Fault monitoring and early warning method for operation state of non-circular gear emulsion motor

Abstract

The invention relates to the technical field of non-circular gear emulsion motor operation data processing, in particular to a non-circular gear emulsion motor operation state fault monitoring and early warning method. The method comprises the following steps: acquiring a sample data set, setting an initial penalty factor, and classifying samples in the sample data set to obtain outliers; calculating the importance degree of the vibration data in the outliers when the outliers are positive samples; calculating the importance degree of the vibration data in the outliers when the outliers are negative samples; calculating the importance degree of the current data in the outliers when the outliers are positive samples; calculating the importance degree of the current data in the outlier when the outlier is a negative sample; the importance degree of the outliers is obtained; obtaining an optimized penalty factor according to the importance degree of the outlier and the initial penalty factor; and obtaining an optimized SVM model based on the optimized penalty factor of each outlier, and monitoring the running state of the motor. According to the invention, the accuracy of monitoring the operation fault of the non-circular gear emulsion motor can be improved.

Description

Technical Field

[0001] The present invention relates to the technical field of operation data processing of non-circular gear emulsion motors, and specifically relates to a method for fault monitoring and early warning of the operation state of non-circular gear emulsion motors. Background Art

[0002] The core component of a non-circular gear emulsion motor is a non-circular gear train, which usually consists of a non-circular plum-shaped internal gear ring, a non-circular square sun gear, and planetary gears. Its working principle mainly utilizes the mechanical properties of the liquid, and drives the equipment to operate by changing the pressure difference and flow direction of the liquid. Specifically, when the emulsion enters the motor, through the action of the rotor and stator, the flow direction and speed of the liquid change, thereby generating a pressure difference. This pressure difference drives the rotation of the rotor, and then drives the equipment to operate.

[0003] Non-circular gear emulsion motors are widely used in flammable and explosive environments such as underground coal mines because they use a non-flammable liquid - high water-based emulsion as the working medium. Once a failure occurs, it will cause a series of safety problems. Moreover, as a key device, the operation state of the motor directly affects the production efficiency. Therefore, it is necessary to detect and early warn of the failure of the motor operation state to improve the safety and reliability of the equipment.

[0004] Traditionally, by monitoring the data generated during the operation of a non-circular gear emulsion motor and judging whether the data exceeds the threshold for early warning. However, due to the differences in the working environment and failure types, the method of judging whether it exceeds the threshold is not accurate enough for early warning of operation data with a relatively complex linear relationship. Also, because the support vector machine (SVM) can handle complex non-linear problems and is good at processing various sources of data (such as vibration data, current data, etc.) collected during motor fault monitoring, the support vector machine (SVM) is used for fault monitoring and early warning. However, since the support vector machine is affected by the size of the penalty factor during anomaly recognition, if the penalty factor is too large, it will lead to overfitting, making the generalization ability of the model poor; if the penalty factor is too small, it may cause the model to ignore some important classification information during training; both situations will affect the anomaly classification result of the current vibration data. Summary of the Invention

[0005] In order to solve the above technical problems, the purpose of the present invention is to provide a method for fault monitoring and early warning of the operation state of a non-circular gear emulsion motor, and the specific technical solution adopted is as follows: An embodiment of the present invention provides a method for fault monitoring and early warning of the operation state of a non-circular gear emulsion motor, and the method includes: Positive and negative samples are obtained based on the current data and vibration data during the operation of the motor, and a sample data set is formed; the current sequence and vibration sequence in which each current data and vibration data are located are determined; an initial penalty factor is set and the samples in the sample data set are classified to obtain outliers; If the outlier is a positive sample, calculate the importance degree of the vibration data according to the outliers, coefficient of variation in the vibration sequence where the vibration data in the outlier is located, and the data before and after the vibration data; if the outlier is a negative sample, calculate the importance degree of the vibration data according to the outliers and spectral entropy in the vibration sequence where the vibration data in the outlier is located; If the outlier is a positive sample, calculate the importance degree of the current data according to the information entropy of the current sequence where the current data in the outlier is located and the data before and after the current data; if the outlier is a negative sample, calculate the importance degree of the current data according to the outliers in the current sequence where the current data in the outlier is located and the data before and after the current data; Calculate the importance degree of the outlier according to the importance degrees of the vibration data and current data in an outlier; obtain the optimized penalty factor of the outlier according to the importance degree of the outlier and the initial penalty factor; obtain an optimized SVM model based on the optimized penalty factors of each outlier, and monitor the operating state of the motor.

