features to be detected, and obtain a plurality of third sets based on the frequency domain feature matrix to be detected, wherein the third sets all contain a plurality of normalized frequency domain features to be detected; Respectively acquiring features of the third set and the first feature set to obtain a plurality of third features and a plurality of fourth features; Based on the second weight coefficient, obtaining the Jaccard distance between the third feature and the fourth feature, obtaining a plurality of second Jaccard distances, and obtaining a first distance to be detected based on the maximum of the second Jaccard distances; Based on the first distance threshold and the first distance to be detected, a first detection result of the motor of the electric knife to be detected is obtained.

2. The method for detecting abnormality of an electric knife motor according to claim 1, characterized in that: The method further comprises: Extracting sample time domain features of each motor vibration signal sample, obtaining a sample time domain feature matrix based on the sample time domain features, and obtaining a plurality of second feature sets based on the sample time domain feature matrix, wherein the second feature sets each include a plurality of normalized sample time domain features; Obtain any two of the second feature sets, obtain a fourth set and a fifth set, obtain features of the fourth set and the fifth set respectively, and obtain a plurality of fifth features and a plurality of sixth features; Based on the third weight coefficient, obtaining the Jaccard distance between the fifth feature and the sixth feature, obtaining a plurality of third Jaccard distances, and obtaining a second distance threshold based on the maximum third Jaccard distance; Extracting the time domain features to be detected of the vibration signal to be detected, obtaining a time domain feature matrix to be detected based on the time domain features to be detected, and obtaining a plurality of sixth sets based on the time domain feature matrix to be detected, wherein the sixth sets each include a plurality of normalized time domain features to be detected; Acquire the features of the sixth set and the second feature set respectively, and obtain a plurality of seventh features and a plurality of eighth features; acquire the Jaccard distance between the seventh feature and the eighth feature based on a fourth weight coefficient, and obtain a plurality of fourth Jaccard distances; and acquire a second distance to be detected based on the maximum fourth Jaccard distance; Based on the second distance threshold and the second distance to be detected, obtaining a second detection result of the motor of the electric knife to be detected; Based on the first detection result and the second detection result, a third detection result of the motor is obtained, and the first detection result of the motor is updated to the third detection result of the motor.

3. The method for detecting abnormality of an electric knife motor according to claim 2, characterized in that: The first weight coefficient, the second weight coefficient, the third weight coefficient and the fourth weight coefficient are all obtained in the following manner: Obtaining a first mean of the feature, and obtaining a first variance based on the first mean; Obtaining a first covariance of any two features, and obtaining a first correlation coefficient based on the first covariance and the first variance; A weight coefficient of each feature is obtained based on the first correlation coefficient and the first variance.

4. The method for detecting abnormality of an electric knife motor according to claim 3, characterized in that: The specific steps of obtaining the sample frequency domain feature matrix include: Normalizing the sample frequency domain features, and obtaining the sample frequency domain feature matrix based on the normalized sample frequency domain features; The specific steps of obtaining the frequency domain feature matrix to be detected include: normalizing the frequency domain features to be detected, and obtaining the frequency domain feature matrix to be detected based on the normalized frequency domain features to be detected; The specific steps of obtaining the sample time domain feature matrix include: normalizing the sample time domain features, and obtaining the sample time domain feature matrix based on the normalized sample time domain features; The specific steps of obtaining the time domain feature matrix to be detected include: normalizing the time domain features to be detected, and obtaining the time domain feature matrix to be detected based on the normalized time domain features to be detected.

5. The method for detecting abnormality of an electric knife motor according to claim 4, characterized in that: The calculation formula for normalizing the sample frequency domain characteristics is: in, Indicates The normalized sample frequency domain features, Indicates The frequency domain features of samples before normalization, and Respectively The minimum and maximum values ​​of .

6. The method for detecting abnormality of an electric knife motor according to claim 5, characterized in that: The calculation formula of the first Jaccard distance is: in, express The first Jaccard distance between the first feature set A and the first feature set B in , Represents the sample frequency domain characteristics The first weight coefficient of The first feature set A represents Features, represents the first feature set B Features, represents the number of features in the first feature set, Representation characteristics With features The number of intersection elements, Representation characteristics With features The number of elements in the union of .

7. The method for detecting abnormality of an electric knife motor according to claim 6, characterized in that: The calculation formula of the first correlation coefficient is: in, Represents the sample frequency domain characteristics The first correlation coefficient of Represents the sample frequency domain characteristics And the sample frequency domain characteristics The covariance of and Represent the sample frequency domain characteristics And the sample frequency domain characteristics The mean square error, Indicates The normalized sample frequency domain features, Represents an integer greater than or equal to 1.

8. The method for detecting abnormality of an electric knife motor according to claim 7, characterized in that: The calculation formula of the first weight coefficient is: in, Represents the sample frequency domain characteristics The first weight coefficient of and Represent the sample frequency domain characteristics And the sample frequency domain characteristics The mean square error, Indicates the number of features in the first feature set.

9. The method for detecting abnormality of an electric knife motor according to claim 4, characterized in that: The method further comprises: Classifying the motor vibration signal samples to obtain a number of classified sample signals, and obtaining the category of the vibration signal to be detected; Based on the category, obtaining audio signals corresponding to the classified sample signal and the vibration signal to be detected, respectively obtaining a first audio signal and a second audio signal, and obtaining an abnormal audio signal based on the first audio signal and the second audio signal; Based on the second audio signal and the classified sample signal, respectively acquiring a first abnormal vibration signal corresponding to the abnormal audio signal and a second abnormal vibration signal corresponding to the classified sample signal; The first waveform graph and the second waveform graph of the first abnormal vibration signal and the second abnormal vibration signal are respectively obtained, the abnormal similarity of the first waveform graph and the second waveform graph is obtained, the abnormal value is obtained based on the abnormal similarity, and the usage time is obtained based on the abnormal value and the category.

10. The method for detecting abnormality of an electric knife motor according to claim 9, characterized in that: The specific steps of obtaining the abnormal similarity include: Acquire an extreme value of the second waveform graph, and adjust the first waveform graph based on the extreme value to obtain a third waveform graph; Based on a preset period, first data and second data of the second waveform graph and the third waveform graph are respectively obtained, waveform similarity is obtained based on the fluctuation directions of the first data and the second data, and a ratio value is obtained based on the values ​​of the first data and the second data; Based on the waveform similarity and the ratio value, the abnormality similarity is obtained.

Images