[0006] Preferably, obtaining positive and negative samples based on the current data and vibration data during the operation of the motor and forming a sample data set includes: Preprocess the current data and vibration data, including data cleaning and standardization processing; form a feature vector with the current data and vibration data at a moment after preprocessing, one feature vector is one sample, the sample in the normal operating state of the motor is a positive sample, and the sample in the abnormal operating state of the motor is a negative sample; divide the positive samples and negative samples into a training set and a test set, and the test set and the training set form a sample data set.

[0007] Preferably, setting an initial penalty factor and classifying the samples in the sample data set to obtain outliers includes: Train an initial SVM model using the sample data set according to the set initial penalty factor, and classify the sample data set using the initial SVM model to obtain outliers.

[0008] Preferably, calculating the importance degree of the vibration data according to the outliers, coefficient of variation in the vibration sequence where the vibration data in the outlier is located, and the data before and after the vibration data includes: Obtain the normalized value of the ratio of the number of outliers after box plot screening for all vibration sequences to the number of outliers after box plot screening in the vibration sequence where the vibration data in an outlier is located, denoted as the non-outlier proportion; obtain the normalized value of the ratio of the sum of the coefficient of variation of all vibration sequences to the coefficient of variation of the vibration sequence where the vibration data in this outlier is located, denoted as the stability degree; obtain the difference between the vibration data in this outlier and the mean of the data before this vibration data in the vibration sequence where this vibration data is located, and take the absolute value to obtain the first difference; obtain the difference between the vibration data in this outlier and the mean of the data after this vibration data in the vibration sequence where this vibration data is located, and take the absolute value to obtain the second difference; obtain the absolute value of the difference between the first difference and the second difference, and normalize it to obtain the possible degree of the turning point; perform weighted summation on the non-outlier proportion, stability degree, and possible degree of the turning point to obtain the importance degree of this vibration data.

[0009] Preferably, calculate the importance degree of the vibration data according to the outliers and spectral entropy in the vibration sequence where the vibration data in the outlier is located, including: Obtain the normalized value of the ratio of the number of outliers after box plot screening in the vibration sequence where the vibration data in an outlier is located to the number of outliers after box plot screening for all vibration sequences, denoted as the outlier proportion; obtain the ratio of the spectral entropy of the vibration sequence where the vibration data in an outlier is located to the sum of the spectral entropy of all vibration sequences, and normalize it to obtain the outlier degree; perform weighted summation on the outlier proportion and outlier degree to obtain the importance degree of this vibration data.

[0010] Preferably, calculate the importance degree of the current data according to the information entropy of the current sequence where the current data in the outlier is located and the data before and after this current data, including: Compare the sum of the information entropy of all current sequences with the information entropy of the current sequence where the current data in an outlier is located and normalize it to obtain the stability degree of the current sequence; calculate the slopes between this current data and the previous adjacent current data and the next adjacent current data respectively, denoted as the first slope and the second slope; take the reciprocal of the absolute value of the difference between the first slope and the second slope and normalize it to obtain the slope change situation; perform weighted summation on the stability degree of the current sequence and the slope change situation to obtain the importance degree of this current data.

[0011] Preferably, calculate the importance degree of the current data according to the outliers in the current sequence where the current data in the outlier is located and the data before and after this current data, including: Obtain the normalized value of the ratio of the number of outliers screened by the box plot in the current sequence where the current data in an outlier is located to the number of outliers screened by the box plot in all current sequences, denoted as the abnormal current ratio; normalize the absolute value of the difference between the first slope and the second slope to obtain the slope difference; perform weighted summation on the abnormal current ratio and the slope difference to obtain the importance degree of the current data.

[0012] Preferably, calculate the importance degree of an outlier according to the importance degrees of the vibration data and the current data in the outlier, including: For a discrete point that is a negative sample, perform weighted summation on the importance degrees of the vibration data and the current data in the discrete point obtained under the negative sample to obtain the importance degree of the discrete point; for a discrete point that is a positive sample, perform weighted summation on the importance degrees of the vibration data and the current data in the discrete point obtained under the positive sample to obtain the importance degree of the discrete point.

[0013] Preferably, obtain the optimized penalty factor of the outlier according to the importance degree of the outlier and the initial penalty factor, including: Add the importance degree of the discrete point to the initial penalty factor to obtain the optimized penalty factor of the discrete point.

[0014] Preferably, obtain an optimized SVM model based on the optimized penalty factors of each outlier and monitor the operating state of the motor, including: Use each discrete point, its corresponding optimized penalty factor, and other samples in the sample dataset to train the initial SVM model to obtain an optimized SVM model; use the optimized SVM model to classify the operating data of the non-circular gear emulsion motor and monitor the operating state of the motor.

[0015] The embodiments of the present invention have at least the following beneficial effects: This application will set an initial penalty factor, use the initial penalty factor combined with the initial SVM model to classify the samples in the sample dataset to obtain outliers; further, on the premise that the outlier is a positive sample or a negative sample, calculate the importance degrees of the current data and the vibration data in the outlier, and then comprehensively combine the importance degrees of the current data and the vibration data to obtain the importance degree of the discrete point, use the importance degree of each outlier to optimize the initial penalty factor to obtain the optimized penalty factor of the discrete point, and then substitute the optimized penalty factor into the original optimization problem, and use the sample dataset for training to obtain a more accurate optimized SVM model, improving the accuracy of fault monitoring of the operating state of the non-circular gear emulsion motor. Description of the Drawings

[0016] To more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0017] Figure 1 It is a method flow chart of a fault monitoring and early warning method for the operating state of a non-circular gear emulsion motor provided by an embodiment of the present invention; Figure 2 It is a fault monitoring and early warning method for the operating state of a non-circular gear emulsion motor provided by an embodiment of the present invention. Detailed implementation manners

[0018] In order to further elaborate on the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the following combines the drawings and preferred embodiments to detail the specific implementation manners, structures, features, and effects of a fault monitoring and early warning method for the operating state of a non-circular gear emulsion motor proposed according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. In addition, the specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0019] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the technical field to which the present invention belongs.

[0020] The following specifically describes the specific solution of a fault monitoring and early warning method for the operating state of a non-circular gear emulsion motor provided by the present invention with reference to the drawings.

[0021] Embodiment: The main application scenario of the present invention is: collecting the operating data during the operation of a non-circular gear emulsion motor, including current data and vibration data, then analyzing the operating data to obtain abnormal operating conditions, and further achieving the purpose of fault monitoring and early warning for the operation of the non-circular gear emulsion motor.

[0022] Please refer to Figure 1 , which shows a method flow chart of a fault monitoring and early warning method for the operating state of a non-circular gear emulsion motor provided by an embodiment of the present invention. The method includes the following steps: Step S1, obtaining positive samples and negative samples based on the current data and vibration data during the motor operation, and forming a sample data set; determining the current sequence and vibration sequence to which each current data and vibration data belong; setting an initial penalty factor and classifying the samples in the sample data set to obtain outliers.

[0023] During the operation of the non-circular gear emulsion motor, the current data can reflect the load condition of the motor. Too high or too low current may indicate problems with the motor. The change in vibration data can provide a warning at the initial stage of a fault and offer more specific fault information. Therefore, the current data and vibration data of the non-circular gear emulsion motor during operation are collected in this application. Since the operation of the non-circular gear emulsion motor is at irregular time intervals, the current data and vibration data are collected during its operation period. The current data of one operation period form a current sequence, and the vibration data of one operation period form a voltage sequence. The current data and vibration data are obtained according to the operation period in which they are located.

[0024] Furthermore, preprocessing of the current data and vibration data is required, mainly including data cleaning and normalization. Since the data generated during the current motor operation needs to be classified for anomalies using SVM, the historical emulsion motor operation data needs to be cleaned first, including removing outliers, filling in missing values, and denoising. After data cleaning, the current data and vibration data need to be normalized to ensure that each feature has the same dimension. At the same time, since it is necessary to perform an inverse mapping on the vectors in the feature space to obtain the corresponding positions of the two features in the original data sequence, the timestamps corresponding to the vibration data and current data before mapping to the feature space need to be retained at this time.

[0025] Furthermore, the current data and vibration data at a certain moment after preprocessing are combined into a feature vector. A feature vector is a sample. Samples in the normal operating state of the motor are positive samples, and samples in the abnormal operating state of the motor are negative samples. It should be noted that when dividing positive and negative samples, manual division is required to divide them into positive samples and negative samples. The positive samples and negative samples are divided into a training set and a test set. The test set and the training set form a sample data set. The method of dividing into a training set and a test set is the cross-validation method. The test set is used to test the performance of the model. Common cross-validation methods include K-fold cross-validation and leave-one-out cross-validation.

[0026] Then, it is also necessary to select an appropriate kernel function to handle the non-linear relationship between the positive and negative samples of the sample data set to make the data more separable, such as the radial basis function (RBF), sigmoid kernel function, or polynomial kernel function to reduce the computational complexity.

[0027] In an ideal situation, when using a support vector machine (SVM) for classification, the motor operation training data will be divided into normal data and fault data. However, due to the non - linear and dynamic change characteristics and noise during the operation of the non - circular gear emulsion motor, there will be outliers in the initial classification model. Since the penalty factor represents the tolerance for outliers, the larger the penalty factor, the less willing to discard outliers. Therefore, it is necessary to evaluate the importance of outliers so that each outlier corresponds to a corresponding penalty factor. At this time, the penalty factors corresponding to each outlier are brought into the model to train the training data set, and then a more accurate classification model can be obtained, thus realizing the fault assessment of the current motor operation state.

[0028] Therefore, it is necessary to set an initial penalty factor and classify the samples in the sample data set to obtain outliers. Specifically, an initial SVM model is obtained by training the sample data set according to the set initial penalty factor, and the sample data set is classified using the initial SVM model to obtain outliers. As Figure 2 shown, H is the hyperplane, and are support vectors, with equal distances from the hyperplane. The squares represent the positive samples of the motor operation data in the feature space, that is, the normal state of the motor operation; the circles represent the negative samples of the motor operation data in the feature space, that is, the fault state of the motor operation; the squares and circles on the two straight lines are the closest to the hyperplane, representing support vectors; the squares within the circular area are outliers, and the outliers are recorded and corresponding to the initial vibration data and current data. It should be noted that after model training, a hyperplane is obtained, and then the data is classified.

[0029] Step S2: If the outlier is a positive sample, calculate the importance of the vibration data according to the outliers, coefficient of variation, and the data before and after the vibration data in the vibration sequence where the vibration data of the outlier is located; if the outlier is a negative sample, calculate the importance of the vibration data according to the outliers and spectral entropy in the vibration sequence where the vibration data of the outlier is located.

[0030] Since the penalty factor represents the tolerance for misclassifying outliers, and whether to attach importance to the classification results of these outliers needs to be judged according to the characteristics of the outliers themselves, that is, it is necessary to calculate the importance of the outliers at this time. The higher the importance, the more attention should be paid to the classification results of the outliers, and the corresponding penalty factor should be larger at this time; since the outliers are composed of two characteristics, namely vibration data and current data, at this time, the corresponding importance should be obtained according to the data characteristics of the outliers in these two types of data, and then the comprehensive importance of the current outliers can be obtained.

[0031] Since the criteria for judging importance are different due to the different positive and negative categories to which the outliers belong, it is necessary to calculate separately according to the positive and negative sample categories to which the outliers belong.

[0032] If the outlier is a positive sample, calculate the importance degree of the vibration data according to the outliers, coefficient of variation in the vibration sequence where the vibration data in the outlier is located, and the data before and after the vibration data. Assume there are vibration sequences in total, and the vibration data in the i-th outlier is in the -th vibration sequence.

[0033] Specifically, obtain the normalized value of the ratio of the number of outliers screened by the box plot in all vibration sequences to the number of outliers screened by the box plot in the vibration sequence where the vibration data in an outlier is located, denoted as the non-outlier proportion; obtain the normalized value of the ratio of the sum of the coefficients of variation of all vibration sequences to the coefficient of variation of the vibration sequence where the vibration data in the outlier is located, denoted as the stability degree; obtain the difference between the vibration data in the outlier and the mean of the data before the vibration data in the vibration sequence where the vibration data is located, and take the absolute value to get the first difference; obtain the difference between the vibration data in the outlier and the mean of the data after the vibration data in the vibration sequence where the vibration data is located, and take the absolute value to get the second difference; obtain the absolute value of the difference between the first difference and the second difference, and normalize it to get the possible degree of the turning point; perform a weighted sum on the non-outlier proportion, stability degree, and possible degree of the turning point to obtain the importance degree of the vibration data.

[0034] The specific calculation formula is:

[0035] Among them, represents the importance degree in the corresponding vibration sequence when the i-th outlier is a positive sample, that is, the importance degree of the vibration data in the i-th outlier; represents the number of outliers screened by the box plot in the vibration sequence where the vibration data in the i-th outlier is located when the i-th outlier is a positive sample, is the number of outliers screened by the box plot in all vibration sequences; is the coefficient of variation of the j-th vibration sequence; represents the value of the coefficient of variation in the -th vibration sequence segment where the vibration data in the i-th outlier is located; is the value of the vibration data in the i-th outlier, represents the mean of the vibration data before the vibration data in the i-th outlier in the vibration sequence where the vibration data in the i-th outlier is located, is the mean of the vibration data after the vibration data in the i-th outlier in the vibration sequence where the vibration data in the i-th outlier is located, and norm represents the normalization operation.

[0036] is the non - abnormal proportion, The larger the value of is, the smaller the number of abnormal points in the vibration sequence where the i - th outlier is located, indicating that the stability degree of the vibration sequence where the i - th outlier is located is higher. At this time, a relatively large importance should be given to this point; is the total value of the coefficient of variation of all vibration sequences, is the stability degree, The larger is, the lower the stability degree in the vibration sequence segment where the i - th outlier is located, that is, the higher the stability degree of the vibration data in the vibration sequence segment where the i - th outlier is located. At this time, a relatively large importance should be given to this point, is the absolute value of the difference between the value of the vibration data in the i - th outlier and the mean value of the previous vibration data in its vibration sequence, is the absolute value of the difference between the value of the vibration data in the i - th outlier and the mean value of the subsequent vibration data in its vibration sequence, is the absolute value of the difference between the two absolute values. This term is because the outlier is a positive sample. If the current vibration data continues to decline or rise after a turning point at the current point when it is a positive sample, it means that the current point is an important turning point. Therefore, the larger the third term in the formula, the more significant the situation that the vibration data in the outlier is a positive sample and is a turning point in the vibration sequence, and the greater the corresponding importance at this time, is the possibility degree of the turning point. The larger this value is, the higher the possibility that the vibration data in the outlier as a positive sample is a turning point in its vibration sequence. Also, because the third term has a greater impact on the importance as the discrimination criterion for the turning point, a weight is given, .

[0037] Furthermore, when the outlier is a negative sample, the importance of the vibration data is calculated according to the abnormal points and spectral entropy in the vibration sequence where the vibration data in the outlier is located.

[0038] Specifically, obtain the normalized value of the ratio of the number of abnormal points screened by the box - plot in the vibration sequence where the vibration data in an outlier is located to the number of abnormal points screened by the box - plot in all vibration sequences, denoted as the abnormal proportion; obtain the ratio of the spectral entropy of the vibration sequence where the vibration data in an outlier is located to the sum of the spectral entropies of all vibration sequences, and normalize it to obtain the abnormal degree; perform weighted summation on the abnormal proportion and the abnormal degree to obtain the importance of the vibration data.

[0039] Its calculation formula is:

[0040] Among them, is the importance degree of the $i$-th outlier as a negative sample in the corresponding vibration data sequence, that is, when the $i$-th outlier is a negative sample, it is the importance degree of the vibration data in the $i$-th outlier; denotes the ratio of the number of outliers screened by the box plot in the $i$-th vibration sequence where the $i$-th outlier is located to the number of outliers in all vibration sequences, which represents the outlier proportion. The larger this value is, the more outliers there are in the vibration sequence segment where the $i$-th outlier is located, indicating a higher degree of abnormality in the vibration sequence where the $i$-th outlier is used as a negative sample. At this time, a greater importance degree should be given to this point; is the spectral entropy of the vibration data in the $i$-th outlier in the -th vibration sequence, is the spectral entropy of the $j$-th vibration sequence, is the sum of the spectral entropies of all vibration sequences, represents the degree of abnormality. The larger this value is, the greater the degree of abnormality of the vibration sequence where the $i$-th outlier is located. At this time, a greater importance degree should be given to this point. Since the first term in the above formula has a greater impact on the importance degree calculation, a weight is given, .

[0041] Thus, the importance degree of the vibration data in the outlier when the outlier is a positive sample or a negative sample can be calculated.

[0042] Step S3: If the outlier is a positive sample, calculate the importance degree of the current data according to the information entropy of the current sequence where the current data in the outlier is located and the data before and after the current data; if the outlier is a negative sample, calculate the importance degree of the current data according to the outliers in the current sequence where the current data in the outlier is located and the data before and after the current data.

[0043] After calculating the importance degree of the vibration data in the outlier, it is also necessary to analyze the current data in the outlier. Assume that there are current sequences in total, and the current data of the $i$-th outlier is in the -th current sequence. Similarly, the calculation method of the importance degree of the current data is different when the outlier is a positive sample or a negative sample.

[0044] If the outlier is a positive sample, calculate the importance degree of the current data according to the information entropy of the current sequence where the current data in the outlier is located and the data before and after the current data.

[0045] Specifically, the sum of the information entropy of all current sequences is compared with the information entropy of the current sequence where the current data in an outlier is located and normalized to obtain the stability degree of the current sequence. The slopes between the current data and the previous adjacent current data and the next adjacent current data are calculated respectively, denoted as the first slope and the second slope. The reciprocal of the absolute value of the difference between the first slope and the second slope is normalized to obtain the slope change situation. The stability degree of the current sequence and the slope change situation are weighted and summed to obtain the importance degree of the current data.

[0046] Its calculation formula is: , where, represents the importance degree of the current data in the discrete point when the i-th discrete point is a positive sample; represents the information entropy of the j-th current sequence; represents the information entropy of the current sequence where the current data in the i-th discrete point is located; represents the sum of the information entropy of all current sequences, represents the stability degree of the current sequence, which is the normalized result of the ratio of the sum of the information entropy of all current sequences to the information entropy of the current sequence where the current data in the outlier is located. The larger this value is, the higher the stability degree of the current sequence where the current data in the current outlier is located. At this time, the outlier as a positive sample should be given a greater importance degree; represents the slope between the current data in the i-th outlier and the previous adjacent current data, that is, the first slope, represents the slope between the current data in the i-th outlier and the next adjacent current data, that is, the second slope, is the slope change situation, which represents the normalized result of the reciprocal of the absolute value of the difference between the two slopes. The larger this value is, that is, the smaller the denominator of the second term in the formula, the smaller the slope difference between the current data as a positive sample and the adjacent current data on the left and right, indicating the greater the stability degree of this point, and a greater importance degree should be given to this point, is the weight.

[0047] Finally, calculate the importance degree of the current data in the discrete point when it is a negative sample, and calculate the importance degree of the current data according to the outliers in the current sequence where the current data in the outlier is located and the data before and after the current data.

[0048] Specifically, obtain the normalized value of the ratio of the number of outliers screened by the box plot in the current sequence where the current data in an outlier is located to the number of outliers screened by the box plot in all current sequences, denoted as the current anomaly ratio; normalize the absolute value of the difference between the first slope and the second slope to obtain the slope difference; the current anomaly ratio and the slope difference are weighted and summed to obtain the importance degree of the current data.

[0049] Its calculation formula is as follows: , wherein, represents the importance degree of the current data in the discrete points when the i-th discrete point is a negative sample; represents the number of outliers screened by the box plot in the current sequence where the current data in the i-th outlier is located when the i-th outlier is a negative sample, is the number of outliers screened by the box plot for all current sequences; represents the slope of the current data in the i-th outlier and the previous adjacent current data, that is, the first slope, represents the slope of the current data in the i-th outlier and the next adjacent current data, that is, the second slope.

[0050] is the current anomaly ratio. The larger this value is, the more outliers there are in the current sequence where the current data in the i-th outlier is located, indicating that the anomaly degree of the current sequence where the i-th outlier is a negative sample is higher. At this time, a greater importance degree should be given to this point; represents the slope difference, which represents the slope difference between the current data in the i-th outlier and the current data of the previous and next two current data. The larger this value is, the higher the anomaly degree of the current point, and a greater importance degree should be given to the current outlier.

[0051] Thus, the importance degree of the current data in the discrete points when the discrete points are positive samples or negative samples is obtained.

[0052] Step S4: Calculate the importance degree of the outlier according to the importance degrees of the vibration data and the current data in an outlier; obtain the optimized penalty factor of the outlier according to the importance degree of the outlier and the initial penalty factor; obtain an optimized SVM model based on the optimized penalty factors of each outlier, and monitor the operating state of the non-circular gear emulsion motor.

[0053] In steps S2 and S3, the importance degrees of the vibration data and the current data therein are obtained respectively when the discrete points are positive samples or negative samples. Further, it is necessary to comprehensively calculate the importance degree of the discrete points according to the importance degrees of the vibration data and the current data. When calculating, it is also necessary to divide the discrete points into positive samples or negative samples for calculation.

[0054] For the discrete points that are negative samples, the importance degrees of the vibration data and the current data in the discrete points obtained under the negative samples are weighted and summed to obtain the importance degree of the discrete points; for the discrete points that are positive samples, the importance degrees of the vibration data and the current data in the discrete points obtained under the positive samples are weighted and summed to obtain the importance degree of the discrete points.

[0055] The specific calculation formula is as follows: , , wherein, and are the importance degrees when the discrete point is a positive sample and a negative sample respectively; and are the importance degrees of the vibration data and current data when the discrete point is a positive sample; and are the importance degrees of the vibration data and current data when the discrete point is a negative sample.

[0056] Thus, the importance degree of each discrete point can be obtained.

[0057] Since the penalty factor represents the tolerance for misclassifying outliers, the higher the importance degree of an outlier, the more attention is paid to the classification result of this point, and the more we hope that this point is correctly classified. Therefore, a larger penalty factor should be corresponding to the outlier with a high importance degree to ensure that this point is classified correctly as much as possible.

[0058] The optimized penalty factor of the outlier is obtained according to the importance degree of the outlier and the initial penalty factor. Specifically, the importance degree of the discrete point is added to the initial penalty factor to obtain the optimized penalty factor of the discrete point. For the discrete point that is a positive sample, the importance degree of the discrete point calculated under the positive sample is added to the initial penalty factor to obtain the optimized penalty factor. For the discrete point that is a negative sample, the importance degree of the discrete point calculated under the negative sample is added to the initial penalty factor to obtain the optimized penalty factor.

[0059] Before obtaining the optimized penalty factor, it is necessary to process the importance degree of the discrete point for convenient subsequent use. The importance degree of the discrete point is reserved to one decimal place and then added to the initial penalty factor.

[0060] After obtaining the optimized penalty factor, the optimized penalty factor is brought into the original optimization problem, and the training model is used to train the operation data, so as to obtain a more accurate classification model. Specifically, the initial SVM model is trained using each discrete point, its corresponding optimized penalty factor and other samples in the sample dataset to obtain an optimized SVM model; the optimized SVM model is used to classify the operation data of the non-circular gear emulsion motor to monitor the operation state of the motor. If abnormal operation data is found, a warning is given to remind the staff to carry out maintenance.

[0061] This solution first classifies the motor operation training data initially, obtains the outliers according to the initial classification results, then calculates the importance of the outliers corresponding to the respective data sequences, adjusts the penalty factor according to the importance, and then uses the adjusted training model to train the operation data, thereby obtaining a more accurate classification model, so as to achieve more accurate anomaly recognition of the current motor operation data. This method of adaptively adjusting the penalty factor according to the importance of the outliers is more flexible than the traditional SVM classification and is more applicable to the unique environment of the non-circular gear emulsion motor operation, improving the accuracy of fault monitoring.

[0062] It should be noted that the above sequence of the embodiments of the present invention is only for description and does not represent the superiority or inferiority of the embodiments. And the above describes specific embodiments of this specification. In addition, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0063] The various embodiments in this specification are all described in a progressive manner. The same or similar parts among the various embodiments can be referred to each other, and the key points of each embodiment are the differences from other embodiments.

[0064] The above are only the preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A fault monitoring and early warning method for the operating state of a non-circular gear emulsion motor, characterized in that The method includes: Obtaining positive samples and negative samples based on the current data and vibration data during the operation of the motor, and forming a sample data set; determining the current sequence and vibration sequence to which each current data and vibration data belongs; setting an initial penalty factor and classifying the samples in the sample data set to obtain outliers; If the outlier is a positive sample, calculating the importance degree of the vibration data according to the outliers, coefficient of variation in the vibration sequence where the vibration data in the outlier is located, and the data before and after the vibration data; if the outlier is a negative sample, calculating the importance degree of the vibration data according to the outliers and spectral entropy in the vibration sequence where the vibration data in the outlier is located; If the outlier is a positive sample, calculating the importance degree of the current data according to the information entropy of the current sequence where the current data in the outlier is located and the data before and after the current data; if the outlier is a negative sample, calculating the importance degree of the current data according to the outliers in the current sequence where the current data in the outlier is located and the data before and after the current data; Calculating the importance degree of the outlier according to the importance degrees of the vibration data and current data in an outlier; obtaining the optimized penalty factor of the outlier according to the importance degree of the outlier and the initial penalty factor; obtaining an optimized SVM model based on the optimized penalty factors of each outlier, and monitoring the operating state of the motor.

2. The fault monitoring and early warning method for the operating state of a non-circular gear emulsified liquid motor according to claim 1, characterized in that, The obtaining positive samples and negative samples based on the current data and vibration data during the operation of the motor, and forming a sample data set includes: Performing preprocessing on the current data and vibration data, including data cleaning and normalization processing; forming a feature vector with the current data and vibration data at a moment after preprocessing, a feature vector being a sample, a sample in the normal operating state of the motor being a positive sample, and a sample in the abnormal operating state of the motor being a negative sample; dividing the positive samples and negative samples into a training set and a test set, and the test set and the training set form a sample data set.

3. The fault monitoring and early warning method for the operating state of a non-circular gear emulsified liquid motor according to claim 1, characterized in that, The setting an initial penalty factor and classifying the samples in the sample data set to obtain outliers includes: Training an initial SVM model using the sample data set according to the set initial penalty factor, and classifying the sample data set using the initial SVM model to obtain outliers.

4. The fault monitoring and early warning method for the operating state of a non-circular gear emulsion motor according to claim 1, characterized in that, The calculating the importance degree of the vibration data according to the outliers, coefficient of variation in the vibration sequence where the vibration data in the outlier is located, and the data before and after the vibration data includes: Obtain the normalized value of the ratio of the number of outliers after box plot screening in all vibration sequences to the number of outliers after box plot screening in the vibration sequence where the vibration data in an outlier is located, denoted as the non-outlier proportion; obtain the normalized value of the ratio of the sum of the coefficient of variation of all vibration sequences to the coefficient of variation of the vibration sequence where the vibration data in this outlier is located, denoted as the stability degree; obtain the difference between the vibration data in this outlier and the mean of the data before this vibration data in the vibration sequence where this vibration data is located, and take the absolute value to get the first difference; obtain the difference between the vibration data in this outlier and the mean of the data after this vibration data in the vibration sequence where this vibration data is located, and take the absolute value to get the second difference; obtain the absolute value of the difference between the first difference and the second difference, and normalize it to get the possible degree of the turning point; perform weighted summation on the non-outlier proportion, the stability degree, and the possible degree of the turning point to get the importance degree of this vibration data.

5. The fault monitoring and early warning method for the operating state of a non-circular gear emulsion motor according to claim 1, characterized in that, Calculating the importance degree of the vibration data according to the outliers and spectral entropy in the vibration sequence where the vibration data in the outlier is located includes: Obtain the normalized value of the ratio of the number of outliers after box plot screening in the vibration sequence where the vibration data in an outlier is located to the number of outliers after box plot screening in all vibration sequences, denoted as the outlier proportion; obtain the ratio of the spectral entropy of the vibration sequence where the vibration data in an outlier is located to the sum of the spectral entropy of all vibration sequences, and normalize it to get the outlier degree; perform weighted summation on the outlier proportion and the outlier degree to get the importance degree of this vibration data.

6. The fault monitoring and early warning method for the operating state of a non-circular gear emulsified liquid motor according to claim 1, wherein, Calculating the importance degree of the current data according to the information entropy of the current sequence where the current data in the outlier is located and the data before and after this current data includes: Compare the sum of the information entropy of all current sequences with the information entropy of the current sequence where the current data in an outlier is located and normalize it to get the stability degree of the current sequence; calculate the slopes between this current data and the previous adjacent current data and the next adjacent current data respectively, denoted as the first slope and the second slope; take the reciprocal of the absolute value of the difference between the first slope and the second slope and normalize it to get the slope change situation; perform weighted summation on the stability degree of the current sequence and the slope change situation to get the importance degree of this current data.

7. A fault monitoring and warning method for the operating state of a non-circular gear emulsion motor according to claim 1, characterized in that, Calculating the importance degree of the current data according to the outliers in the current sequence where the current data in the outlier is located and the data before and after the current data includes: Obtain the normalized value of the ratio of the number of outliers after box plot screening in the current sequence where the current data in an outlier is located to the number of outliers after box plot screening in all current sequences, denoted as the current outlier proportion; normalize the absolute value of the difference between the first slope and the second slope to get the slope difference; perform weighted summation on the current outlier proportion and the slope difference to get the importance degree of this current data.

8. A fault monitoring and early warning method for the operating state of a non-circular gear emulsified liquid motor according to claim 1, characterized in that, Calculating the importance degree of the outlier according to the importance degrees of the vibration data and the current data in an outlier includes: For the discrete points that are negative samples, the importance levels of the vibration data and current data in the discrete points obtained under the negative samples are weighted and summed to obtain the importance level of the discrete points; for the discrete points that are positive samples, the importance levels of the vibration data and current data in the discrete points obtained under the positive samples are weighted and summed to obtain the importance level of the discrete points.

9. The fault monitoring and early warning method for the operating state of a non-circular gear emulsion motor according to claim 1, characterized in that, The obtaining of the optimized penalty factor for the outlier according to the importance level of the outlier and the initial penalty factor includes: Adding the importance level of the discrete point to the initial penalty factor to obtain the optimized penalty factor of the discrete point.

10. The fault monitoring and early warning method for the operating state of a non-circular gear emulsified liquid motor according to claim 1, characterized in that The obtaining of the optimized SVM model based on the optimized penalty factors of the respective outliers and the monitoring of the operating state of the motor include: Training the initial SVM model by using each discrete point, its corresponding optimized penalty factor, and other samples in the sample dataset to obtain the optimized SVM model; classifying the operating data of the non-circular gear emulsion motor by using the optimized SVM model to monitor the operating state of the motor.

Images